Scientists Build Custom Protein Add-Ons to Rewire a Nanopore’s Electrical Behavior

Protein nanopores have become workhorses of modern biotechnology, most famously as the sensing elements inside DNA sequencing devices that read individual molecules as they thread through a nanoscale channel. Yet the pores used in these applications are almost always natural proteins, assembled from many identical subunits, and their performance is dictated by the geometry and chemistry of their internal channel, or lumen. In a study published in Nature Structural & Molecular Biology, a team led by Lee Schnaider, A. Katherine Hatstat and Alistair J. Scott, working in the laboratory of William F. DeGrado at the University of California, San Francisco, together with collaborators at Oxford Nanopore Technologies, reports a strategy that departs from the usual playbook. Rather than designing an entirely new pore or relying on point mutations to tune a natural scaffold, the researchers installed additional de novo designed protein subunits onto an existing native pore complex, achieving large-scale architectural changes to a working biological nanopore.

The scaffold they chose was CsgG, a nonameric bacterial secretion pore that has found wide use in sensing applications because of its stable, well-characterized conductance. The challenge was formidable: the new designed components had to fit inside a confined, nonuniform space governed by ninefold rotational symmetry, and they had to do so without collapsing the channel. A functioning pore needs an open conducting lumen and a stable, low-noise baseline current; any design that distorted the assembly or obstructed the pathway would be electrically useless. The team’s solution was to design new protein subunits, which they named CsgX, that integrate seamlessly with the native CsgG ring, producing semisynthetic conduction pores containing eighteen subunits in total and weighing approximately 315 kilodaltons.

The computational approach leaned heavily on structural databases and established protein design tools. Using the MASTER fragment-search method, the designers first identified helical segments from the Protein Data Bank that could pack against an anchor helix derived from CsgF, a natural peptide that associates with the CsgG pore. Sequential searches of a non-redundant set of structures yielded roughly one thousand structural matches, which were clustered by root-mean-square deviation, and the geometrically recurring pairs were carried forward. Ninefold symmetry was then applied to generate complete assemblies, and only clash-free models with appropriate lumen geometry, in the range of roughly 19 to 23 angstroms in diameter, advanced to designability scoring. From there, loop-building connected the helices into helix-loop-helix motifs, and three backbone models were selected on the basis of structural prevalence, interface geometry and feasibility, with poorly packed or sterically occluded candidates excluded.

Interestingly, the team also attempted to generate the same CsgX-like topologies with RFdiffusion, the deep-learning-based protein design method that has transformed much of the field. Those attempts largely failed in a revealing way: the outputs either folded toward the CsgG body rather than into the lumen, or simply extended the anchor helix instead of creating the desired hairpin. The authors note that with more careful tuning this might be overcome, but they instead favored the fragment-based route, which allowed them to design against interfaces as close to optimal as the available structural data permitted, while keeping the design process transparent enough to extract physical principles for constructing functional pore architectures.

The designed variants came in two flavors, linear and cyclized versions of two distinct CsgX designs, CsgX1 and CsgX2, which differed in their engineered hydrogen-bonding networks. In CsgX1, intermolecular hydrogen bonds were designed into the helix-loop-helix motif, with hydrophobic interactions anchoring the C-terminal region; in CsgX2, the hydrogen-bonding network within the motif was intramolecular, while intermolecular contacts were reserved for the C-terminal region. These subtle differences in stabilizing interactions gave the team a set of closely related but structurally distinct pores with which to probe how lumen architecture translates into electrical behavior.

Structural validation came from cryo-electron microscopy. The team determined structures of four complexes, deposited in the Protein Data Bank and Electron Microscopy Data Bank under identifiers including 9PNA, 9PN8, 9PN9 and 9PNB, and the resulting maps confirmed the designed lumen architecture. For the CsgX1 variants and the cyclized CsgX2 design, density was observable for the full designed hairpin, demonstrating that the computationally specified helix-loop-helix motifs fold as intended within the confined pore interior. The one exception was the linear CsgX2 construct, where no density was visible beyond residue 30, precluding structural elucidation of its designed hairpin. The cryo-EM workflows, processed with tools including MotionCor2, Scipion and cryoSPARC, showed the usual mixture of single pores and pore dimers in two-dimensional class averages, with the cyclized CsgX2 variant uniquely showing only single pores.

The electrical measurements were where the designs truly proved themselves. Single-channel recordings performed in Oxford Nanopore Technologies MinION flow cells, using LSK114 sequencing buffer at 180 millivolts, showed that the semisynthetic pores insert into synthetic membranes and conduct ions with stable, low-noise open-pore currents. Crucially, the complexes exhibited distinct current-voltage responses relative to the native pore, including clear current rectification, meaning the pores conduct ions more readily in one direction than the other. Rectification is a valuable property in sensing because it can sharpen the electrical signature of molecules passing through the channel, and the fact that it emerged from designed subunits rather than from mutations to the native pore demonstrates that the approach can genuinely reprogram a pore’s electrical personality.

The broader significance of the work lies in the design philosophy. Natural multi-subunit pores are constrained by evolution, and modifying them through mutation alone offers only incremental control over lumen geometry and chemistry. De novo design of entirely new transmembrane pores, while advancing rapidly, still faces enormous challenges in achieving the stability and low-noise conductance that practical sensing demands. The semisynthetic strategy occupies a powerful middle ground: it preserves the proven conductance and assembly behavior of a native scaffold while allowing designed components to reshape the channel interior on a large scale. The successful integration of nine designed subunits into a working eighteen-subunit complex, under stringent symmetry and structural constraints, establishes a general blueprint for modifying existing nano-assemblies with custom protein parts.

The implications for nanopore sensing could be substantial. Because conductance depends directly on lumen geometry and chemistry, the ability to install designed subunits with tailored internal surfaces opens the door to pores optimized for specific analytes, from DNA and RNA to proteins and small molecules, a direction the field is actively pursuing in the push toward single-molecule proteomics. The design scripts used in the study have been released via Zenodo, and the structural models are publicly available, lowering the barrier for other groups to adopt and extend the approach. With support from the National Institutes of Health, the National Science Foundation and Oxford Nanopore Technologies, and with several of the authors named as coinventors on a related patent application, the work sits at the intersection of open science and commercial development, and it signals that the next generation of nanopores may be built as much by design as by evolution.

For a field that has spent three decades refining natural pores for sequencing and sensing, the demonstration that de novo protein components can be grafted onto a native pore complex and validated at atomic resolution marks a conceptual shift. The semisynthetic pores described by Schnaider and colleagues are not merely modified versions of CsgG; they are hybrid machines in which designed and natural protein parts cooperate to produce electrical behavior that neither could achieve alone. As the design toolkit matures, the same strategy could be applied to other symmetric nano-assemblies, extending the reach of computational protein design into some of the most demanding and commercially important structures in biotechnology.

Subject of Research: De novo protein design of semisynthetic nanopore conduction pores based on the CsgG scaffold

Article Title: De novo design of semisynthetic conduction pores

Article References: Schnaider, L., Hatstat, A. K., Scott, A. J., Tan, S. K., Hambley, R. G., Dawson, W. M., Griffiths, R. C., Kormos, R. C., Melo, A. A., Tse, E., Polizzi, N. F., Wallace, E. J., Merz, G. E., & DeGrado, W. F. (2026). De novo design of semisynthetic conduction pores. Nature Structural & Molecular Biology. https://doi.org/10.1038/s41594-026-01881-w

Image Credits: AI Generated

DOI: 10.1038/s41594-026-01881-w

Keywords: nanopore, protein design, de novo design, CsgG, cryo-EM, ion channels, current rectification, semisynthetic pores, structural biology, molecular sensing, DNA sequencing, protein engineering