Page 42 - Read Online

Alekseeva et al. Microbiome Res Rep 2023;2:10 https://dx.doi.org/10.20517/mrr.2023.06 Page 9 of 15

Three-dimensional models of 2D FN3 and ΔFN3.1 proteins were constructed on the basis of homologous
modeling using previously obtained crystal structures of eukaryotic proteins containing FN3 domains from
PDB: crystal structure of the full ectodomain of human Gp130 (PDB code 3L5H), human Ifnw-Ifnar ternary
complex (code PDB 3SE4), structure of a cytokine receptor complex (PDB code 4NKQ), and crystal
structure of PTPdelta ectodomain in complex with IL1RAPL1 (PDB code 4YH7). The 2D FN3 protein
model is formed by two linear FN3 domains dominated by antiparallel β-sheets [Figure 4A]. The ΔFN3.1
protein model includes five structural elements (epitopes), with three structural elements that tightly
interact with each other forming a V-shaped block [Figure 4C]. We conjecture that the binding between the
ΔFN3.1 protein of the B. longum GT15 strain and TNFα or other cytokines occurs specifically in this region
(pocket). Since no homologous structures were found for the CD FN3 protein fragment (C-terminal
region), a 3D model was built de novo. This model is formed by three structural elements, which are also
dominated by antiparallel β-layers [Figure 4B].

Polymorphisms of amino acid sequences of FN3 protein fragments of B. longum strains, the impact
of identified substitutions on predicted 3D structures
Previously, the comparative analysis of sequences encoding the regions 2D FN3 in the sequenced genomes
[26]
of B. longum subsp. longum allowed us to divide them into four groups . The revealed polymorphism of
amino acid sequences of ΔFN3.1 proteins in 203 sequenced genomes of B. longum subsp. longum also made
it possible to distinguish four groups of strains [Supplementary Figure 2]. Group 1 encompassed all strains
harboring identical amino acid sequences of ΔFN3.1, including that of the strain GT15 and was considered
as a reference. As for group 2, the strains contained a single substitution: 43 A→V. Strains belonging to
group 3 contained four substitutions: 43 A→V, 51 A→T, 417 P→Q and 424 A→T. Strains belonging to
group 4 contained three substitutions: 111 T→I, 417 P→Q and 424 A→T.

Using trRosetta software, we predicted the spatial structures of ΔFN3.1 proteins from all four groups of B.
longum subsp. longum strains [Figure 5].

The amino acid substitutions detected in ΔFN3.1 proteins lead to conformational changes in structural
elements of the predicted protein models. Significant differences were observed in the region of three
structural elements that form a V-shaped block, in which binding to cytokines can occur. Thus, amino acid
differences in the sequences of ΔFN3.1 proteins from four groups of B. longum subsp. longum can account
for the specificity of binding to cytokines.

Bioinformatics analysis of the occurrence of proteins containing FN3 domains in commensal
bacteria inhabiting the human intestinal microbiota
Proteins containing fibronectin domains carry out diverse functions and are widespread among all bacteria,
including commensal members of the human and animal intestinal microbiota. Previously, we identified
and characterized the species-specific PFNA operon, which was found only in the genus Bifidobacterium.
One of the key genes of the cluster is the fn3 gene, which encodes a protein harboring two type 3 fibronectin
domains (FN3) containing MCRs [22,23] .

Bioinformatics analysis of proteins with FN3 domains in the sequenced genomes of bifidobacterial strains
isolated from the human gastrointestinal tract showed that, apart from the FN3 domain encoded by the
PFNA operon, most bifidobacterial species also contain proteins annotated as glycosyl hydrolase (GH)
family 3 (beta-glucosidase); these proteins typically contain in the central or C-terminal regions an FN3
domain lacking MCRs. B. dentium strains each contained 12 proteins belonging to the GH family 3
[Supplementary Table 1].