The largest protein family of mitochondrial solute transporters is the SLC25 (mitochondrial carrier) family. In human there are a total of 53 members that fulfil the vital roles of uniport or exchange of ions, metabolites and other solutes across the IMM [28]. It has been recognized early on that the sequence of the ADP/ATP translocase (SLC25A4) has a repeating sequence element that contains two hydrophobic segments and repeats 3 times in the sequence [29]. Such an internal repeat symmetry is commonly found in membrane transporters [30]. Indeed, such features were later found in several other members of the protein family [31,32,33,34], along with conserved proline, glycine and acidic amino acid residues [35]. The conserved residues were later compiled into a consensus characteristic or “signature motif” for mitochondrial carriers, Px(D/E)xx(K/R) [36], and the conserved charged residues were indeed found to take part in specific conserved salt-bridge contacts, termed the matrix network [37,38,39,40,41,42], in the 3D structure of the proteins [43,44]. The signature motif can be extended in the C-terminal direction to include a proximal glutamine (Q) residue, which helps stabilize the salt-bridge contacts on the matrix side of the carrier, forming the “Q brace” [44,45]. There is a similar cluster of charged residues on the intermembrane side of the even-numbered TM helices, forming the cytoplasmic salt-bridge network, where the charged residues of a (Y/F)(D/E)xx(K/R) motif engage in salt-bridge contacts [42,44]. This network is stabilized by the so-called tyrosine (Y) brace, formed by the hydrogen bonding of the tyrosine residues of the motif to these salt bridges [45]. A further sequence motif, (Y/W/L/F)(K/R)GxxP, present in the connecting loop between each short matrix helix and the following transmembrane helix, has been described, replacement of which disrupts the function of the transporter [46]. In addition, close helix-helix contacts in the matrix-facing state are formed by two conserved sequence motifs, π G π x π G on the odd-numbered, and π xxx π on the even-numbered helices, where π stands for a residue with a small side-chain [45,47,48]. Further details of these sequence motifs, including their roles in the transport mechanism and a functional interpretation of individual amino acid residues is covered in detail by Kunji et al. in this special issue. In particular, disease-causing mutations in context of the 3D structure of SLC25 carriers and their sequence motifs are extensively reviewed by Palmieri et al. in the present review series [49].

Interestingly, the sequence motif (QYKGxxDCxRK) in the short matrix helices has also been described, which is only conserved in a subset of mitochondrial carriers, such as ADP/ATP (SLC25A4–6, SLC25A31), aspartate/glutamate (SLC25A12–13), ornithine (SLC25A2, SLC25A15), glutamate (SLC25A18, SLC25A22), and carnitine (SLC25A20) carriers, one ATP/P_i carrier (SLC25A24) and three carriers with unknown function (SLC25A9, SLC25A34, SLC25A45) [50]. This motif was proposed to harbor residues that go through post-translational modification thereby locally altering the protein structure and thus modulating function. One example for this is acetylation at K163 of SLC25A5 (AAC2), corresponding to the last residue of the motif, while the cysteine residue of the motif might interact with oxidizing/reducing agents [50]. Further investigation is required to reveal the precise functional role of this sequence motif.

Most SLC25 family proteins do not contain an MTS at their N-termini. Instead, they seem to hold mitochondrial targeting information in all three segments of their three-fold repeat sequence [51,52]. SLC25 proteins are embedded into the IMM by the TIM22 machinery, as described in the previous chapter. It was suggested that the three repeating segments of SLC25 proteins act in a cooperative manner to facilitate receptor binding and translocation into mitochondria [53], nevertheless, a single unit consisting of a matrix loop and the following transmembrane helix is enough for mitochondrial localization [52]. The net positive charge of the short matrix loop helices have been shown to be essential for import into mitochondria, and these regions are thought to interact with the Tom40 channel component of the TOM complex, which effectively functions as a selectivity filter [52].

Interestingly, certain SLC25 proteins, such as the mitochondrial phosphate carrier (SLC25A3) and citrate/tricarboxylate carrier (SLC25A1) have been proposed to harbor an N-terminal MTS, based on physicochemical composition of their N-terminal sequences and observation of mature protein forms truncated at the anticipated cleavage site [31,54]. This is also partially supported by TargetP-2.0 predictions, which report an MTS of 49 amino acids with likelihood 0.5749 for SLC25A3, exactly as anticipated from experiments where a protein fragment N-terminally truncated at the same position was identified [31]. For SLC25A1, the presence of an MTS was predicted with likelihood 0.3753. Whether a functional N-terminal MTS is present in these proteins has not been investigated experimentally. For all other human SLC25 proteins, the likelihood of an MTS at the N-terminus was less than 0.17 according to TargetP-2.0 predictions.

3. SLC54—Mitochondrial Pyruvate Carriers (MPC)

The mitochondrial pyruvate carriers have been identified as IMM transporters responsible for pyruvate uptake into the mitochondria [55,56]. MPC1 (SLC54A1) and MPC2 (SLC54A2) function as heterodimers [56,57], while humans as well as other placental mammals also contain a paralog of MPC1 called MPC1L (SLC54A3), the two sharing 48.2% sequence identity in human [58]. A detailed discussion of the biological role and function of MPCs and their links to disease can be found in a comprehensive review by Martinou et al. and Taylor et al. in this special issue.

MPCs have been predicted to harbor 2–3 transmembrane helices (TMHs) [55,56,59]. Interestingly, they are unrelated to SLC25 carriers and instead have been shown to be homologous to the 3-TMH repeating element in SWEET (“Sugars will eventually be exported transporters”) transporters [60,61], which show a 3 + 1 + 3 TMH architecture [62,63]. Proteins from the SWEET family also exist as homodimers of half-transporters encompassing the 3-TMH repeat, called SemiSWEET [62,64]. While there are relatively few studies on the structure and transport mechanism of MPCs, the structure and mechanism of SWEET and SemiSWEET transporters are quite well described [62,65,66,67,68,69]. Based on the suggested similarity, it can be speculated that the structure of a functional MPC transporter is similar to those of the homodimers of SemiSWEET proteins, or a single protomer of a SWEET-fold transporter. Based on this proposed similarity and the available structures for SWEET and SemiSWEET transporters, it should be possible to interpret disease-causing mutations in a structural context in the future, such as L79H and R97W in human SLC54A1/MPC1 [55,70,71].

Nevertheless, even without a structural context, one can analyze residue conservation in MPC proteins based on sequence alignment and the help of the “MPC” (PF03650) domain from the Pfam database [72]. The information of residue conservation encoded by profile hidden Markov models (HMM) of Pfam domains can be visualized as a HMM logo by a suitable software, such as Skylign [73] (Figure 2). We submitted the amino acid sequences of human SLC54A1–3 to three different methods to predict the location of transmembrane regions (HMMTOP [74], SPOCTOPUS [75], MEMSAT-SVM [76]) and secondary structural elements (PSIPRED [77]). Combining information on conservation with transmembrane region prediction pinpoints possible conserved charged/polar residues in transmembrane regions, which would imply that they might have a functional or structural role. Such residues may be S52/S54/S68, R68/R70/R84, H84/H86/N100 in MPC1/MPC1L/MPC2, respectively (Figure 2). Interestingly, residues N33/S35/K49 in MPC1/MPC1L/MPC2, respectively, represent a position which shows slight preference for polar/charged residues, but is asymmetric between MPC1/SLC54A1 and MPC2/SLC54A2 proteins, which could hint at a possible substrate-binding role [42]. We can also observe based on Figure 2 that the disease-associated mutation R97W [55] modifies an amino acid at a location that is considerably conserved in the family with a preference for basic sidechains, explaining the deleterious effects of the mutation. Intriguingly, the other currently known point mutation, L79H, is located at a poorly conserved position, and so is not expected to have a direct impact on the structure or function of the mature protein monomer. Several positions might contain structurally important residues with a clear preference for aromatic sidechains, such as F27/F29/F43, W28/W30/W44, W34/W36/W50, F66/F68/W82, F69/F71/Y85 in MPC1/MPC1L/MPC2, respectively. The detailed investigation of these residues in future studies might reveal more about their role in transporter function.

Figure 2. Sequence analysis of human MPC (SLC54) proteins. The diagram above the alignment shows the part of the HMM logo corresponding to the sequence (see text). At each position in the HMM logo, higher columns correspond to higher sequence conservation, and letters are drawn proportional to their frequency of occurrence at that position. Vertical lines in the HMM logo show positions with non-zero insertion frequency. Transmembrane regions predicted by three different methods (HMMTOP, SPOCTOPUS, MEMSAT-SVM) are marked by black lines. Green lines mark transmembrane regions predicted to by pore-lining by MEMSAT-SVM. Secondary structure as predicted by PSIPRED is shown as magenta lines ( α -helix) and golden lines ( β -strand). Uniprot sequence identifiers are shown next to protein names for each sequence.

4. SLC56—Sideroflexins

Sideroflexin 1 (SLC56A1/SFXN1) has been described as the gene whose defects are responsible for the flexed-tail mouse phenotype with sideroblastic anemia [78]. Later, it was found that SLC56A1/SFXN1 plays a role in serine transport into the mitochondria, which in turn fuels one-carbon metabolism [79]. SLC56A3/SFXN3 and SLC56A2/SFXN2 can compensate for this function, as well as yeast and Drosophila homologues, indicating an ancestral function, while SLC56A4/SFXN4 and SLC56A5/SFXN5 were unable to do so, indicating a possible altered transport rate or substrate selectivity [79]. Regarding their mitochondrial targeting, sideroflexins do not show any canonical MTS [78], and instead are targeted to the mitochondria via the TIM22 import complex [80,81,82], similarly to SLC25 proteins.

Very little is currently known about the structure and function of sideroflexin transporters. No structural homologs exist and, interestingly, no internal repeat symmetry has been described for SLC56 proteins. Five transmembrane helices were proposed to be conserved among family members and a few consensus motifs have been described [78]. Nevertheless, sequence analysis can potentially hint at functionally important regions in SLC56 proteins. Since the 3D structure and thus the membrane-spanning regions of SLC56 proteins are not known, we submitted the amino acid sequences of human SLC56A1–5 to various methods to predict the location of transmembrane regions and secondary structural elements as for MPC/SLC54 proteins. Interestingly, while four regions are more-less robustly predicted as transmembrane by various methods (Figure 3, alignment positions 140–160, 175–195, 225–250, 260–285, marked as Region 3–6, respectively), there appears to be significant inconsistency in predicting the first TM helical region (alignment region 75–125). The predictions by HMMTOP suggest two distinct regions (alignment positions 75–100 and 105–125, marked Region 1 and 2), and a third region overlapping with the first two (alignment position 90–110). However, the third region seems less likely due to the lack of sequence conservation as reported by the HMM logo, and the probability of residue insertions in alignment region 100–105 (Figure 3). SPOCTOPUS consistently does not report a TM segment in alignment region 75–125, while MEMSAT-SVM consistently reports the presence of a TM segment in Region 1 (Figure 3). Interestingly, MEMSAT-SVM also reports a sixth TM segment in Region 2 in SLC56A4, which is also predicted to be transmembrane by HMMTOP. This region is predicted to be helical by PSIPRED, and marked as pore-lining by SPOCTOPUS, likely hinting at its amphipathic nature. Region 2 also contains a sequence motif that is seemingly conserved in the protein family, and that has been described as “asparagine rich” already after the identification of SLC56A1/SFXN1 [78]. This motif could be described by the consensus WQWxNQSxNxxxN motif, where polar residues N, S and Q are in conserved positions. According to PSIPRED predictions, occasional coil and strand content can occur near the second Q and second N residues (Figure 3). Such local distortions of helical geometry typically signal the location of substrate-binding sites in transmembrane transporter proteins [42,43,83,84], which, combined with the conservation pattern of residues, suggests that this region could have a functional role. Given that this putative functional region is often missed by transmembrane segment prediction software, while region 75–100 is consistently predicted by MEMSAT-SVM as transmembrane, combined with the presence of a non-conserved, insertion-prone region at 100–105, it is tempting to speculate that in fact regions 75–100 and 105–125 constitute two independent transmembrane helices, leading to an overall topology with six transmembrane helices in all SLC56 proteins.

Analysis of the other regions proposed as transmembrane also show remarkably conserved polar or charged residues that might play a role in shaping transport function. In Region 1, a conserved Arg residue is apparent from the HMM logo (Figure 3), corresponding to R92/R91/R91/R109/R108 in SLC56A1–5, respectively. Region 3 contains a conserved aromatic/hydrophobic residue (Y151/Y150/Y150/L166/Y167 in SLC56A1–5, respectively). Region 4 shows the consensus sequence pattern RxVPFxxVxxAxxxNxxxMR, of which P181/P180/P180/P201/P204 (SLC56A1–5, resp.) are conserved, indicating a possible structural role, while the N192/N191/N191/N212/N215 residues in the second half of the proposed transmembrane helix might play a functional role, and the highly conserved basic residue R197/R196/R196/R217/R220, proposed to be at the edge of the TM helix, might play a role in gating. Region 5 contains a conserved Ser-Arg (SR) motif with R233/R232/R232/R253/R256 highly preferred to be basic. In Region 6, the sequence pattern PxAxAxFPQ is apparent, with the first proline residue (P281/P280/P280/P301/P304, SLC56A1–5, resp.) being highly conserved, suggesting the presence of a functionally or structurally relevant kink in the transmembrane helix in this region. Interestingly, regions outside the predicted transmembrane segments also contain seemingly conserved sequence motifs. The region just N-terminal to the first predicted transmembrane segment (Region 1) harbors the HPDT motif that is well conserved in human sideroflexin proteins and has been recognized early on [78]. The proposed loop between Regions 4 and 5 also contains some remarkably conserved amino acid residues, such as E200/E199/E199/E220/E223, S218/S217/S217/S238/S241, (in SLC56A1–5, resp.), which might play a functional role; and G204/G203/G203/G224/G227, which could be important in maintaining structure, such as close helix-helix contacts. Finally, the last five C-terminal residues of human sideroflexin proteins contain the consensus sequence FNKGL of unknown function, with a highly conserved glycine residue. Notably, no disease-linked genetic defects that modify a single amino acid position have been described for SLC56 proteins so far, instead, only mutations causing frameshift and/or premature termination have been reported [85].

Since the above descriptions are only based on computer predictions, they are speculative at this point. However, given the current scarcity of structure-function studies on sideroflexin proteins, these amino acid positions can be helpful to identify functionally relevant hot spots. These hypotheses, however, await validation by future experimental studies.

5. The SLC55/LETM Mitochondrial Cation/Proton Exchanger Family

LETM1 (SLC55A1) has had a controversial role, as it was first proposed to be a part of the mitochondrial K⁺/H⁺ exchanger (KHE) pathway [86,87]. However, later, through a genome-wide genetic screen, it was found to be responsible for altered Ca²⁺ levels in mitochondria [88], and subsequently the purified LETM1 protein in liposomes was shown to mediate Ca²⁺/H⁺ exchange [88,89,90]. However, later it was argued that K⁺/H⁺ and unspecific and electroneutral cation/H⁺ exchange better explain the effect seen upon knockout of LETM1 in model cells and the effects seen in the mitochondria of model animals and patients with LETM1 deletion [91]. For a more detailed review of mitochondrial K⁺ homeostasis, see the comprehensive reviews by Szabo et al. and Szewczyk et al. in this special issue. Nevertheless, the structure of LETM1 was proposed to be a hexamer and the predicted single TMH (alignment region 60–80 in Figure 4 [86]) contains a conserved acidic residue (E221), the mutation of which causes loss of the ability for the protein to take up Ca²⁺ into liposomes [90], which can be interpreted as E221 being part of the substrate-binding site. Interestingly, this residue is not conserved in LETMD1/SLC55A3 (N151, Figure 4), while other polar residues are present in the transmembrane region, such as S145, hinting at a substrate spectrum or a function that is likely different from LETM1/SLC55A1 and LETM2/SLC55A2. Nevertheless, the transport mechanism of the protein remains unknown, even though it has been shown that a change in external pH causes conformational changes in LETM1 [90]. In the proposed transmembrane region of SLC55 proteins, two proline residues seem to be conserved based on our sequence alignment and the HMM logo (Figure 4, P211/P180/F140 and P219/P188/P148 in SLC55A1–3, respectively). Conserved proline residues often signal positions with structurally important kinks in helices [92,93]. LETM1 contains the mitochondrial signal peptide [90], similarly to LETM2 (likelihood 0.3338) and LETMD1 (likelihood 0.9889) according to TargetP-2.0 [94] predictions, indicating that SLC55 proteins likely get translocated into the mitochondria via the TIM23 translocation complex. Currently, no single-point mutations linked to disease have been reported for SLC55 proteins.

Figure 4. Sequence analysis of human LETM proteins (SLC55 family). HMM logo is shown over the alignment, based on the “LETM1” HMM model from Pfam (see text). Only the alignment region covered by the HMM model is shown. Certain non-conserved residue positions with ambiguous matches with the protein sequences have been removed from the HMM logo.

6. ATP-Binding Cassette (ABC) Transporters in Mitochondria

Currently, 3 of the 48 human ATP-binding cassette (ABC) transporters have been shown to be present in mitochondria, which are ABCB7, ABCB8 and ABCB10 (also known as ABC7 [95], M-ABC1 [96], and M-ABC2 [97], respectively). ABCB6 has also been suggested to be a mitochondrial transporter [98], but this has later been challenged [99,100]. The three mitochondrial ABC transporter proteins, in contrast to solute carriers, contain a mitochondrial targeting sequence at their N-termini [95,96,97,101], indicating that they are most likely translocated into the mitochondrial inner membrane via the TIM23 machinery. The presence of an N-terminal cleavable targeting signal is also supported by TargetP-2.0 predictions (likelihoods 0.9634, 0.7184, and 0.207 for human ABCB7, ABCB8, and ABCB10, respectively), while human ABCB6 was not predicted to contain a signal peptide (likelihood 0.0001).

A yeast ABC transporter protein (Mdl1) showing considerable sequence similarity to human ABCB proteins [102], was shown to use both the conservative Oxa1-mediated import machinery and lateral release from the TIM23 complex to translocate to the IMM [20]. Based on this, it is likely that human ABCB proteins also use the same pathway for mitochondrial targeting. In addition, the homologous yeast Mdl2 protein and Atm1 (human ABCB7 ortholog) have also been suggested to employ OXA-mediated membrane insertion into the IMM [103]. Nevertheless, while it is plausible that human mitochondrial ABCB proteins also employ a similar mechanism, this has not yet been experimentally studied.

ABCB transporters in the mitochondria are so-called “half transporters”, consisting of one nucleotide-binding domain (NBD) and one transmembrane domain (TMD), with homodimeric complexes forming the functional transporter unit. Recent structural studies of human ABCB10 [104] and yeast Atm1 [105] have highlighted conserved sequence motifs that are located along cavities forming the putative substrate-binding site in the TMD. Interestingly, they are not lying in overlapping locations in the two different proteins. For human ABCB7, residues R315, R319, N378, N425, T429, R432 and E433 (corresponding to R280, R284, N343, N390, S394, R397, D398 in yeast Atm1, respectively) in TMH4, TMH5 and TMH6 near the cytoplasmic membrane interface have been proposed to take part in substrate binding based on the yeast Atm1 structure with bound glutathione [105]. While one of these positions, E433 have been found to be mutated to lysine in patients with X-linked sideroblastic anemia (XLSA) [106], biochemical validation of these binding site residues is still missing [105]. For human ABCB10, a conserved signature sequence of (N/I)xxR (containing N229 and R232) and NxxDGxR (containing N289, D292, R295) in TMH2 and TMH3, respectively, were found and have been proposed to take part in substrate binding [104]. In the case of ABCB10, the missing identity of the transported substrate hampers biochemical studies on the functional role of individual residues. It has been proposed that ABCB10 transports an intermediate in the heme biosynthesis pathway [107]. Structure-function studies elucidating functionally relevant residues or sequence motifs of human ABCB8 are still missing. In neither ABCB8 nor in ABCB10 have disease-linked single-point mutations been described yet.

7. Mitochondrial Calcium Transport via SLC8 Family

SLC8B1 (NCLX, Na⁺/Ca²⁺/Li⁺ exchanger) was identified as the ion exchanger protein responsible for the exit of Ca²⁺ from the mitochondria [108]. SLC8B1 is thought to exchange 3 Na⁺ ions for 1 Ca²⁺ ion based on similarity to other Na⁺/Ca²⁺ exchangers [109,110,111], and is unique in the property that Li⁺ ions can replace Na⁺ ions in transport [112]. Recently, it has been reported that Na⁺ taken up into the mitochondrial matrix by SLC8B1 in exchange for Ca²⁺ derived from calcium precipitates upon matrix acidification controls hypoxic signaling via the mitochondrial respiratory chain [113]. Specifically, it was reported that the Na⁺ imported into the matrix reduces membrane fluidity through interaction with phospholipids. This in turn was shown to lead to the generation of reactive oxygen species (ROS) by altering certain elements of the electron transport chain, thereby promoting an adaptive short-time elevation of mitochondrial complex III-dependent ROS production during acute hypoxia. Indeed, inhibition of Na⁺ import via SLC8B1 was sufficient to prevent this pathway leading to adaptation to acute hypoxia [113].

While the mechanism of mitochondrial targeting of SLC8B1 has not yet been studied, the sequence of human SLC8B1 does not seem to contain a mitochondrial targeting sequence according to TargetP-2.0 predictions (likelihood 0.0016). This suggests that other internal targeting signals are present that direct the protein into the mitochondria, possibly similarly to SLC25, MPC and sideroflexin proteins.

Disease-releated mutations have not been reported for SLC8B1 thus far. However, residues responsible for the unique Li⁺-exchange capacity of SLC8B1 have been investigated in detail. Based on the determined structure of an archaeal NCX homologue, NCX_Mj, it was found that only 3 of the 12 ion-coordinating residues were shared with human SLC8B1 [114]. By mutating the 9 different residues to their human counterparts, it was possible to engineer a mutant of NCX_Mj that can also mediate Li⁺-dependent Ca²⁺ exchange [114]. In a later study, it was found that mutation of residue D471 to alanine can shift the selectivity toward Na⁺, while mutations at several positions can render SLC8B1 a Li⁺-selective exchanger [115]. All these positions cluster close to the Na⁺-binding sites shown by the X-ray structures of NCX_Mj [116,117] and the homologous H⁺/Ca²⁺ exchanger CAX_Af from the euryarchaeota A. fulgidus [118].

8. Additional Families with Members Proposed to be Localized in the IMM