Supplementary data 5

Structural context of classifier selected features for:

(A) Rattus norvegicus FCGRT

(B) Mus musculus RAE1B

Each MHC-I-like heavy chain (in gray) comprises the [D1] (in the back) and [D2] (in the front) extracellular domains. B2M (in green) is in complex with FCGRT, but virtually placed for RAE1B (see Method). The C-LIKE extracellular domain of FCGRT is not shown. Each feature is labeled with domain, position and amino acid observed in the 3D structure. (A) (3fru coordinate file) The features located in potential B2M contact zone are colored in yellow and those located out of this zone are in cyan. (B) (1jfm coordinate file) The features located in potential B2M contact zone are colored in green, the others are in orange; [D1] 21 W and [D1] 12 K (in green, in the back) are not labeled.



Dynamic visualization of the structural context of the 18 selected features

Structural context of the features conserved in B2M bound (9 features) or unbound (9 features) MhcSF proteins are analyzed using the 3D structure of Rattus norvegicus FCGRT (3fru) and Mus musculus RAE1B (1jfm), respectively. Installation of PyMOL software and download of PyMOL script and PDB files are required for dynamic visualization.

(i) Install the PyMOL software:

(ii) Download the PyMOL scripts for visualization of:

(iii) Download the PDB files of:

(iv) Execute PyMOL scripts



Correspondence between IMGT and PDB numbering

The MHC-I-like heavy chain corresponds to the chain A in each of the two 3D structures (Rattus norvegicus FCGRT in 3fru and Mus musculus RAE1B in 1jfm).

IMGT domain and position numbers of features conserved in B2M bound proteins

Corresponding numbers in chain A

of 3fru PDB file

[D1] 27

29

[D1] 32

34

[D1] 51

51

[D1] 86

87

[D2] 10

97

[D2] 27

113

[D2] 32

118

[D2] 83

170

[D2] 85

172

IMGT domain and position numbers of features conserved in B2M unbound proteins

Corresponding numbers in chain A

of 1jfm PDB file

[D1] 8

8

[D1] 11

11

[D1] 12

12

[D1] 21

21

[D1] 25

25

[D1] 35

35

[D1] 74

72

[D1] 88

86

[D2] 39

125



Method for B2M placement in Mus musculus RAE1B 3D structure

In order to understand how the 9 features conserved in B2M unbound MhcSF proteins prevent their binding to B2M, we virtually placed B2M in Mus musculus RAE1B 3D structure. We used ProFit program to fit the 3D structure of Mus musculus RAE1B protein (1jfm, considered as MOBILE) to the 3D structure of Mus musculus H2-D1 (1wbx, considered as REFERENCE), according to their sequence alignment in Protein sequence alignment (IMGT web site). The H2-D1 protein (MHC-I) is bound to B2M, and in complex in the 1wbx coordinate file. The fitting is processed only on alpha carbon atoms (ATOMS CA) of the [D1] and [D2] domains, and results in structures with 4,17 Å rms. The new 3D coordinates of RAE1B have same orientation than those of H2-D1, allowing us to place virtually B2M (chain B in 1jfm_B2M 3D coordinate file).