In Silico Analysis of Phosphomannomutase-2 Dimer Interface Stability and Heterodimerization with Phosphomannomutase-1

Abstract

Phosphomannomutase 2 (PMM2) catalyzes the interconversion of mannose-6-phosphate and mannose-1-phosphate, a key step in the biosynthesis of GDP-mannose for N-glycosylation. Its deficiency is the most common cause of congenital disorders of glycosylation (CDGs), accounting for the subtype known as PMM2-CDG. PMM2-CDG is a rare autosomal recessive disease characterized by multisystemic dysfunction, including cerebellar atrophy, peripheral neuropathy, developmental delay, and coagulation abnormalities. The disease is associated with a spectrum of pathogenic missense mutations, particularly at residues involved in dimerization and catalytic function (i.e., p.Phe119Leu and p.Arg141His). The dimerization of PMM2 is considered essential for enzymatic activity, although it remains unclear whether this supports structural stability alone, or whether both subunits are catalytically active-a distinction that may affect how mutations in each monomer contribute to overall enzyme function and disease phenotype. PMM2 has a paralog, phosphomannomutase 1 (PMM1), which shares substantial structural similarity-including obligate dimerization-and displays mutase activity in vitro, but does not compensate for PMM2 deficiency in vivo. To investigate potential heterodimerization between PMM1 and PMM2 and the effect of interface mutations over PMM2 dimer stability, we first assessed the likelihood of their co-expression using data from GTEx and the Human Protein Atlas. Building on this expression evidence, we modeled all possible dimeric combinations between the two paralogs using AlphaFold3. Models of the PMM2 and PMM1 homodimers were used as internal controls and aligned closely with their respective reference biological assemblies (RMSD < 1 Å). In contrast, the PMM2/PMM1 heterodimer model, the primary result of interest, showed high overall confidence (pLDDT > 90), a low inter-chain predicted alignment error (PAE∼1 Å), and robust interface confidence scores (iPTM = 0.80). Then, we applied PISA, PRODIGY, and mmCSM-PPI to assess interface energetics and evaluate the impact of missense variants specifically at the dimerization interface. Structural modeling suggested that PMM2/PMM1 heterodimers were energetically viable, although slightly less stable than PMM2 homodimers. Interface mutations were predicted to reduce dimer stability, potentially contributing to the destabilizing effects of disease-associated variants. These findings offer a structural framework for understanding PMM2 dimerization, highlighting the role of interface stability, paralogs co-expression, and sensitivity to disease-associated mutations.

Keywords: PMM2-CDG; PPI; human PMMs; structural bioinformatics.

Figure 1

Human phosphomannomutase 1 (PMM1, tan, UniProt ID: Q92871) and 2 (PMM2, gold, UniProt ID: O15305). (a) Global pairwise alignment of the two paralogs using Needleman–Wunsch algorithm. (b) Structural superimposition of subunit crystal structures (2FUC for PMM1 and 7O4G for PMM2). Single, fully conserved residue “|”; groups of strongly similar properties—(>0.5 Gonnet PAM250) “:”; groups of weakly similar properties—(=<0.5 Gonnet PAM250) “.”.

Figure 2

PMMs’ expression levels across tissues. (a) Scatter plot of PMM1 vs. PMM2 median transcript abundance (TPM) across GTEx tissues. Points associated with protein co-detection in the Human Protein Atlas (HPA) immunohistochemistry data are colored in light blue. The red dashed line marks the TPM threshold used to define robust transcript-level expression. The cluster of brain tissues is marked with a convex hull. Tissues with TPM > 10 for both genes and/or confirmed protein co-detection are labeled. (b) Violin plots of PMMs’ expression levels across tissues.

Figure 3

Models of dimer structure produced by AlphaFold3 server. Each model is associated with its PAE matrix (blue: low PAE, yellow: high PAE) and with the detail of interacting residues spanning ≤4 Å with predicted aligned error ≤5 Å (interactions marked by thick red pseudobonds). (a) PMM1 homodimer, (b) PMM2 homodimer, (c) PMM1/PMM2 heterodimer.

Figure 4

Comparison of interface composition and stability across the possible PMM dimers. For each dimer, the interface is shown as a cartoon, and residues involved in the formation of the most prominent salt bridges are displayed as sticks. Data from PISA analysis are reported in the figure.

Figure 5

Analysis of contacts at the interface for the PMM1 homodimer (a), the PMM1/PMM2 heterodimer (b), and the PMM2 homodimer (c).

Figure 6

Left: Comparison of predicted $\Delta \Delta G^{\mathrm{binding}}$ values calculated using mmCSM-PPI. Colors indicate mutation classification while marker shape denotes presence (×) or absence (∘) of each variant in population datasets (gnomAD). Muted red shading marks the region of significant destabilization ($|\Delta \Delta G^{\mathrm{binding}}|$ > 2 kcal/mol, [18]). Right: Structural representation of the PMM2 dimer interface. Interface residues are shown as surfaces, colored according to their clinical annotation status in ClinVar, labeled based on UniProt residue numbering. Residues are classified as follows: never associated (not reported in ClinVar), pathogenic (at least one variant at that position is labeled pathogenic), and uncertain (only variants of uncertain significance reported). Coloring reflects these categories as described in the legend.