The shock absorbers of biology
Kevin Verstrepen and his team on buffer proteins and the genetic variation they mask
A recent Nature news feature dives into how buffer proteins can protect us from harmful mutations and how that could inspire new drugs. We asked Kevin Verstrepen, as one of the experts quoted in the piece, and team members Mo Tawfeeq, Karin Voordeckers, Hala Kasmo, and Iris Goeminne, to tell us more about what buffering really is and why it matters for disease and evolution.
Let’s start with the concept of buffering itself. What do scientists mean when they talk about “buffering”?
Kevin Verstrepen: “Well, first off, they don’t always mean exactly the same thing. Some studies make it sound like buffering is something very special, with only a few genes being true buffers, capable of masking the effects of genetic mutations. As we elaborated in a recent review,1 we see that differently. We consider buffering as a special and strong case of genetic interactions.”
Mohammed Tawfeeq: “Genetic interaction is the phenomenon where the combination of two changes in DNA (i.e. 2 mutations) has a different phenotypic effect than what would be expected from what happens when each change (mutation) happens separately. With negative interactions, the combined effect is larger than expected, while the opposite is true for positive interactions. Genetic interactions are very common.
Buffering is then the case where changing the activity of one gene (mutation 1) influences the phenotypic outcome of an exceptionally broad subset of mutations.”
"The exact definition of when we can call a gene a buffer gene is not set in stone—and that is causing quite a bit of confusion and discussion in the field."
Karin Voordeckers: “Translated to human disease, this would mean that if someone misses a buffer gene, or has a buffer gene with lower activity, that person may be more susceptible to the effects of new, random, or already present mutations. Random mutations happen constantly in our cells, and even more so when cells are exposed to carcinogens such as UV light, alcohol, and cigarette smoke. If these random mutations affect a crucial pathway such as cell division, cancers and other diseases can emerge. Inactivating a buffer gene could thus in principle increase the chance of disease and cancer.”
Kevin Verstrepen: “Importantly, there is no clear definition of the magnitude of genetic interactions that a gene must have before we call it a buffer gene; it is a bit of an arbitrary cutoff. In other words: the exact definition of when we can call a gene a buffer gene is not set in stone—and that is causing quite a bit of confusion and discussion in the field.”
Yet, HSP90 is the clear poster child?
Iris Goeminne: “Yes. One of the other key limitations of what we currently know about buffering is that almost all of our knowledge is based on studies that focus on one specific buffer gene, Hsp90. Since Hsp90 is a chaperone that helps proteins fold into their active conformation, this also led to perhaps a bit of tunnel vision. Specifically, most studies have looked at how Hsp90 can buffer mutations by stabilizing mutant proteins [recent examples of biomedical relevance include the following papers2,3]. However, the precise mutations whose effects are modified by Hsp90 have remained mostly obscure.
The lack of a clear definition of buffering and the focus on only one gene, Hsp90, implies that we currently do not know the answer to several key questions about buffer genes: how general is the phenomenon of buffering? Is it specific to a limited set of proteins, or is it common to many proteins? In other words, how unique is Hsp90? Are some buffer genes more important in specific conditions than others? What type of mutations are buffered, and what are the exact mechanisms?”
.jpg)
How is the Verstrepen lab contributing to finding those answers?
Kevin Verstrepen: “We recently performed a genome-wide screen for genes that can act as genetic buffers, by combining the genetic toolbox of the model eukaryote Saccharomyces cerevisiae with high-throughput phenotyping.4 Measuring the fitness of 1.8 million mutated strains enabled us to calculate the average number of negative and positive interactions that each gene has with a set of 100-200 random mutations. We then defined buffer genes as those genes that showed an extreme level of negative interactions with a very large number of random mutations—but again, there is no clear cutoff, it is just a long list of all 5000 genes that were tested, ranked according to their ‘buffering effect’.
Interestingly, the genes that show the most buffering activity are a small set of evolutionary conserved buffer genes involved in protein folding and chromatin organization, including GIM3, SSA2, HOG1, and FKH2. Deletion of these genes influenced the fitness effect of de novo mutations (i.e. new genetic variation that has not yet experienced natural selection) as well as standing genetic variation (i.e. genetic variation present within a population that has been subjected to natural selection)."
Hala Kasmo: “What’s more, we were able to show that losing a buffer gene resulted in a decline of standing genetic variation, because some of the mutations cause a sharp decline in fitness when they are no longer ‘buffered’. Additionally, loss of some buffer genes caused higher fitness variation in some environmental conditions than others, suggesting that the ability of these genes to buffer mutations depends on the environmental context.”
Mo Tawfeeq: “As a follow-up, we are currently investigating exactly what type of mutations are buffered by these different buffer genes, which will hopefully lead to more insight into the exact mechanisms. We have a follow-up paper on this that was recently published, where we focus on GIM3 and show that under stress from the antifungal drug fluconazole, Gim3 stabilizes mutant Erg3 proteins, allowing these mutations to have immediate phenotypic effects that support growth, antifungal resistance, and cross-resistance to other antifungal drugs.”5
Hsp90 has been studied for decades, but as you say, buffering is a much broader concept. Why the increased interest for drug development now?
Kevin Verstrepen: “There are several recent papers about how biology can mask ‘problem’ mutations that would otherwise hurt the fitness of the organism. Importantly, to be able to really say something about buffering, the mechanisms involved, and how it can affect human health, we need large screens, high-throughput editing tools to test the effect of specific mutations, large genomic datasets, as well as extensive health records—things that have only become available and/or affordable in the past years.
Recent advances in machine learning have made it possible to process these large datasets and uncover links or predict penetrance.6 A key caveat here is that even if you could predict penetrance relatively accurately, without knowing exactly the mechanisms involved (why does a patient with a specific mutation develop a disease?), there still is no path to clinical treatment. This again shows the importance of knowing exactly how a buffer gene can impact a phenotype!”
What are some of the areas where you see progress and excitement on this topic?
Iris Goeminne: “One recent study that we found particularly exciting is a study by the group of Dan Jarosz at Stanford. They identified more than 5,000 Hsp90-dependent loci, with around 50% of these being cis-regulatory variants.7 This came as quite a surprise, since almost all past studies had focused on coding variants impacting protein folding. This can have profound clinical implications as well. Many clinical datasets only contain exome data, since many clinical tests do not sequence the full genome, only the exome.
Kevin Verstrepen: “For example, ClinVar, a key public archive of reports of human variations classified for diseases and drug responses, which is often used in biomedical studies, is mainly dominated by coding variants. This also means we might be missing out on key (buffered) mutations relevant for disease.”
Can you give some examples of buffering in action in biomedicine?
Hala Kasmo: “Yes, there are several. Besides the recent BRCA1 paper,2 and the work on Fanconi anaemia,8 both mentioned in the Nature news feature by Philip Ball, there is recent work showing TP53, a key tumor suppressor gene that helps protect cells against cancer, is buffered by HSF1.3 HSF1 is a regulator of Hsp90, again showing how many studies circle back to Hsp90. There are also papers that find that cancers adapt to their mutational load by buffering protein misfolding stress.9
Kevin Verstrepen: “Several of these studies indicate that buffering (Hsp90) can shape the mutational space that is accessible. In other words, Hsp90 buffering can allow mutations to arise without an immediate phenotypic effect, and could then allow these mutations to further spread in the population. Note that in some cases this buffering effect can be positive (e.g. in the case of the BRCA1 paper, where Hsp90 buffering reduces the clinical severity) or detrimental (i.e. in some cancers, where Hsp90 buffering allows cancer cells to survive). While there are several Hsp90 inhibitors tested as cancer treatment,10 these results show there is no one approach that suits all cancers.”
What about the link between buffering and adaptability?
Hala Kasmo: “Since most mutations are neutral or negative, buffering decreases a cell’s susceptibility to ‘bad’ mutations. However, evolution is fueled by ‘positive’ mutations, those that make an organism more fit. At first sight, buffering could then also suppress the effect of these positive, adaptive mutations and thus slow down evolution.
Susan Lindquist heavily promoted the idea that buffering can in some cases increase evolvability and adaptability because buffering is often reduced in times of stress, when an organism is not well adapted to its environment and thus ‘in need’ of evolution and adaptation. This was initially criticized, since the vast majority of phenotypes released by Hsp90 inhibition in plants and animals is detrimental. Think of fruit flies with deformed wings for example.
Recent studies, however, have now shown that in some cases the released variation can be adaptive. There is, for example, the currently well-known example of eye size in Mexican cavefish,11 and just last year, a study in red flour beetle showed adaptive eye size reduction upon Hsp90 inhibition, and, importantly, also provided insights into the exact mechanism behind this.10 ”
There is a tempting narrative that buffering evolved in order to help organisms adapt under stress. You are more cautious than that. Why?
Kevin Verstrepen: “Yes, it is very important to stress that it is absolutely not clear—and in our view very unlikely—that variation in buffering in times of stress has evolved to support or guide evolution. It is equally possible that it is simply an indirect effect. While Lindquist was a strong advocate for the first hypothesis, i.e. that buffering is the result of adaptive evolution, with buffer genes having evolved to allow the accumulation of hidden genetic variation that can be released in times of stress, others have argued that the effects that are observed with Hsp90 are biased."
"In our view, it's very unlikely that variation in buffering in times of stress has evolved to support or guide evolution. It is equally possible that it is ismply an indirect effect."
Iris Goeminne: “Specifically—and this is where we do need to get (even) more technical—most studies only look at standing genetic variation, variation that has been under selection. Therefore, mutations that are buffered by Hsp90 have been allowed to accumulate, but those that are not buffered have been removed by selection, resulting in a bias towards standing genetic variation that is buffered.”
If buffering can hide some mutational effects, how can protein function evolve at all, without reducing fitness?
Kevin Verstrepen: “This is again a very complex question that requires a full textbook for a complete and nuanced answer. Firstly, evolution is a never-ending process, and not all, maybe even no proteins have reached their (theoretical) maximum. Many may, however, be locked in a so-called ‘local maximum’, a state where small changes that often occur naturally indeed do not result in an improved function, or even a small or large decrease.”
Hala Kasmo: “One of the key things here is that often protein functions are optimized to function in a specific environment, be it a genetic or ecological environment. However, when this environment changes, for example due to mutations in the genome or a change in the external conditions of a cell or organism, what was previously optimal might no longer be so, and the other way around: what was previously considered as not optimal, might suddenly be.”
Iris Goeminne: “People have hypothesized that buffer genes allow the accumulation of mutations without immediate phenotypic effects, with the buffer genes acting as so-called shock absorbers. In changing conditions, buffer gene activity could be compromised, and the previously hidden variation can be released. Please note that all this is from a buffer gene perspective, but there are other ways of masking effects of mutations and allowing proteins to evolve new functions, for example gene duplications.13
Another effect that can occur is that some random mutation in a gene encoding a protein, or a gene that somehow influences how a protein works, completely changes the working of the protein in question, often reducing its activity and making it less optimal. Still, if they are not overly harmful in a short timeframe, such seemingly ‘negative’ mutations can allow a protein to accumulate more positive mutations and find a completely new form or function that is not very different from the ‘local maximum’ where it was locked in before.”
The picture that emerges is both exciting and unfinished. Buffering is not just a striking idea attached to one famous protein, but is becoming a broader framework for thinking about how genomes tolerate change, how disease risk is modulated, and how variation may accumulate over time. But, as Kevin stresses, we’re just at the beginning.
Kevin Verstrepen: “Many things related to buffering and evolvability remain unresolved. For example, we still don’t know if and how buffering shapes variation accumulating in the genome, and we lack detailed experimental studies on the effect of buffering on evolvability.”
References
1. Tawfeeq, M. T. et al. Mutational robustness and the role of buffer genes in evolvability. EMBO J. 43, 2294–2307 (2024).
2. Gracia, B. et al. HSP90 buffers deleterious genetic variations in BRCA1. Mol. Cell 85, 4365-4378.e10 (2025).
3. Halim, S. et al. Dominant-negative TP53 mutations potentiated by the HSF1-regulated proteostasis network. Mol. Cell 86, 345-361.e6 (2026).
4. Frickel, J. et al. Genes involved in protein folding and chromatin organization buffer genetic variation. 2024.09.24.614041 Preprint at https://doi.org/10.1101/2024.09.24.614041 (2024).
5. Tawfeeq, M.T. et al. Gim3 buffers and potentiates de novo mutations that affect fluconazole susceptibility in yeast. EMBO Rep. 27(6):1510-1539 (2026).
6. Forrest, I. S. et al. Machine learning–based penetrance of genetic variants. Science 389, eadm7066 (2025).
7. Jakobson, C. M., Aguilar-Rodríguez, J. & Jarosz, D. F. Hsp90 shapes adaptation by controlling the fitness consequences of regulatory variation. 2023.10.30.564848 Preprint at https://doi.org/10.1101/2023.10.30.564848 (2023).
8. Karras, G. I. et al. HSP90 shapes the consequences of human genetic variation. Cell 168, 856-866.e12 (2017).
9. Tilk, S., Frydman, J., Curtis, C. & Petrov, D. A. Cancers adapt to their mutational load by buffering protein misfolding stress. eLife 12, RP87301 (2024).
10. Sayed, R. et al. HSP90 as an evolutionary capacitor drives adaptive eye size reduction via atonal. Nat. Commun. 16, 9277 (2025).
11. Rohner, N. et al. Cryptic variation in morphological evolution: HSP90 as a capacitor for loss of eyes in cavefish. Science 342, 1372–1375 (2013).
12. Geiler-Samerotte, K. A., Zhu, Y. O., Goulet, B. E., Hall, D. W. & Siegal, M. L. Selection Transforms the Landscape of Genetic Variation Interacting with Hsp90. PLOS Biol. 14, e2000465 (2016).
13. Soskine, M. & Tawfik, D. S. Mutational effects and the evolution of new protein functions. Nat. Rev. Genet.11, 572–582 (2010).
