How Choice of Model Membrane Affects Protein–Glycosphingolipid Interactions: Insights from Native Mass Spectrometry

Abstract

Interactions between glycan-binding proteins (GBPs) and glycosphingolipids (GSLs) are involved in numerous physiological and pathophysiological processes. Many model membrane systems are available for studying GBP-GSL interactions, but a systematic investigation has not been carried out on how the nature of the model membrane affects binding. In this work, we use electrospray ionization mass spectrometry (ESI-MS), both direct and competitive assays, to measure the binding of cholera toxin B subunit homopentamer (CTB₅) to GM1 ganglioside in liposomes, bilayer islands [styrene maleic acid lipid particles (SMALPs), nanodiscs (NDs), and picodiscs (PDs)], and micelles. We find that direct ESI-MS analysis of CTB₅ binding to GM1 is unreliable due to non-uniform response factors, incomplete extraction of bound GM1 in the gas phase, and nonspecific CTB₅-GM1 interactions. Conversely, indirect proxy ligand ESI-MS measurements show that the intrinsic (per binding site) association constants of CTB₅ for PDs, NDs, and SMALPs are similar and comparable to the affinity of soluble GM1 pentasaccharide (GM1_os). The observed affinity decreases with increasing GM1 content due to molecular crowding stemming from GM1 clustering. Unlike the smaller model membranes, the observed affinity of CTB₅ toward GM1 liposomes is ∼10-fold weaker than GM1_os and relatively insensitive to the GM1 content. GM1 glycomicelles exhibit the lowest affinity, ∼35-fold weaker than GM1_os. Together, the results highlight experimental design considerations for quantitative GBP-GSL binding studies involving multisubunit GBPs and factors to consider when comparing results obtained with different membrane systems. Notably, they suggest that bilayer islands with a low percentage of GSL, wherein clustering is minimized, are ideal for assessing intrinsic strength of GBP-GSL interactions in a membrane environment, while binding to liposomes, which is sub-optimal due to extensive clustering, may be more representative of authentic cellular environments.