Molecular insights into the dual-glycoprotein surface layer of the oral bacterium Tannerella serpentiformis

Surface (S-) layers are widespread proteinaceous lattices that form the outermost cell envelope in many bacteria and archaea, yet the molecular organization of multi-protein S-layers remains poorly understood. Tannerella serpentiformis, a long, segmented oral commensal, expresses two heavily glycosylated S-layer proteins (TssA and TssB) that are homologous to the dual S-layer system of the periodontal pathogen Tannerella forsythia. However, the structural arrangement and assembly principles of the T. serpentiformis S-layer have not previously been resolved. Here, we combined electron microscopy, sequence conservation analysis, glycosylation mapping, and integrative structural modelling to reconstruct plausible molecular architectures of the TssA-TssB lattice. Surface imaging confirmed a square (p4) lattice with a unit cell size of ∼10-12 nm and an outer membrane-to-surface spacing of ∼21 nm, paralleling the T. forsythia S-layer. Using AlphaFold2 Multimer, AlphaFold3, and HADDOCK docking, we generated and evaluated TssA-TssB heterodimers and assembled p4-symmetric unit cell candidates. AlphaFold2 Multimer favoured an elongated sequential dimer (ipTM + pTM = 0.71), whereas both AlphaFold3-when supplied with experimentally verified glycosylation sites-and HADDOCK 2.4 prefered a side-by-side arrangement (ipTM + pTM = 0.55). Symmetric docking of these dimers produced a pit-like unit cell candidate with a ∼3.5 nm wide pore in the center in which TssA forms a compact central core and TssB creates a peripheral ring, with glycans projecting outward from the surface and conserved residues facing the cell envelope. This architecture is consistent with lattice flexibility observed in other S-layer systems and allows nutrient exchange through the central pore. Our results suggest a plausible molecular model for the T. serpentiformis S-layer and establish a potential framework for structurally resolving multi-protein S-layers through glycan-informed integrative modelling.