Protein subunit interfaces: heterodimers versus homodimers.

Protein dimers are either homodimers (complexation of identical monomers) or heterodimers (complexation of non-identical monomers). These dimers are common in catalysis and regulation. However, the molecular principles of protein dimer interactions are difficult to understand mainly due to the geometrical and chemical characteristics of proteins. Nonetheless, the principles of protein dimer interactions are often studied using a dataset of 3D structural complexes determined by X-ray crystallography. A number of physical and chemical properties govern protein dimer interactions. Yet, a handful of such properties are known to dominate protein dimer interfaces. Here, we discuss the differences between homodimer and heterodimer interfaces using a selected set of interface properties.

The analyses report on the role of inter-subunit H-bonds in protein subunit association. The numbers of H-bonds vary in different studies. Previous studies also show that hydrophobic effect plays an important role in protein association [3, 7, 12], yet not as much as in protein folding. [3] There studies showed that protein interfaces are more hydrophobic than surfaces, but less than interior. Hydrophobic effect was measured by the buried non-polar surface area (or percent burial) of residue types. [3] The study showed that the ratio between buried hydrophobic and buried hydrophilic residues is approximately 1.

[3]
Hydrophobic residues (except ALA) and the charged residue ARG are predominantly present at proteinprotein interfaces with TYR and TRP having highest propensity. Interface size is yet another important property widely used to describe protein-protein interfaces and it is usually characterized by interface area. The number of interface residues is linearly correlated to interface area ( 96 . 0 ≥ r ) in several studies. [5,7] However, the mean number of interface residues varies between these studies. It is shown that the mean is 52 [7], 57 [5], 53.7 [14], 44.4 (for homodimers) and 42.2 (for heterodimers).
[9] Thus, the number of interface residues vary within a narrow rang of 42 and 57 in these studies.
Here, we created two extended datasets of mutually exclusive homodimers and heterodimers. We believe that these exclusive datasets can reduce data bias to differentiate heterodimer and homodimer interfaces.

Methodology: Creation of heterodimer and homodimer dataset:
A total of 2488 heterodimer candidates and 1324 homodimer candidates were downloaded from PDB (Protein Databank) and PQS (Protein Quaternary Structure Server). We then created a non-redundant dataset of 156 heterodimers and 170 homodimers (Table 1) such that they satisfy the following conditions. These include: (1) each chain ≥ 50 residues; (2) structures determined by x-ray crystallography; (3) resolution ≤2.5 Å; (4) the structure with the highest resolution was selected where more than one structure was available; (5) redundant entries were removed at a sequence similarity cut-off of ≥ 30%. [15] Calculation of interface parameters: Interface area ASA (accessible surface area) was calculated using NACCESS [16] with a probe radius of 1.4 Å and interface area is defined by ΔASA (change in ASA upon complexation from monomer to dimer state) as described elsewhere. [10] Inter-subunit H-bonds A hydrogen bond is a polar interaction between two electronegative atoms, where a donor and an acceptor participate. The number of H-bonds formed between subunits was calculated using the program HBPLUS.

Interface residues propensity
Interface residues show an Δ ASA (change in accessibility) of ≥ 5% upon complexation. Interface residue propensities were calculated using the percentage frequencies of 20 residues using the following functions:

Results and Discussion:
Dimer interactions are characterized by a large combination of physical-chemical parameters. Analysis of dimer structures can provide insight into the principles of protein-protein complexation and help develop models to predict interaction sites. The multi dimensional scaling method applied in a recent study reduced a large pool of interface parameters to a small set of six critical properties for heterodimers. [10] Zhanhua et al., 2005, showed that the six selected parameters were sufficient to describe subunit interfaces instead of the complete parameter space. Here, we use these selected set of properties to discuss the interface differences between 156 heterodimers and 170 homodimers. The properties used in this study are (1) interface residues, (2) interface H-bonds, (3) interface hydrophobicity, (4) interface residue-composition.  Interface residues: The number of interface residues is proportional to interface area. [5,7] Stronger protein subunit associations were generally associated with larger interface areas.
[11] In our study, the range of heterodimer interface residues varies from 18 to 162 with a mean value of 51. While, the range of homodimer interface residues extends from 15 to 308 with a mean value of 81. Like Hbonds, interface residues also varied with different studies and are affected by dataset size and data type. [5, 7, 9, 14] Hence, we created mutually exclusive datasets of homodimers and heterodimers for this analysis to reduce bias due to data type heterogeneity. Thus, we show that the amount of interface residues is significantly different for homodimers and heterodimers. The results also suggest that the previous studies are based on datasets biased with heterodimers. The relation between number of interface residues and monomer length is shown in Figure 1 E and F. They show that interface residues increase with both heterodimer and homodimer monomer length. However, the relation is causal. Figure 1 C and D show a causal relationship between interface area and monomer length for both homodimers and heterodimers. The mean interface residues are larger in homodimers than heterodimers. This is consistent with previous studies. [ 7,9] Interface residue composition: Several studies show the prevalence of certain types of residues at the dimer interfaces. [4,6,7,12,13] However, the significance of hydrophobic, hydrophilic, and charged residues at the interface of homodimers and heterodimers is not well documented. Figure 1G show the fractional distribution of hydrophobic, hydrophilic and charged residues in homodimer and heterodimer interfaces. Hydrophobic residues (M, F, P, A, B, L), except for I and G are dominant in homodimer interfaces. However, hydrophilic residues (W, C, H, Q, N, Y, S), except for T, are dominant in heterodimer interfaces. This observation is interesting and not surprising because homodimers being made of identical monomer subunits tend to associate by hydrophobic interactions. This is in contrast to the observation in heterodimer interfaces being made of non-identical monomer subunits, associating generally by hydrophilic interactions. Figure 1H, shows the ration of interface/surface and interface/interior residue propensity difference between heterodimers and homodimers. Interestingly, the ratio of interface to interior charged residues (D, E, K, R) is significantly larger in heterodimers compared to homodimers ( Figures 1H, 1I, 1J). On the other hand, the ratio of interface to interior hydrophobic residues (A, V, L, M, I, F) are prevalent in homodimers than in heterodimers ( Figures 1H, 1I, 1J). Similarly, hydrophilic residues (N, Q, H, Y, S, T) are prevalent in heterodimer interfaces ( Figures 1H, 1I, 1J). However, the propensity difference in the ratio of interface to surface hydrophobic/hydrophilic/charged for homodimers and heterodimers is almost zero ( Figure 1H).