Mammalian GLUT5 and SGLT1 substrate specificity biosensor platform

Overview

This project develops a high-throughput discovery platform for substrates and competitive inhibitors of two clinically important mammalian sugar transporters: human GLUT5 (fructose transporter) and rabbit SGLT1 (sodium-glucose cotransporter). The work converts these membrane transporters into intramolecular fluorescence biosensors expressed in yeast, enabling direct detection of small-molecule binding. Screening data trains explanatory and generative machine learning models that reveal the molecular basis of substrate specificity and design new candidate molecules de novo. The approach addresses a long-standing challenge in transporter biology: conformational changes induced by substrate or inhibitor binding are difficult to measure experimentally.

Technical specifications

Core technology:

• Intramolecular insertion of a conformation-sensitive fluorescent protein into GLUT5 and SGLT1, inspired by the validated SweetTrac1 biosensor for the plant SWEET1 transporter
• Heterologous expression in yeast, where both GLUT5 and SGLT1 have been demonstrated to be functional
• High-throughput fluorescence screening of sugar and sugar analog chemical libraries
• Binary classifier machine learning models trained on screening results to predict additional binders
• Generative models to propose novel substrates and competitive inhibitors computationally
• Secondary validation using a second transporter converted into a biosensor (analogous to SWEET2) to confirm bona fide transport versus inhibition
• A transacceleration-based assay to differentiate substrates from competitive inhibitors when radiolabeled versions are unavailable

Key advantages:

• Builds directly on a published, validated methodology (Park et al., 2022; Park et al., 2023)
• Applicable to transporters implicated in diabetes and metabolic disease
• Generates both experimental hits and computationally predicted candidates for chemical synthesis and testing

Technology readiness level

The underlying biosensor concept and screening pipeline have been validated for the plant SWEET1 transporter, where the approach identified known diabetes drugs such as 1-deoxynojirimycin and voglibose as unexpected binders. Extension to mammalian GLUT5 and SGLT1 is expected to require approximately one year for biosensor conversion and a second year for chemical screening and computational model training. The project is currently seeking financial support for a technician and access to relevant chemical libraries, along with synthesis capabilities for computationally generated candidate molecules. This places the offering at an early-to-mid stage of translational readiness, with strong proof-of-concept in a related system and a clear path to validation in the target transporters.