Project Description

Abstract

Chronic Rhinosinusitis (CRS) poses a severe global health challenge due to its high prevalence, painful symptoms, and the fact that it is often underestimated by patients. Current clinical approaches remain significantly limited, making timely diagnosis and complete recovery difficult to achieve. Therefore, we are dedicated to developing a practical solution for early diagnosis and precision treatment.

Granzyme K (GZMK) has recently been identified as a pivotal enzyme driving both the initiation and progression of CRS. Centered on this target, we developed an integrated diagnostic–treatment strategy. On the diagnostic side, a colloidal-gold-based rapid test enables early and accessible screening of rhinosinusitis. On the treatment side, we are developing small-molecule inhibitors to achieve precise enzymatic inhibition and inflammation control. This integrated diagnostic–therapeutic framework centered on GZMK establishes a new paradigm for early screening and precision intervention in rhinosinusitis.

In the process, we also established two derivative technology platforms: the Prometheus system for de novo protein design and the Nexus system for high-throughput affinity measurement. These two platforms provide a systematic solution for efficiently designing and validating high-affinity protein binders, contributing more powerful research tools to the field of synthetic biology.

We are confident that this work will not only advance the clinical management of rhinosinusitis but also contribute powerful new tools and insights to the broader biomedical community.

Introduction

The early diagnosis of Chronic Rhinosinusitis (CRS) presents a dual challenge. On one hand, current definitive diagnostic methods—such as imaging and endoscopy—depend on expensive, specialized equipment. Their diagnostic value often becomes conclusive only when the disease has advanced to moderate or severe stages, often accompanied by intense facial pain, congestion, and a high likelihood of developing nasal polyps as a complication. On the other hand, patients frequently mistake early symptoms for a common cold or allergies, leading to delays in medical consultation. As a result, diagnosis often lags behind disease progression, causing patients to miss the optimal window for intervention.

On the therapeutic front, while Functional Endoscopic Sinus Surgery (FESS) can remove lesions and provide rapid symptom relief, postoperative discomfort and a recurrence rate of up to 60% trap many patients in a distressing cycle of repeated surgeries. Moreover, steroid-based medications, although capable of suppressing inflammation, remain a suboptimal long-term solution due to their limited specificity and the inherent risk of systemic side effects.

To optimize current rhinosinusitis management and improve patient outcomes, we posed a central question: How can we achieve early diagnosis and precision treatment for rhinosinusitis? Driven by this inquiry, we conceived DOCTOR — Diagnostic Optimization & Chemical Treatment Of Rhinosinusitis — with the mission of establishing a novel, integrated strategy for the early diagnosis and precision treatment of rhinosinusitis.

Recent studies published in early 2025 have revealed a new pathogenic mechanism underlying CRS. Recurrent stimulation of the nasal epithelium by pathogens or allergens compromises the mucosal barrier and recruits immune cells, particularly CD8⁺ T cells. A distinct subset of these cells secretes large amounts of Granzyme K (GZMK), a pivotal enzyme that drives disease progression. Extracellular GZMK induces epithelial cells to release inflammatory cytokines such as IL-6 and IL-8, while simultaneously activating the complement system. This dual action amplifies local inflammation, leading to persistent complement activation, tissue damage, and remodeling—a vicious cycle that promotes the transition from acute to chronic rhinosinusitis and the eventual formation of nasal polyps.

Importantly, these findings underscore the dual value of GZMK as both a diagnostic biomarker and a therapeutic target. As a biomarker, GZMK demonstrates superior sensitivity and specificity compared with conventional indicators, providing an earlier reflection of disease activity and recurrence risk. As a therapeutic target, selective inhibition of its enzymatic activity can effectively suppress inflammation and restore tissue homeostasis, halting the progression of CRS and nasal polyps.

Inspired by this discovery, we devised a novel GZMK-centered diagnostic-treatment approach, integrating early detection with precise enzymatic inhibition. This strategy opens new avenues for the effective control—and potentially the eventual cure—of rhinosinusitis.

Diagnosis of Rhinosinusitis

To enable early diagnosis and disease monitoring of CRS, we set out to develop a rapid and reliable detection tool for GZMK. Leveraging the principles of lateral flow immunochromatography, we engineered a colloidal gold test strip named Rhinolens for swift, on-site detection of GZMK. The key innovation lies in replacing traditional antibodies with a computationally designed protein binder that recognizes GZMK with high specificity and affinity.

Traditional antibody production is time-consuming, costly, and dependent on animal immunization. To overcome these limitations, we adopted a de novo protein design approach, which allows for fast, low-cost development and flexible optimization. Our final choice—a de novo-designed Binder—offers exceptional design versatility, high predicted affinity, and robust expression in E. coli.

The design process integrates a series of advanced computational tools. First, key interaction hotspots on GZMK were identified. The RFdiffusion model was then used to generate initial protein backbones, which were further refined using ProteinMPNN, AlphaFold2, and Rosetta to optimize both sequence and structure. Based on computational scoring, the most promising candidates were selected for experimental validation. These binders were expressed and purified, and their affinities were quantified via Surface Plasmon Resonance (SPR). The binder demonstrating the highest performance was subsequently incorporated into the test strip design.

Building upon these results, we implemented a dual-binder sandwich format to streamline test strip development. In this design, GZMK is “sandwiched” between two binders recognizing distinct epitopes, replacing the conventional antibody–antigen model. The architecture maintains assay performance while substantially lowering development cost, shortening preparation time, and improving adaptability across targets.

The test strip consists of five layers: a backing card, sample pad, conjugation pad, nitrocellulose (NC) membrane, and absorbent pad. For the control line, SUMO1 conjugated with colloidal gold is applied to the conjugation pad, while its binding partner Ubc9 is immobilized on the control line to ensure proper strip performance. For GZMK detection, Binder 1 is conjugated to colloidal gold on the conjugation pad, and Binder 2, which binds a separate GZMK site, is immobilized on the test line of the NC membrane.

During testing, if GZMK is absent, only the SUMO1–Ubc9 interaction occurs, producing a visible C-line. When GZMK is present, the Binder 1–GZMK–Binder 2 complex forms and accumulates at the test line, resulting in visible C-line and T-line signals.

Through this work, we successfully established a binder-based colloidal gold test strip capable of rapid, specific, and user-friendly detection of GZMK, providing a foundation for the early diagnosis of rhinosinusitis.

Treatment of Rhinosinusitis

With GZMK serving as a potential therapeutic target for CRS, our project aims to discover and validate candidate drugs that can inhibit its activity through high-throughput screening. The ultimate goal of this effort is to realize a precision medicine approach for the treatment of rhinosinusitis.

To enable efficient drug screening, we first expressed and purified catalytically active GZMK protein. Using the HEK 293F expression system, we obtained secreted GZMK, which was subsequently purified through affinity chromatography followed by size-exclusion chromatography. The final protein product showed high purity and correct molecular weight, as confirmed by SDS-PAGE and mass spectrometry, providing a solid basis for downstream biochemical assays.

To verify enzymatic function and establish a reproducible screening system, we developed an in vitro fluorescence-based GZMK activity assay. This assay confirmed the catalytic activity of GZMK and enabled us to determine key kinetic parameters, offering a robust and quantitative foundation for inhibitor evaluation.

Leveraging this assay, we conducted a high-throughput screening (HTS) campaign of 1,813 FDA-approved drugs, systematically measuring their inhibitory effects on GZMK activity. From this screen, we identified three primary hit compounds exhibiting more than 90% inhibition. The most potent candidate was selected for further dose–response and mechanism-of-action studies, laying the groundwork for future optimization and preclinical research.

We also initiated a parallel virtual screening campaign against GZMK. Lacking an experimental structure, we first built a high-confidence model of GZMK using AlphaFold3 and then performed an initial screen with the Glide docking model.

By comparing these results with our physical screen, we established a new virtual screening workflow. This process uses a known inhibitor as a "seed" to find structurally similar molecules, then employs more precise computational methods to validate their binding. This workflow balances efficiency and accuracy, yielding multiple lead candidates for GZMK. Furthermore, it provides a new methodological paradigm for discovering drugs against targets with flexible active-site pockets.

Towards an Optimized Workflow

Software

To enhance the detection sensitivity and reliability of the developed sinusitis test strips, we aim to further improve the binding affinity between GZMK and the de novo designed proteins. We developed BetterMPNN, a brand-new one-shot protein design method, which enables the efficient and precise design of GZMK binders with higher affinity under limited experimental throughput.

Our workflow with BetterMPNN is based on “exploration–evaluation–optimization” loop. First, use RFdiffusion to generate backbones. Then, for each backbone, execute the "exploration - evaluation - optimization" loop: (1) Design multiple sequences for the same backbone using ProteinMPNN; (2) Evaluate the sequence properties in the Environment and calculate the Reward; (3) Use GRPO to optimize the parameters of ProteinMPNN. Repeat the loop until the model converges to the target, which means it has learned the ability to generate better binders as we expected.

Through this workflow, we successfully designed potential protein inhibitors verified by in silico experiments that directly target the active pocket of GZMK within 20 hours. Meanwhile, we obtained binders that can bind to other hot-spots of GZMK, which can be used for detection work. Most of the results are superior to the binders designed by traditional methods in in-silico experiments. Our proposed BetterMPNN model can not only be used to optimize our test strips but also serves as a general tool, providing a novel and reliable solution for the precise and efficient de novo design of proteins, representing a fundamental advance.

Hardware

During the design of our colloidal gold test strip, we identified a key bottleneck: the low success rate of de novo protein design. Traditional methods require extensive expression and screening to find an effective binder, making the process inefficient and costly. To address this, we developed the high-throughput affinity screening platform, Nexus, which utilizes a cell-free system and automated detection to rapidly express and initially screen for binders with higher affinity. As a rapid and cost-effective tool for high-throughput affinity screening and measurement, the Nexus platform provides broadly applicable technical support for accelerating various de novo protein design projects.

Science & Society

Our Education and Human Practices work was deeply integrated with our project's development. Through diverse educational outreach, we raised public awareness of synthetic biology and Chronic Rhinosinusitis . Crucially, we established a continuous feedback loop with clinicians, patients, and industry stakeholders. Their insights, gathered through interviews and surveys, were directly integrated into our project, guiding its design and refinement. Close consultation with engineers and professors further ensured our work remained scientifically robust and socially valuable.

These integrated efforts have been vital. They have not only improved our project's scientific design but also ensured our research is aimed squarely at real-world application, paving a path from the laboratory to tangible improvements in human health.