Computational design of pH-sensitive binders

Why I picked this preprint

I want to highlight this preprint not just because it’s from David Baker’s lab, whose work I greatly admire, but because it demonstrates how computational design can replace trial-and-error approaches with systematic, mechanism-driven solutions. While previous methods relied on laborious combinatorial histidine scanning and testing, this work uses deep-learning-based protein design integrated with physics-based scoring to generate pH-dependent binders. The researchers developed two complementary strategies addressing different use cases—polymer stabilization and monomer destabilization—showing how computational methods can provide genuine insights into the molecular principles underlying pH-dependent protein interactions.

Background

Proteins inside cells need to be carefully controlled—where they go and what they do affects everything from cell signaling to disease progression (1, 2). One promising way to control proteins is using circulating antibodies that can enter cells. These cell-penetrating antibodies, recycled through a cellular pathway called FcRn, can grab onto multiple target proteins, break down harmful enzymes, trigger-controlled protein degradation, and even block key checkpoints in tumors (1, 12). Despite these exciting applications, most current methods rely on random trial-and-error to find pH-sensitive antibodies. This approach, combined with simple library screening, is labor-intensive and doesn’t help us understand why certain designs work.

The researchers’ goal was to develop computational design methods that systematically encode pH-sensitivity into protein binders, leveraging recent advances in deep-learning protein design combined with physics-based modeling of both protonation states and electrostatic interactions.

Key Findings

Unbiased histidine incorporation fails to generate pH-sensitivity, but interface histidine-cationic interactions create electrostatic repulsion

Baker and colleagues designed pH-sensitive binders using Protein MPNN (a deep-learning protein design tool) with histidine bias, generating 12,000 candidates. However, simply adding histidines yielded minimal pH-sensitivity (Figure 2A). Structural analysis revealed only a few histidines actually formed pH-dependent binding interactions (bonds that change strength with pH) at low pH.

Screening identified four EphA2 binders (proteins that bind to the EphA2 receptor) with up to 1000-fold weaker binding at acidic pH (preprint-Figure 2B). Bilayer interferometry (a technique that measures binding strength in real-time) showed rapid dissociation at pH 5.4 (Figure 2B). His15 and His19 residues (specific histidine positions) drove this pH-sensitivity, located 5.7 Å and 7.8 Å from target arginines (positively charged amino acids) (Figure 2C).

The key finding: pH-sensitive binders require strategic positioning of histidines near positively charged residues (like arginine and lysine), not just high histidine content overall.

Figure 2: Engineering pH-responsive binding interfaces: A. SPR binding curves for EphA2 and PCSK9 binders designed with histidine-biased Protein MPNN, tested at pH 7.4 and 5.5. B. Structural models of pH-responsive EphA2 binders from yeast display screening. SPR measurements at pH 7.4 and 5.4 show binding strength (first two panels), with accelerated release at lower pH (third panel). C. Testing individual histidine mutations in EphA2_pH_1 binder using BLI at pH 7.4 and 5.4.

Buried histidine-charged networks destabilize binders under acidic conditions

The first approach required strategically placing histidines near charged residues at the binding interface—but this depends on the target protein having the right residues in the right places. To overcome this limitation, the authors developed a complementary strategy: installing buried histidines (histidines hidden inside the protein core) within the binder itself (preprint-Figure 3A). At acidic pH, electrostatic repulsion destabilizes the binder from the inside, causing dissociation regardless of the binding interface.

ILVBP binders (targeting the insulin-like growth factor binding protein) with 5 buried histidines showed a striking 10^8-fold reduction in binding at pH 6.4 (preprint-Figure 3B). IL-6 binders (targeting interleukin-6, an inflammatory protein) achieved up to 6-fold reduction (Figure 3C), and PCSK9 binders (targeting proprotein convertase subtilisin/kexin type 9, a cholesterol regulator) up to 3.5-fold reduction from 40 designs (Figure 3D, preprint-Figure 5), demonstrating versatility across targets.

The key insight: Networks of buried, cooperating charged residues can destabilize binders at mildly acidic pH 6.4 (preprint-Figure 3E, preprint-Figure S10A-B), with cooperativity enhancing pH-sensitivity. This raises an important question: how can we computationally design these networks in new binders?

Figure 3: Engineering pH-sensitivity through buried charged networks: C. IL-6 pH-sensitive binder showing the buried hydrogen bond network (left panel). SPR binding curves at pH 7.4 and 5.4 comparing the original binder (IL-6 Parent) to the pH-sensitive version (IL6_pH). D. PCSK9 pH-sensitive binder showing the buried hydrogen bond network (left panel). SPR binding curves at pH 7.4 and 5.4 comparing the original binder (PCSK9 Parent) to the pH-sensitive version (PCSK9_pH).

pH-sensitive LYTACs achieve catalytic, multiturnover protein degradation

The researchers then combined their pH-sensitive binders with LYTACs (lysosome-targeting chimeras—molecules that tag proteins for degradation in lysosomes). They fused zinc-targeting chimeras using a pH-insensitive receptor binder (Rigid1 against IGF2R, the insulin-like growth factor 2 receptor) to a pH-sensitive target binder, enabling degradation.

A pH-sensitive EphA2-targeting LYTAC showed target-dependent EphA2 degradation in cells (Figure 4C-D) and eliminated phosphorylated EphA2 at Ser897 (preprint-Figure 4E). Following pH-sensitive LYTAC treatment, cells internalized significant target amounts through the lysosomal recycling pathway (preprint-Figure 4G), confirming catalytic multiturnover degradation through this recycling mechanism (preprint-Figure 4F). Batimastat (a matrix metalloproteinase inhibitor) blocked degradation (preprint-Figure 4G), confirming the pathway’s dependence on endolysosomal acidification (the process of lysosomes becoming acidic). A pH-sensitive IL-6 degrader achieved higher potency at subnanomolar concentrations (preprint-Figure 4H), enabling efficient cytokine depletion.

Figure 4: Catalytic degradation of membrane and secreted proteins using pH-sensitive binders: C. EphA2 degradation in HCT116 cells measured by flow cytometry (cell-surface levels, left) and western blot (total protein levels in HCT116 and MDA-MB-231 cells, middle and right). Cells treated with 250 nM designed proteins for 24h. D. Dose-response curve for cell-surface EphA2 degradation in HCT116 cells after 24h treatment. E. Western blot showing phosphorylated EphA2 (p-EphA2, Ser897) levels after 24h treatment. F. EphA2 degradation persists after washout. HCT116 cells treated with 250 nM proteins for 24h, washed, then incubated in fresh media for 6h. Cell-surface levels measured by flow cytometry (mean fluorescence intensity, MFI).

What makes this exciting: The study establishes generalizable computational approaches for designing pH-sensitive binders through two complementary mechanisms.

Main Takeaways

1. Strategic placement beats random addition: Simply adding histidines doesn’t create pH-sensitivity. Instead, histidines must be strategically positioned near charged residues (like arginines) at the binding interface to generate electrostatic interactions that weaken at low pH.
2. Buried networks offer target-independent control: Installing networks of buried histidines inside the binder’s core provides an alternative strategy that destabilizes the protein from within at acidic pH—working independently of the target protein’s surface properties.
3. Computational design enables therapeutic applications: These generalizable computational approaches successfully created pH-sensitive LYTACs (lysosome-targeting chimeras) that achieve catalytic, multiturnover protein degradation through lysosomal recycling—demonstrating real therapeutic potential.

Questions/Future Directions

• Fine-tuning pH sensitivity: How can pH thresholds be optimized for specific environments? For example, can we design binders that activate precisely at tumor microenvironment pH (around 6.5) by adjusting histidine pKa values (the pH at which histidines gain or lose charge)?
• System capacity: What limits exist on the number of degradation cycles these systems can achieve? Are there practical limits to their catalytic turnover?
• Performance benchmarking: How do the expression levels and stability of these computationally designed binders compare to natural antibodies or antibody-derived therapeutics in clinical settings?

References

1. Hu, Y.-B., Dammer, E. B., Ren, R.-J. & Wang, G. The endosomal-lysosomal system: from acidification and cargo sorting to neurodegeneration. Transl. Neurodegener. 4, 18 (2015).
2. Müller, S., Dennemärker, J. & Reinheckel, T. Specific functions of lysosomal proteases in endocytic and autophagic pathways. Biochim. Biophys. Acta 1824, 34–43 (2012).
3. Watson, J. L. et al. De novo design of protein structure and function with RFdiffusion. Nature 620, 1089–1100 (2023).
4. Dauparas, J. et al. Robust deep learning-based protein sequence design using ProteinMPNN. Science 378, 49–56 (2022).
5. Zhou, F. et al. Tumor microenvironment-activatable prodrug vesicles for nanoenabled cancer chemoimmunotherapy combining immunogenic cell death induction and CD47 blockade. Mater. 31, e1805888 (2019).
6. Thisted, T. et al. VISTA checkpoint inhibition by pH-selective antibody SNS-101 with optimized safety and pharmacokinetic profiles enhances PD-1 response. Commun. 15, 2917 (2024).
7. Zou, W. et al. A stepwise mutagenesis approach using histidine and acidic amino acid to engineer highly pH-dependent protein switches. 3 Biotech 12, 21 (2022)
8. Murtaugh, M. L., Fanning, S. W., Sharma, T. M., Terry, A. M. & Horn, J. R. A combinatorial histidine scanning library approach to engineer highly pH-dependent protein switches. Protein Sci. 20, 1619–1631 (2011).

Tags: #proteindesign #computationalbiology #phsensitivebinders #proteinengineering #machinelearning #targetedproteindegradation #lytac #bakerlab #biorxiv #biotechnology

doi: https://doi.org/10.1242/prelights.41957

Read preprint (No Ratings Yet)