*4.1. Model Building*

We built homology models for the structural ABC transporter core of hABCC6 and rAbcc6, including residues of TMD1, NBD1, TMD2, and NBD2. First, we generated a multiple sequence alignment (MSA) using MAFFT [54] including the sequences of the hABCC6, hABCC1, and hABCC5 proteins from multiple organisms (Table S1), retrieved from UniProtKB [55]. Based on the alignment of hABCC6 and rAbcc6 sequences with bAbcc1, homology models of hABCC6 and rAbcc6 were generated with Modeller 9v15 [56], using the inwardand outward-facing cryoEM structures of bAbcc1 as templates [22,23]. For hABCC6 and rAbcc6, 20 models were generated for both the inward- and the outward-facing states, by applying a slow refinement protocol and 20 cycles of simulated annealing as in our previous work on hABCC7 (CFTR) [57]. For the inward-facing state, the ABCC6 models were generated considering the presence of LTC<sup>4</sup> in the template. The final hABCC6 and rAbcc6 models for each conformation were chosen based on the Discrete Optimized Protein Energy (DOPE) value implemented in Modeller. The rAbcc6 and hABCC6 models are provided in the Supporting Information as PDB files. The MSA is provided in the form of two separate files, covering residues of TMD1 and NBD1 (see TMD1-NBD1\_Alignment.pdf), and residues of TMD2 and NBD2 (see TMD2-NBD2\_Alignment.pdf). These files were generated using Jalview [58] and the residues are colored according to the Clustal X coloring scheme implemented in Jalview. The residues highlighted in bold correspond to the residues investigated in the present study and the rAbcc6 numbering of amino acids is indicated. The residues of the TM helices of TMD1 and TMD2 are also annotated. The percentage amino acid identity among bAbcc1, rAbcc6, and hABCC6 was calculated on the TMD1-NBD1 and TMD2-NBD2 alignment using the id\_table command available in Modeler. Sequences of the hABCC1 and hABCC6 proteins shown in Figure S5 were retrieved from UniProtKB [55] and were aligned using Clustal Omega [59].

Electrostatic potential calculations were performed using the PDB2PQR and APBS webservers [60,61], using a pH of 7 and a NaCl concentration of 0.15 M. The electrostatic potential was visually mapped on the molecular surface of the models using UCSF Chimera [62], with a surface offset parameter of 1.4. Figure 5 was also generated using UCSF Chimera, after calculating the cavity volumes with the 3V webserver, using the default parameters for the Channel Finder module [34]. Other figures were generated using PyMOL [63].
