Structural and functional diversity calls for a new classification of ABC transporters

Members of the ATP-binding cassette (ABC) transporter superfamily translocate a broad spectrum of chemically diverse substrates. While their eponymous ATP-binding cassette in the nucleotide-binding domains (NBDs) is highly conserved, their transmembrane domains (TMDs) forming the translocation pathway exhibit distinct folds and topologies, suggesting that during evolution the ancient motor domains were combined with different transmembrane mechanical systems to orchestrate a variety of cellular processes. In recent years, it has become increasingly evident that the distinct TMD folds are best suited to categorize the multitude of ABC transporters. We therefore propose a new ABC transporter classification that is based on structural homology in the TMDs.

Members of the ATP-binding cassette (ABC) transporter superfamily translocate a broad spectrum of chemically diverse substrates. While their eponymous ATP-binding cassette in the nucleotide-binding domains (NBDs) is highly conserved, their transmembrane domains (TMDs) forming the translocation pathway exhibit distinct folds and topologies, suggesting that during evolution the ancient motor domains were combined with different transmembrane mechanical systems to orchestrate a variety of cellular processes. In recent years, it has become increasingly evident that the distinct TMD folds are best suited to categorize the multitude of ABC transporters. We therefore propose a new ABC transporter classification that is based on structural homology in the TMDs.
Keywords: ABC transporters; ATPases; cryo-EM; membrane proteins; molecular machines; phylogeny; primary active transporters; sequence alignment; structural biology; X-ray crystallography We suggest a new classification of the ABC transporter superfamily that is based on the TMD fold. Historically, first hints of the ABC protein superfamily came from sequence alignments of bacterial proteins that revealed highly conserved motifs in their ATPase domains [1]. The superfamily of ABC proteins was subsequently divided into three main classes [2][3][4]: exporters, nontransporter ABC proteins, and a third class consisting primarily of importers. The mammalian ABC systems, in particular, were grouped into seven subfamilies (ABCA to ABCG), based on NBD and TMD sequence homology, gene structure, and domain order [5][6][7]. It should be noted that ABCE and ABCF are not transporters, but exist as twin-NBDs without TMDs and are involved in mRNA translation control [8]. Detailed membrane topology and sequence analyses of exporters uncovered that, in contrast to the NBDs, the TMDs are polyphyletic and can serve as references to categorize ABC transporters into three distinct types (ABC1-3) [9,10]. According to this classification, the cystic fibrosis transmembrane conductance regulator (CFTR), the transporter associated with antigen processing (TAP), and the drug efflux pump P-glycoprotein (P-gp) belong to the ABC1 transporters; ABCG2 and ABCG5/G8 are members of the ABC2 group, which also comprises importers; and the macrolide translocator MacB is categorized as an ABC3 system. Yet, another classification scheme currently in use differentiates between the three types of importers predominantly found in prokaryotes [11][12][13][14] and two types of exporters, exemplified by Sav1866 [15] and ABCG5/8 [16], in addition to the LptB 2 FGtype [17,18] and MacB-type [19][20][21][22] transporters.
Our motivation for proposing a revised nomenclature stems from the recent wealth of ABC transporter structures determined by X-ray crystallography and single-particle cryo-electron microscopy, which has unveiled a remarkable diversity of TMD folds and evolutionary relationships between bacterial and eukaryotic/mammalian transporters [16][17][18][19][20][21][23][24][25][26]. This affluence of structural information provides the opportunity to introduce a universal nomenclature that  Tables 1 and 2). This classification is supported by quantitative analyses using TM-scores based on pairwise structural alignment of TMDs (Tables S1-S6, Fig. S1). The classification focuses on the transporterforming TMDs and does not consider additional membrane integrated domains, as for example observed in TAP1/TAP2 [27,28]. As before, types I-III of the new nomenclature cover the three different importer architectures (Fig. 1, Table 1, Tables S2 and S3; TM-score for pairwise structural alignment between the type III systems CbiQ (PDB code 5X3X) and EcfT from Lactobacillus brevis (PDB code 4HUQ): 0.736). It is noteworthy that prokaryotic importers typically operate with periplasmic, extracellular, or membrane-embedded substratebinding proteins whose structural features correlate with the type of TMD fold [29].
Based on the characteristic structure of the founding member Sav1866, which includes a domain-swapped TMD arrangement, type IV members of the new nomenclature have previously been classified as type I ABC exporters [15]. However, a significant and growing number of these ABC proteins have nonexporter functions, i.e., the gated chloride channel CFTR, the regulatory K ATP channel modules SUR1/2, the lysosomal cobalamin (vitamin B 12 ) transporter ABCD4 [30], the bacterial siderophore importers YbtPQ and IrtAB, and the cobalamin/antimicrobial peptide importer Rv1819c [31][32][33], as well as several type IV systems with importer functions in plants [34][35][36][37][38][39]. This striking functional diversity mediated by the same structural framework ( Fig. 1, Tables 1 and 2, Tables S4 and S5) makes the type IV ABC transporters stand out and is also the main reason why we suggest the more universal taxonomy based on structural principles.
Finally, LptB 2 FG and MacB are the founding members of type VI and type VII ABC transporters, respectively. We are aware that LptF and LptG have TMD folds that resemble type V members, and the TMD of MacB is reminiscent of type V systems and LptF/G. Yet, they exhibit distinct features that warrant classifications into separate groups. These include the lack of an amphipathic N-terminal 'elbow helix' and no extracellular reentrant helices between TM5 and TM6. In addition, MacB contains only four proper TM helices as well as an additional coupling helix, thereby defining a separate transporter architecture. In accordance with differences in TMD topologies, the LptFG and MacB transporters also display diverging dimerization interfaces. Thus, we have chosen to assign LptFG and MacB to separate types. This notion is corroborated by the TM-score-based quantitative analysis (Table S6 and Fig. S1). Of note, at the time of writing, publicly available, yet unpublished structures of the lipid transporter complex MlaFEDB of Gram-negative bacteria reveal some resemblance of MlaE to LptF/G and MacB. However, the number of TM helices differs between LptFG (six TM helices), MlaE (five TM helices), and MacB (four TM helices) [45][46][47][48] (Table S6 and Fig. S1).
We would like to point out that the classification of the mammalian ABC transporters into the ABCA-G subfamilies can be maintained as subcategories of type IV (subfamilies B-D) and type V (subfamilies A and G) within the new nomenclature (Table 2). We are also not proposing any changes to gene symbols. Most importantly, the new nomenclature based on TMD architecture can be universally applied to ABC transporters beyond their particular physiological functions and across the three domains of life. Hence, it allows any newly discovered transporter fold to be compared with the existing types and seamlessly incorporated into the classification scheme, possibly as a new type. Since the new nomenclature depends on TMD architecture, it requires structural information in order to classify new transporter systems. At the same time, we regard the nomenclature as a dynamic platform that can be upgraded, adjusted, or refined whenever necessary due to novel insights that add extra dimensions to our understanding of ABC systems.
The recent advances in structural mapping of the diverse superfamily of ABC transporters have revealed a vast area of mechanistically uncharted territory. One key objective of future research should be to fully comprehend how type IV systems perform so many different functions, i.e., as importer, exporter, lipid floppase, ion channel, and regulator, by employing a single structural scaffold. However, we do not exclude that other types might turn out to be as functionally diverse as type IV systems. Exploring the different modes of operation and accompanying conformational landscapes [49] and the dynamics of the multifarious ABC systems will require integrative experimental

Type VI Type VII
Exporter ABCG2

Ion channel CFTR (ABCC7)
Exporter PCAT1  experiments. We are confident that future studies of such kind will provide major new insights into the mechanisms of these fascinating molecular machines.   (2020). The discussion included a vivid exchange of thoughts via hundreds of emails and remote video sessions during the global COVID-19 pandemic. In addition to the authors listed, we received positive feedbacks on our proposed classification from several further leading scientists in the ABC transporter field. Yet, as they felt that their contribution was too small, they decided not to accept authorship.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Fig. S1. Phylogenetic tree based on TM-scores of structural TMD alignments. Table S1. TM-scores based on pairwise structural alignment of representatives of the different TMD types. Table S2. TM-scores based on pairwise structural alignment of type I TMDs. Table S3. TM-scores based on pairwise structural alignment of type II TMDs. Table S4. TM-scores based on pairwise structural alignment of type IV TMDs in inward-facing conformations. Table S5. TM-scores based on pairwise structural alignment of type IV TMDs in (semi-) occluded/outward-facing conformations. Table S6. TM-scores based on pairwise structural alignment of type V, VI, and VII TMDs a .