The peptide sensor motif stabilizes the outward-facing conformation of TmrAB

The ATP binding cassette (ABC) family of transporters move diverse small molecules across membranes in nearly all organisms. Transport activity requires conformational switching between inward-facing and outward-facing states driven by ATP-dependent dimerization of two nucleotide binding domains (NBDs). The allosteric mechanism that connects ATP binding and hydrolysis in the NBDs to conformational changes in a substrate binding site in the transmembrane domains (TMDs) presents an unresolved question. Here we use sequence coevolution analyses together with biochemical characterization to investigate the role of a highly conserved motif called the peptide sensor in coordinating domain rearrangements in the heterodimeric peptide exporter from Thermus thermophilus, TmrAB. Mutations in the peptide sensor motif alter ATP hydrolysis rates as well as substrate release. Disulfide crosslinking, evolutionary trace, and evolutionary coupling analysis reveal that these effects likely destabilize a network between the peptide sensor motif and the Q-loop and X-loop, two known allosteric elements in the NBDs. We further find that disruption of this network in TmrA versus TmrB has different functional consequences, hinting at an intrinsic asymmetry in heterodimeric ABC transporters extending beyond that of the NBDs. These results support a mechanism in which the peptide sensor motifs help coordinate the transition of TmrAB to an outward open conformation, and each half of the transporter likely plays a different role in the conformational cycle of TmrAB.

They consume ATP to transport biosynthetic components such as lipids or toxic molecules including drugs, toxins, and peptides across membranes [1,2]. Their dysfunction is genetically linked to many human diseases including cystic fibrosis, Stargardt disease, Tangier's disease, and multiple dyslipidemias [3]. Because of their medical relevance, the transport mechanism has long been a focal point of ABC transporter research, especially of exporters which predominate in humans. Structural snapshots of isolated states reveal that transport in both ABC exporters and importers is accompanied by large rearrangements between a pair of nucleotide binding domains (NBDs) that bind and hydrolyze ATP and a pair of transmembrane domains (TMDs) that bind and transport substrate.
The domain rearrangements that underlie the transport cycle in ABC transporters progresses through three general steps: 1) ATP-driven NBD dimerization, 2) conformational change from a high affinity inward-facing (IF) state to a low affinity outward-facing (OF) state, and 3) ATP hydrolysis to reset the transport cycle ( Figure 1A). Alternative models that build on this general mechanism have been proposed by biochemical and genetic studies [4][5][6], high resolution structures determined by x-ray crystallography and cryogenic electron microscopy (cryo-EM) in different states and conformations [7][8][9][10][11][12], as well as distance measurements by electron paramagnetic resonance (EPR) [13,14]. The most widely recognized models proposed can be classified into two primary groups based on mechanistic commonalities [15]. Models in the first category suggest a complete separation of NBDs occurs after each cycle and includes the Switch model [16], the Tweezers-like model [17], and processive clamp model [18]. In contrast, the second category proposes a continuous contact of at least one NBD during the transport cycle and is comprised of the nucleotide occlusion model [19,20], the alternating catalysis model [21] and the constant contact model [22]. These models serve to define the mechanism behind substrate release and the precise role of ATP binding versus hydrolysis in transporter function. Together, their findings suggest transport cycles can proceed via symmetric or asymmetric NBD dimerization -the latter being associated with asymmetric transporter architectures. Nevertheless, all of them strongly support the role of ATP binding to bridge two NBDs and initiate a conformational change in the TMD during transport.
A key element for conformational coupling is allosteric communication between ATP binding sites and the substrate binding sites. As an example, substrate binding often stimulates or, in some cases, inhibits ATPase activity [23]. Investigations into their allosteric relationship have focused on two motifs in each NBD, the Q-loop and X-loop [24][25][26] (Supplemental Figure   1). Both motifs are found on loops that bridge the NBDs to the TMDs and are highly conserved, with the X-loop found only in ABC exporters [27]. Mutations in the Q-loop of the drug transporter human P-glycoprotein (PgP) result in defects in the ability to transition to outward open and a decoupling of ATPase activity and substrate transport [28]. Similarly, X-loop mutations in ABCB4 found in progressive familial intrahepatic cholestasis type-3 (PFIC-3), a liver disease involving aberrant bile formation, result in uncoupling of substrate binding and ATPase activity [29]. Both the X-and Q-loop interact with the TMD through helical extensions of the transmembrane forming intracellular loops (ICL) bridging these NBD motifs to the substrate binding site in the TMD ( Figure 1A) [24].
Comparatively less emphasis has been placed on the ICLs compared to the X-and Qloops despite being poised to provide a mechanical link between the ATP and substrate binding sites. Sequence similarity is generally lower in the TMDs than in the highly conserved NBDs, likely owing to the variability of transported substrates. This variability makes identifying conserved allosteric pathways difficult, but several recent investigations support a sequence specific role of intracellular loop 1 (ICL-1) in the transport cycle. Mutations in a conserved patch of charged residues in ICL-1 of the bacterial lipid A transporter, MsbA, destabilize the outwardfacing state during its transport cycle [30]. Furthermore, cysteine crosslinking in the human T-cell Antigen Presenting complex, TAP1/TAP2, peptide transporter showed that residues defined as the "peptide sensor" motif (abbreviated as the PSM here) decreased transport and altered interactions throughout the transport cycle [2]. The peptide sensor sequence (Gly(Arg/Asp)x3Argx3Asp) is conserved in the evolutionarily related homolog of TAP1/TAP2, TmrAB from Thermus thermophilus ( Figure 1B). TmrAB is a peptide transporter with similar substrate specificity to TAP1/TAP2, and notably, can functionally replace TAP1/TAP2 in cellbased experiments [31]. The conservation in sequence and function by a bacterial ancestor of this class of proteins suggests an important role of this understudied motif in the regulation of ABC transporters, and further investigations may provide insights into their conformational cycle.
TmrAB is one of the best characterized ABC transporters to date [6,10,[31][32][33]. A recent study reported cryo-EM structures of 6 different states of TmrAB [10]. This complete view of its conformational landscape allowed us to identify the PSM as a potential regulatory motif in a heterodimeric asymmetric ABC transporter. In this class, one NBD lacks the ability to hydrolyze ATP due to a non-canonical walker B motif (ffffDD instead of the canonical ffffDE -where ffff is any hydrophobic residue), a feature shared with many medically relevant human transporters including TAP1/TAP2, the Cystic Fibrosis Transmembrane Regulator (CFTR), and Multidrug Resistance Protein (MRP) proteins. In TmrAB, the canonical walker B motif is present in the TmrA NBD and the non-canonical walker B motif is found in the TmrB NBD. We observed that interactions formed by the peptide sensors in the structures of both TmrA and TmrB are substantially altered as the protein proceeds through the proposed transport cycle ( Figure 1C).
Upon transition to the outward open state, a new helical bundle is formed between the PSMs of both chains and cytoplasmic extensions of the TMDs. The symmetry of the tightly assembled ICLs is noteworthy especially in light of the functional asymmetry between TmrA and TmrB and raises the question whether there is a role for peptide sensor residues in facilitating ATPdependent conformational coupling in TmrAB.
Here, we further investigate the hypothesis that the peptide sensor region acts as a physical couple in TmrAB conformational changes upon ATP binding. Our analysis builds on previous characterization of the peptide sensor region from TAP1/TAP2 [2,24] and further defines the Gly(Arg/Asp)x3Argx3Glu as a peptide sensor motif (PSM). We show consensus in this sequence across ABC exporters using large scale evolutionary analysis with TmrAB and representative members of every class of human exporter that display a similar fold (B, C, and D) and of their bacterial homologs. In the PSM, the charged residues are highly conserved, and the glycine residue is almost absolutely conserved. Biochemically, we observe asymmetric disruption of the ATPase catalytic cycle when either PSM is disrupted, suggesting different roles of the PSM in the catalytically active and inactive chain of the transporter. Furthermore, we find that the highly conserved charged residues specifically are critical for stabilizing the outward open conformation necessary for export. Together, our results shed insights on a potential mechanism for stabilizing the formation of structural intermediates in the conformational transitions underlying ABC transporter activity and helping propagate allosteric information between ATP binding and substrate binding sites.

The peptide sensor motif forms part of an evolutionarily constrained network connecting nucleotide binding to the transmembrane region
To assess the extent to which the peptide sensor motif (PSM) within intracellular loop 1 (ICL-1) may be important for allosteric nucleotide coupling in asymmetric ABC transporters, we first analyzed its conservation across evolution ( Figure 1B TmrA-like (R132 -~90% R or K, R135-100% R or K, D140 -100%) and TmrB-like (D117 -97% D or E, R120 -83% with R or K, D125 -100%) sequences of ABC exporters in bacteria.
Nonetheless, in all sequences a glycine residue precedes three charged residues that follow an i + 4 helical pattern (i.e. are spaced by 3 residues). This putative consensus motif connects the coupling helix-1 (CH1) positioned at the surface of the NBD to transmembrane helix 3 (TM3) in the TMD in TmrAB [2,24] (Figure 1C).

Evolutionary contact predictions show the PSM connects NBDs to the transmembrane in the outward-occluded and outward-facing state of TmrAB
The scale of our sequence alignments allowed us to take a closer look at the functional relevance of the PSM in ABC exporters through sequence coevolution analysis. Many predictive techniques for classifying the structural and functional properties of specific sites through multisequence analysis have emerged in recent years [34]. The Evolutionary Trace method is one such approach that is particularly powerful because it can infer likely functionally important features from phylogenetic relations and information content through evolution of a family [35].
Using the Universal Evolutionary Trace method [36], residues of TmrA and TmrB were assigned a coverage score correlating with functional importance and in which residues with a lower score (higher inverse score) are considered more functionally important. As expected, residues in multiple known motifs in the NBDs ranked in the top 5% (above 95% rank) (Figure 2A Further analysis of all sites above 95% rank in the context of the TmrAB conformational states reveals a distinct continuous network that extends from the nucleotide binding sites at the NBD dimer interface, upwards along the peptide sensor in the TmrA ICL-1, and into the TMD in the outward-open TmrAB structure (PDB ID: 6raj [10]; Figure 2B). This network is anchored by G131 TmrA at the NBD/ICL-2 and ICL-1 interface and by D140 TmrA at the ICL-1/TMD interface.
Interestingly, this putative route is not conserved on the TmrB face of the heterodimer when considering only the top 5% of sites, and lacks the cluster that forms the NBD connection to the TmrB ICL-1 as observed in TmrA, highlighting one possible route for conformational coupling between the ATP binding and substrate binding sites.
To better understand the structural basis for how the peptide sensor motif (PSM) may regulate the NBD and the TMD, we investigated evolutionary sequence covariations in TmrAB using a second evolutionary mapping approach and the EVCouplings software suite (Figure 3) [37]. Evolutionary coupling analysis provides direct insights into 3-dimensional contacts through the use of large scale sequence alignments (typically several hundred to several thousand sequences) and accounts for global coupling of sequences that are better able to filter indirect contacts that arise in methods such as Evolutionary Trace [35]. The success of these methods has been best displayed in the success of using these contacts for de novo structure prediction [38]. The EV couplings analysis of TmrA ( Figure 3A) and TmrB ( Figure 3B (Figure 3).We compared the number of connections in ICL-1 and ICL-2 and detected higher conserved co-evolving residues pairs for ICL-1 over ICL-2 (4.6x).
Important connections between the NBD and ICL-1, but not with ICL-2, were identified. This suggests a potential role of ICL-1-mediated coupling of the transmembrane domain and NBDs.

TmrAB PSM mutants exhibit WT like thermostability
To test the role of specific PSM residues in TmrAB activation, we generated TmrAB  (Figure 4) with similar denaturation profiles. Furthermore, samples were inspected for aggregation using negative stain electron microscopy, which did not reveal perturbations to the overall architecture (Supplemental Figure 2). These results suggest that the PSM mutants do not interfere with stability of the resting folding state of TmrAB.

PSM mutants in each chain have different effects on the TmrAB heterodimer ATPase activity
TmrAB is derived from a thermophilic bacteria that exhibits optimal growth and function at 68°C [33]. To determine the effect of PSM mutations on catalysis in TmrAB at its physiological temperature, we employed a colorimetric endpoint assay using ascorbic acid to measure the rate of ATP hydrolysis ( Figure 5A; Table 1). Because TmrAB contains a single functional NBD only in TmrA and the second site likely has ATP bound through the catalytic cycle, all data were fit with a single site Michaelis Menten model. WT TmrAB displayed characteristic ATPase activity with a measured Km value of 1.35 ± 0.24 mM and a kcat of 3.03 ± 0.19 s -1 , consistent with previously published reports [10,33].
Changes in mutant activity can be divided into two groups (in each case compared to WT) as judged by kcat/Km, which was chosen because it provides an indirect measure of the ability of each productive ATP-bound complex to couple an ATP hydrolysis event to a conformational change, and thus informs on the effect of PSM mutations in conformational coupling ( Figure 5A, Table 1).. In group 1, catalytic efficiency was decreased to 26 -37% of WT TmrAB in the G131A TmrA , R132A TmrA , R121A TmrB , and D125A TmrB mutants. Interestingly, this effect was nearly inverted in group 2 which contained residues from the opposite chain as group 1. In group 2, activity increased ~1.7-fold in G116A TmrB, D117A TmrB , and in R136A TmrA . Activity was only marginally increased (107% of WT) in D140A TmrA .

PSM mutants do not influence ATP binding
To determine whether the effects on ATP hydrolysis for our PSM mutants arise from deficiencies in nucleotide binding affinity, we determined the apparent affinity, KD-app, for the fluorescent ATP analog, TNP-ATP, in WT TmrAB and each PSM mutant ( Figure 5B, Table 2).
The KD-app TNP-ATP ranged from 0.81 µM in WT and 1.57 -6.17 µM in PSM mutants. These values for the dissociation constant are significantly lower than their enzymatic Km values, and likely due to formation of partially occluded states as is the case in PgP [20], but overall still consistent with previous reports of the KD-app TNP-ATP in MsbA [39]. It is important to note that these KD values are apparent KD values since there are two nucleotide binding sites and it has been proposed that one site can be more occluded [6] and that this site has higher affinity [20] . There was no statistically significant difference in affinities for ATP between WT TmrAB and the PSM mutants, suggesting that the PSM residues are likely not a determinant for ATP binding.

ATPase-deficient PSM mutations block entry into low peptide affinity conformations of TmrAB
According to the processive-clamp/switch model [16,18], transport is only achieved when the high affinity inward-facing state changes to a low affinity outward-facing state to release substrate. To test the ability of TmrAB PSM mutants to induce the IF-to-OF switch and thus enable substrate release, we employed a fluorescent polarization binding assay to quantify binding of a fluorescently-labeled peptide substrate (RRY(C Fluorescein )KSTEL) to TmrAB in the absence and presence of ATP (Figure 6). In this assay, a decrease in the polarization signal in response to ATP binding corresponds to a loss of bound peptide substrate, presumably through the ability of TmrAB to sample the low-affinity outward open conformation [6]. Our experiments were performed at 4°C where ATP hydrolysis is not observed in TmrAB to prevent the effects of trans-inhibition by hydrolysis products, and for concentrations of peptide where at least 30% of the total enzyme in the sample is bound to minimize protein consumption, under similar concentrations and conditions as previously described [10,31,40].
The peptide binding data showed a decrease in peptide binding in the presence of ATP for WT TmrAB, as is expected. Group 1 mutants R132A TmrA and R121A TmrB and all group 2 mutants (R136A TmrA , D140A TmrA , G116A TmrB and D117A TmrB ) similarly showed this change, suggesting that they also can attain a low affinity state. These results suggest conformational coupling between the NBD and TMD proceeds as expected in these mutants in which ATP hydrolysis was increased. The change in peptide binding in the G131A TmrA and D125A TmrB variants was not significant, consistent with their decreased efficiency of ATPase activity. While peptide binding was also slightly decreased in an ATP-dependent manner in the ATPase deficient R132A TmrA and R121A TmrB variants, the observed change was less significant than in the catalytically-competent PSM mutants. Together, this data suggests that ATPase deficient PSM mutants (group 1) trapped in the high-affinity conformational state of TmrAB may not induce the IF-to-OF switch in the presence of ATP binding.

Mutation of the 100% conserved PSM glycines block the ability of TmrAB to achieve an outward-facing conformation
To understand how the positioning of the PSM might influence the transport cycle, we focused on mutants of the two most conserved residues with opposite functional effects on catalytic activity and peptide binding, G131A TmrA and G116A TmrB . In order to measure formation of exclusively the OFopen state of TmrAB, we engineered cysteines into the glycine variants for disulfide cross-linking on the outside gate of TmrAB. These double cysteine mutations correspond to positions previously investigated in the homodimeric transporter MsbA [39] (L290C in TmrA and V275C in TmrB) (Figure 7A), and are positioned such that a cross-link only forms in IFopen or OFocc states and cannot form in the OFopen state necessary for transport, thus informing on the formation of this state.
In the absence of ATP, the majority of WT TmrAB forms the IFopen state as assessed by the predominant presence of a ~150kDa band on an SDS-PAGE gel corresponding to a dimer between TmrA and TmrB (Figure 7B,C). This result is consistent with similar previously performed studies on ABC exporters [39]. This higher molecular weight crosslinked band disappeared upon incubation with either ATP alone or ATP plus sodium orthovanadate (ATP-Vi), which traps an OF state as expected. However, when repeated with the PSM glycine mutants, our experiment showed an almost complete recovery of crosslinks in the ATP-Vi condition. This implies that the ATP-Vi bound state is not able to hold TmrAB in the OFopen conformation. Notably, a less pronounced effect was observed in the absence of sodium orthovanadate (ATP alone), with only a partial recovery of cross-linking. It is unclear why ATP-Vi or ATP alone have different effects in the mutants, though the cryo-EM structures of TmrAB find that datasets in the presence of vanadate have a particle distribution of ~65% and ~35% in the OFocc and OFopen respectively, whereas ATP alone primarily adopts the OFopen conformation [10]. Nevertheless, our cross-linking results suggest that both the G131A TmrA variant and the G116A TmrB have an impaired ability to couple the presence of ATP or ATP-Vi to the formation of OFopen. transport. Together, our data identify a mechanism for how an allosteric signal between distal substrate bindings sites may be communicated during transport in asymmetric ABC exporters.

DISCUSSION
Protein allostery is diversely defined and can be executed through global conformational changes and domain movements [41], local rearrangements of side chain networks [42], or by subtle alterations in protein dynamics independently of a conformational change [43]. Analysis of evolutionarily conserved networks is a powerful approach for inferring conserved mechanisms of allostery from sequence. Here, two complimentary evolutionary analysis methods were used to identify such networks in the understudied, but critical, TmrAB exporter ICL region: Evolutionary Trace [35] to identify conserved single residues important for function and Evolutionary Couplings [36,37] to identify conserved contacts between pairs of residues.
Evolutionary couplings are particularly informative since they can infer multiple types of interactions -those important for function, those necessary for stabilizing the folded state of a protein, and those that stabilize functionally important alternate conformations. Both methods clearly highlighted the ICL-1 as a central hub for functionally important interactions with several conserved sites clustered in the NBDs and TMDs. Unexpectedly, the most significant of these contacts form networks within one chain of TmrAB that largely do not cross-over into the other chain. However, one prominent network emerged from our analysis in which conserved residues of the canonical TmrA chain PSM act as a conduit bridging sites on the non-canonical TmrB NBD to the TmrA TMD. Notably this includes residues from the TmrB X-loop, which is overall more conserved than the X-loop in TmrA. Furthermore, nearly every pair predicted by evolutionary couplings could be observed forming a direct interaction when mapped onto  Table   3). In group 1 (G131A TmrA , R132A TmrA , R121A TmrB and D125A TmrB ), the decreased ATPase activity relative to WT TmrAB and impaired formation of a low affinity substrate binding site in the presence of ATP suggests hydrolysis is rate-limited by the stability of the OFopen conformation. In contrast, the second group of mutants (G116A TmrB , D117A TmrB , R136A TmrA and D140A TmrA ) were marked by elevated ATPase activity compared to WT TmrAB suggesting a futile ATPase cycle as is observed for mutations in the Q-loop and X-loop of other ABC transporters [28,29]. Such futile cycles often uncouple ATPase activity from the transport cycle, which we also observe in TmrAB. Interestingly, very few of the evolutionarily important contacts, especially in the TMD, identified in our bioinformatics analyses are formed between TmrA and TmrB and instead are intra-chain. Inter-chain interactions were limited to those formed between the NBDs and ICLs (Figure 2), and the observed enzymatic asymmetry of the PSM mutants likely reflects the functional asymmetry of the NBD to which they are networked.
An unusual aspect of our data is the similarities in the functional properties of the PSM glycine mutants to those of the charged residues of the PSM in the opposite chain. The IF-to-OF switch may be aided by a cluster of glutamates that interact with the positive helix dipole of the conserved motif (E218 TmrA and E470 TmrB with the TmrA PSM; E203 TmrB and E492 TmrA with the TmrB PSM). Notably, E218 TmrA and E470 TmrB were ranked in the top 5% of functionally important residues predicted by ET analysis and are part of the continuous network connecting the Q-loop and X-loop of the TmrB NBD to the TmrA PSM and TMD substrate binding sites ( Figure 2B).
The stability imparted by these interactions may be the key element helping to propagate an allosteric signal from the NBD to the TMD.
The IF-to-OF transition does not result in a large change in the PSM backbone atom dihedral angles suggesting the flexibility of the nearly absolutely conserved glycines (G131 TmrA and G116A TmrB ) is not important for PSM function. Instead, it is likely that the small size of glycine is necessary for both TM3 E helices harboring the PSMs to move past other parts of the intracellular loop region that rearrange and become tightly packed in the OF states of ABC exporters like TmrAB. This constricted assembly may be necessary to stabilize the non-typical p-helix geometry formed by residues of the PSM [44]. Frequently observed in critical conformational junctions of dynamic integral membrane proteins, p-helices are often formed and broken between structural transitions and are observed in the TM3 E helices of several transporters such as the bacterial amino acid transporter LeuT [44,45], in the OF states of ABC transporters [8,10,46], and in the human temperature sensitive ion channels of the TRPV family [47][48][49].
It is unclear if our results are specific to heterodimers or more broadly reflect homodimeric mechanisms in exporters that share the TmrAB fold (families B,C, and D in humans), though there is evidence that asymmetry may also be relevant for exporters with two canonical NBDs. Most significantly, asymmetry in ATP binding has been observed in the human drug transporter PgP, which has two functional NBDs [20]. Investigation of DARPin binding to the bacterial homodimer MsbA also showed asymmetric binding of a single DARPin at a time depending on the specific state of ATP along the hydrolysis cycle in each NBD [50]. Lastly, EPR distance measurements on PgP along its transport cycle reveal inherent conformational asymmetry similar to that observed for heterodimeric transporters, thus reflecting processive hydrolysis of ATP at two NBDs [14]. A later EPR study on PgP showed that asymmetric conformations are associated with the substrate bound states, but not for inhibitor bound states [51]. These proteins contain the same PSM-mediated conserved networks identified in our evolutionary analyses of TmrA and TmrB. This evolutionary significance supports a role of the PSM as a common feature of exporter type ABC transporters with the same fold as TmrAB independent of NBD functional symmetry.
In conclusion, we have identified the peptide sensor motif in TmrAB as an important element for the allosteric mechanism that couples ATP binding to conformational changes during transport. Mutation of the highly conserved motif not only results in alterations in catalysis, but also in an overall loss in the ability to attain an OFopen state. Findings from coevolution analyses support a mechanism in which these effects likely arise from the destabilization of a conserved electrostatic network mediated by the PSM in TmrA. Specifically, we highlight a potential role of the almost absolutely conserved G131A TmrA in aligning the NBDs and TMDs during the transition to the OF state. Together our results suggest the existence of differences in how conformational changes are propagated along each side of ABC exporters.

Sequence alignments
Sequence alignments were performed against a collection of bacterial and mammalian ABC transporters that share the TmrAB fold. To limit hits to only ABC transporters with the TmrAB transmembrane fold, only the transmembrane sequence of TmrA (residues 1 -332) and TmrB (residues 1-317) were used for the search. A blast search of the Uniref90 database for bacterial transporters was conducted with an E value cutoff of 0.01 to enable a robust search that covered wide evolutionary sequence space and multiple bacterial phyla. The sequences were then filtered to only include those containing a Walker B motif and that were between 540 and 900 amino acids (90% and 150% length of TmrA, respectively). To remove redundancy and downweigh closely related transporters, the sequences were next filtered through CD-HIT [52] with an 80% identity cutoff. Resultant sequences were aligned with default parameters in MAFFT [53,54] and manually inspected for gaps or undefined residues (1 sequence removed from TmrB alignments) resulting in 615 TmrA-like sequences and 567 TmrB-like sequences (Supplemental Data 1). Conservation of the PSM was analyzed using the WebLogo server [55]. The alignment presented in Supplemental Figure 1 of a manual selection of representative ABC transporters from bacteria, yeast, and humans was also performed in MAFFT [53,54] [and colored by similarity in ESpript 3.0 [56]. A phylogenetic tree was generated on this alignment using raXML [57] and TidyTree (CDC).

Evolutionary Trace analysis
Evolutionary trace analysis was performed using the custom MAFFT alignments described above. To facilitate analysis, the TmrA and TmrB sequences were respectively added back to the list of sequences that had been filtered for redundancy in CD-HIT [58]. The sequences were realigned with MAFFT [53,54] and uploaded to the UET server [36]. The rvET "coverage score" which ranks a given residue against all other residues in the protein was subtracted from 1 to generate a rank in which the most important residues were scored the highest. Residues ranked in the top 5% were mapped onto multiple structures of TmrAB to evaluate the relationship between functional importance and spatial proximity (Supplemental Table 1).

Evolutionary couplings analysis
Evolutionary couplings in TmrAB were analyzed in three sets of alignments -TmrA (intra-TmrA -TmrA contacts), in TmrB (intra-TmrB-TmrB), and in TmrAB (inter-TmrA to TmrB) using the Evcouplings framework (Supplemental Data 1). To ensure sufficient coverage and robust coverage over as much of each TmrAB half as possible, JackHMMMR [59] was used to make a large sequence alignment of TmrA alone and TmrB alone with a Hidden Markov Model to fit difficult to align loops. HHsuite [60] was used to filter redundancy and cluster residues above an 80% sequence identity to ensure broad phylogenetic representation and to limit the overall number of sequences. Sequences were filtered using a Biopython [61] script for either a canonical or noncanonical Walker B motif to ensure proper comparison of either side of the heterodimer. With these criteria, 25,781 sequences were found related to TmrA and 5,000 sequences related to TmrB. Evcouplings analysis of the aligned sequences was then performed using PLMC and analyzed using tools from Evcouplings [37]. Comparison to known structures simultaneously carried out with Evcouplings.

Molecular biology and cloning
Mutants of TmrAB were generated from a TmrAB template in the pet22b vector previously described in Zutz et. al. [33]. Mutagenesis primers were designed using the technique described by Liu and Naismith [62]. Constructs were verified by sequencing (Elim Biopharmaceuticals, Inc.). Construction of cysteine-less TmrAB templates (C416S) of WT TmrAB, G131A TmrA and G116A TmrA was performed by GenScript and used as the template for introducing two N-terminal cysteine residues at sites L290 and V275 in chains A and B, respectively.

Protein expression and purification
WT and mutant TmrAB were expressed and purified using the method previously described by Zutz et. al [33]. were pooled, concentrated, and stored at -80°C prior to further experiments.

Intrinsic tryptophan and fluorescence analysis
To observe the effect of PSM mutations on TmrAB structural integrity and stability, changes in the emission spectra of tryptophan residues, induced by temperature-dependent unfolding, was monitored from 35°C to 95°C. TmrAB was diluted to 8.5 µM in SEC buffer and prepared in duplicate for measurement on a Tycho NT.6 (NanoTemper Technologies). Data are reported as the ratio of fluorescence intensity at 350 nm and 330 nm normalized to the minimum and maximum values of the whole dataset.

Negative-stain analysis
Samples of WT and mutant TmrAB were prepared for negative-stain imaging to evaluate the quality and architecture of purified protein. Protein was diluted to ~0.2 µM concentration in SEC buffer and 3 µL applied to negative glow discharged carbon coated copper grids (EMS, Cat No. CF300-Cu-UL). Sample was incubated on grids for 30 s, then gently dipped in two drops of SEC buffer and one drop of 0.75% uranyl formate staining solution with blotting performed between drops by touching the grid edges to the surface of Whatman 1 filter paper. Grids were dipped in a second drop of stain for 30 s, blotted, and air dried prior to imaging. Images were collected on a 120 kV Tecnai Spirit (FEI) microscope equipped with an AMT XR80L-B 8 x 8 CCD camera at 49,000X magnification corresponding to a pixel size of 2.21 Å.

Colorimetric ATPase assay
A colorimetric endpoint assay for inorganic phosphate determination described by Sarkadi et al. [63] was used to measure TmrAB ATPase activity near the physiological growth

(Equation 1)
Where Y is the enzyme velocity (mM/s), ET is the total enzyme concentration (mM), kcat is the turnover number (1/s), X is the concentration of ATP (mM), and Km is the Michaelis-Menten constant (mM). Kinetic parameters (Km, kcat, Vmax, and kcat/Km) are reported in Table 1 as the mean ± standard error of the mean (S.E.M.).

ATP binding assay
The binding affinity for nucleotide was measured as described in Doshi et. al. 2013 [39] and where fb is the fraction bound of protein with ligand, 8 9:; is maximum binding, L is ligand, and < = is the dissociation constant. Apparent dissociation constants for TNP-ATP binding (KD-app TNP-ATP ) are reported in Table 2 as the mean ± S.E.M.

Peptide binding assay
Binding of the fluorescent transport substrate peptide RRY(C Fluorescein )KSTEL (Genscript) to WT and mutant TmrAB was determined as a function of the change in fluorescence polarization as described in [10]. In a total volume of 30 µL, 1.  Km and kcat values were obtained from data presented in Figure 5A fitted by nonlinear regression analysis using Equation 1 in Prism (GraphPad). c Vmax was calculated as ? @ * A B:C , where ET is the total enzyme concentration in the assay (0.135 µM). Data reported are the mean ± standard error of the mean (S.E.M.) for data collected in triplicate.  Figure 5B fitted by nonlinear regression analysis using Equation 2 in Prism (GraphPad). Data reported are the mean ± S.E.M. e P-value denotes probability of the null-hypothesis that the mutant mean is equal to that of WT TmrAB.              Figure 1 Supplemental Figure 1. Sequence alignment of conserved elements in representative members of ABC transporter families. Sequences were aligned in MAFFT [52,53] and colored in ESPript 3.0 [64] by residue similarity. For the transporters that contain two nucleotide binding domains on a single polypeptide chain (PgP, MRP1, Ste6, and Ycf1) the sequence containing the second nucleotide binding domain (half 2) was included in the alignment as a separate sequence and aligned against the first half of the sequence containing the first nucleotide binding domain (half 1). Residues of the peptide sensor mutated in this study are denoted by an asterisk. The catalytic glutamate (E) in the Walker B motif is highlighted with a black circle below the sequence. Sequences containing an aspartate (D) in this position are considered non-functional for ATP hydrolysis.