Tetanus Toxin cis-Loop Contributes to Light-Chain Translocation.

How protein toxins translocate their catalytic domain across a cell membrane is the least understood step in toxin action. This study utilized a reporter, β-lactamase, that was genetically fused to full-length, nontoxic tetanus toxin (βlac-TT) in discovery-based live-cell assays to study LC translocation. Directed mutagenesis identified a role for K768 in LC translocation. K768 was located between α15 and α16 (termed the cis-loop). Cellular assays showed that K768 did not interfere with other toxin functions, including cell binding, intracellular trafficking, and pore formation. The equivalent K768 is conserved among the clostridial neurotoxin family of proteins as a conserved structural motif. The cis-loop appears to contribute to LC translocation.

HCN. The ␣14/15-helices and ␣16/17-helices are each ϳ11 nm in length and are connected to four shorter ␣-helices by loops (8). The termini of the long ␣-helices of HCN can also be described relative to their physical distance from the LC-HCN interchain disulfide. While the trans-loop region (␣18-loop-␣18) comprising a CNTconserved phenylalanine (TT 838 F equivalent) is located away from the LC-HCN interchain disulfide, the cis-loop region (␣15-loop-␣16) comprising a CNT-conserved lysine (TT 768 K equivalent) located near the LC-HCN interchain disulfide (1). A long loop termed the membrane penetrating peptide which, in isolation, forms ion conducting channels in lipid membranes aligns with the long kinked ␣14/15 and ␣16/17 helices (15,16). Note that the membrane penetrating peptide is conserved among the CNTs and corresponds to amino acids 641 to 691 of TT (16). Recently, a pleomorphic region in HCN (termed the BoNT switch) may function as a pH-triggered lipid anchor (17). However, how HCN facilitates pore formation and LC delivery remains unknown.
Current models propose that following CNT binding and vesicle entry, acidification triggers HCN insertion into the vesicle membrane to deliver LC into the host cytosol (11,(17)(18)(19). LC translocation is necessary for intoxication of host neuron, as premature reduction of the interchain disulfide or inhibition of endosomal acidification reduces neurotoxicity (20). Previous studies on LC translocation identified regions required for pore formation, identifying a minimum pore-forming domain (19), and intoxication as measured by cation conductance through lipid bilayers and substrate cleavage. Heat shock protein 90 (Hsp90) has been shown to contribute in LC translocation (21). Unresolved issues for HCN-mediated LC translocation remain, including what domains/ regions are responsible for pore formation and pH-dependent conformational changes, whether pore formation and LC delivery are coupled, and how LC localizes to the pore. The mechanistic role of the interchain disulfide in neurotoxicity is unclear; however, the thioredoxin-thioredoxin reductase (Trx-TrxR) system is involved in the reduction of the interchain disulfide of CNT (22). Conservation of the overall structures of the HCN domain within the crystal structures of BT serotypes and TT (Fig. 1) implicate a common LC translocation mechanism among the CNTs, with the potential for unique properties within individual CNTs based on each CNT's unique tertiary organization, especially with respect to changes in pH (10).
Here, we used ␤lac-TT, a discovery-based, live-cell imaging reporter system in cells to map regions of HCN that contribute to LC translocation and identified a helix-loophelix within HCN (cis-loop) required for LC translocation, independently of pore formation.

RESULTS
␤lac-TT variants in the loop between ␣15 and ␣16 (cis-loop) within HCN are defective in LC translocation. ␤lac-TT was developed as a reporter for a cell-based model of LC translocation by tetanus toxin detected as the cytosolic cleavage of a FRET-based substrate, CCF2 (12). Overall, this assay has proven to be a useful discoverybased system (13). In initial experiments, two conserved regions of the HCN, a hydrophobic region within ␣12 and a charged loop between ␣15 and ␣16 (cis-loop) (Fig. 2) were targeted by site-directed mutagenesis to assess potential roles in LC translocation.
Cell imaging showed that mutations within ␣12 of ␤lac-TT that targeted a localized group of aspartic acids or three conserved aromatic amino acids translocated ␤lac [␤lac-TT(F 612 A, W 615 A, F 623 A) and ␤lac-TT(D 618 K, D 621 K, D 622 K)] had a number of ␤lac translocation events similar to the number seen with ␤lac-TT (Fig. 3A). In contrast, ␤lac-TT( 767 AAA 769 ), a cis-loop mutation, did not cleave cytosolic CCF2 (Fig. 3A). ␤lac-TT( 767 RKK 769 ) mediated numbers of ␤lac translocation events similar to the number seen with ␤lac-TT, showing that the acidic residues were not required for ␤lac translocation. ␤lac-TT( 767 DAE 769 ) and ␤lac-TT( 767 AKA 769 ) did not cleave cytosolic CCF2, indicating that the presence of 768 K was necessary but not sufficient for reporter translocation (Fig. 3B). Alignment of TT with the seven BT serotypes A to G, BT serotype X, and two recently identified CNT-like proteins, En and Pmp1 (1,(23)(24)(25), showed that the cis-loop lysine was conserved whereas the acidic amino acids were present but uniquely aligned among the BTs, except BT/C, BT/D, and BT/En, which maintained the DKE sequence (Table 1).
␤lac-TT( 767 AAA 769 ) retains protein structure and cell binding function. To rule out the possibility that the LC translocation defect was a result of reporter instability or of a lack of disulfide formation, ␤lac-TT(cis-loop) variants were subjected to trypsin digestion. ␤lac-TT and ␤lac-TT( 767 AAA 769 ) were assembled with the interchain disulfide and yielded similar tryptic cleavage patterns (see Fig. S1 in the supplemental material). In addition, ␤lac-TT( 767 AAA 769 ) (Fig. 4) and various point mutation ␤lac-TT( 767 DKE 769 ) variants bound neurons as efficiently as ␤lac-TT (Fig. S2), indicating that cell binding by ␤lac-TT was not disrupted by introduction of cis-loop mutations.
␤lac-TT and cis-loop variants form similar pores in Neuro-2a cells. Trypan blue diffusion into the cytosol of Neuro-2a cells was used to measure pore formation by cell-bound ␤lac-TT (26). Tetanolysin was used as a positive control for pore formation. At either 40 nM or 80 nM, ␤lac-TT or cis-loop variants mediated acidification-dependent trypan blue uptake, while ␤lac-HCC/T, which lacks pore-forming ability, mediated background levels of trypan blue uptake ( Fig. 6) (26). This showed that cis-loop variants were not defective in pore formation.
Mutation of cis-loop does not alter low-pH-and TrxR-dependent translocation by ␤lac-TT. LC translocation by CNTs requires low endosomal pH levels and is inhibited by ionophores or inhibitors of vATPase (27). Previously, ␤lac-TT translocation was shown to be arrested in the presence of endosomal vATPase inhibitor (12). PropKa analysis (28) of cis-loop predicted protonation of acidic residues D 767 and E 769 at low pH. Although ␤lac-TT( 767 AAA 769 ) could not be directly assessed, pretreatment of primary neurons with bafilomycin inhibited ␤lac translocation by ␤lac-TT( 767 RKK 769 ) and ␤lac-TT (Fig. S3A). More recently, thioredoxin and thioredoxin reductase inhibitors have been shown to be CNT neuroprotective in cells and animals (22). Neurons pretreated with the TrxR inhibitor, auranofin, inhibited ␤lac-LC translocation by ␤lac-TT and ␤lac-TT( 767 RKK 769 ) (Fig. S3B). Together, these experiments showed that intrinsic mutations to the cis-loop did not alter the canonical pathway of ␤lac-TT into primary neurons.
Molecular simulations detected a polarity in HCN cis-loop orientation with the cell membrane. CNT-HCNs comprise a trans-loop (␣18-loop-␣18 ( 838 F in TT), distanced  (Fig. 1), implicating a similar mechanism of pH-dependent membrane insertion among the CNTs. Molecular dynamics simulations of beltless HCN with a lipid membrane (29) showed the preferred association, i.e., 8 of 9 simulations that yielded a membrane-bound HCN, of the HCN trans-loop with the cell membrane, with the HCN cis-loop angled off the cell membrane (Fig. 7). After the initial association with the membrane, HCN remained bound and the HCN trans-loop was buried in the membrane. The surface orientation of the cis-loop in these simulations is consistent with the cis-loop functioning independently of pore function.

DISCUSSION
LC translocation is the least understood step in CNT intoxication. This study characterized the role of conserved regions of the HCN required for LC translocation in TT, assuming that the mechanism of LC translocation is conserved among CNTs. In neurons, deletion of conserved loops and short helices of HCN in C-terminal regions relative to ␣12 reduced LC translocation efficiency, while deletion of N-terminal loop regions relative to ␣12 was dispensable for LC translocation (unpublished data). In  addition, a short stretch of amino acids, 767 DKE 769 , within the ␣-helix15-loop-␣-helix16 junction (cis-loop) was required for LC translocation but not intracellular trafficking or pore formation. Consistent with a role in orientating the unfolded LC for translocation were molecular simulations that located the cis-loop on the membrane surface (Fig. 7). These studies identified the cis-loop as participating in LC translocation, independently of pore formation. Directed mutagenesis of ␤lac-TT identified a charged loop within the HCN of TT, the cis-loop comprising 767 DKE 769 within ␣-helix15-loop-␣-helix16 that mediated ␤lac-LC translocation. The aliphatic substitution of ␤lac-TT( 767 AAA 769 ) also failed to translocate ␤lac-LC. 768 K was the only conserved amino acid within the cis-loop of the respective CNTs, supporting the idea of the cis-loop being a structural motif.
The cis-loop structure is conserved in all BT serotypes, and the DKE sequence was homologous in BT/C and BT/D, but varied within the cis-loop of the other BT serotypes (Table 1). Thus, the cis-loop appears to be a structural motif rather than a sequencebased motif. Alignments with diphtheria toxin identified a candidate cis-loop, 239 SEEKA 243 , within ␣-TH5-␣-TH6, which is N terminal to the membrane-penetrating ␣-TH8 -␣-TH9 (PDB: 1SGK). Since ␣-TH5-␣-TH6 has some membrane interaction properties, the cis-loop could contribute in membrane stability (30). The TT-cis-loop was similar to the cis-loop of BT/En and homologous to the cis-loop of an Anopheles mosquito protein (25). Thus, the cis-loop is a feature shared among bacterial toxins that possess an HCN-like LC translocation function.
CNT-mediated LC translocation has been predominantly studied using electrophys- Light-Chain Translocation by Tetanus Toxin iological measurements of cation flow across lipid membranes, liposomal release assays, or analysis of substrate cleavage in neuronal cells. These studies used full-length CNT or isolated domains and aided our appreciation that low pH, proteolysis, and an intact disulfide (12,20,31,32) were involved in LC translocation. Two models of LC translocation for BT included the tunnel model and the cleft model ( Fig. 8) (6,11,33). In the cleft model, HCN deforms the membrane and the LC partially unfolds, ratcheting through the deformed lipid membrane where a hydrophobic core of HCN contacts lipids and hydrophilic residues of LC contact the HCN. Supporting this model are the findings that the protein toxin channels in lipid membranes can conduct ions but not , ␤lac-HCC/T, or tetanolysin (TTL) (80 nM each) was incubated with ganglioside-enriched Neuro-2a cells in low-K ϩ buffer for 20 min at 4°C before pulsing was performed with a prewarmed low-K ϩ reaction mixture buffered with citrate (pH 5.5) for 15 min as previously described (26,42). Buffer was aspirated and 0.2% trypan blue applied for 1.5 min before fixation. ␤lac-HCC/T was used to assess translocation-independent trypan uptake, and tetanolysin, a pore-forming toxin, was used as a positive control. Bars, 10 m. (B) Random fields were obtained, and the number of cells with trypan uptake was expressed as a percentage of the total. Data representing results from three independent replicates are shown with SEM. ***, P Ͻ 0.001; ns, not significant (Student's two-tailed t test). larger molecules (33). Both translocation models require energy to drive the processes of pore formation and LC translocation. Protein toxins, such as diphtheria toxin and ricin toxin, require proton gradients and membrane potential to drive membrane insertion (34). The energy requirement for CNT LC translocation has not been resolved but may be due to Brownian motion (35). Brownian motion is a passive diffusion of polypeptides through the translocon, which results from protonation and neutralization of acidic residues that then deprotonate upon cytosolic delivery. Deprotonation of acidic residues increases repulsion between the translocated polypeptide and the acidic phospholipids, driving LC translocation (36,37). Independently of the channel, LC translocation also involves cytosolic host factors, such as Hsps, as chaperones to facilitate delivery that is ATP dependent (21,36,38). The cleft model requires acidic residue neutralization or shielding of charged residues, such as pairing basic catalytic domain residues with acidic phospholipid headgroups or anions, to be energetically favorable (33). The cis-loop function is compatible with either the tunnel model or the cleft model for LC translocation. The cis-loop may function in LC translocation directly by orientating the LC for delivery into the pore; mammalian J proteins have this type of structure-function (39), since the cis-loop is located adjacent to the interchain disulfide (LC-HCN) that was previously shown to be needed for LC translocation (12,40). Alternatively, the cis-loop may contribute to the pH-dependent conformational changes, since each CNT cis-loop possesses, in addition to the conserved K 768 , at least one adjacent acidic amino acid that may change charge with pH (Table 1). Continued assessment of ␣-helix15-loop-␣-helix16 (cis-loop) and amino acids that define the structural components of the cis-loop may provide additional details for LC translocation.

MATERIALS AND METHODS
Tetanus toxin reporter expression. ␤lac-TT(RY) and ␤lac-HCC/T were previously engineered (12,41). DNA encoding atoxic ␤lac-TT(RY) was used as a template for site-directed mutagenesis of HCN. Site-directed mutagenesis primers designed for use an Agilent QuikChange II kit or a New England Biolabs Q5 SDM kit were used to engineer ␤lac-TT cis-loop ( 767 DKE 769 Ͼ 767 AAA 769 and 767 DKE 769 Ͼ 767 RKK 769 ) per kit instructions before transformation into Escherichia coli TG1. Primers for site-directed

Light-Chain Translocation by Tetanus Toxin mutagenesis of the other ␤lac-TT cis-loop variants [residues K 768 A (DKE Ͼ DAE) and D 767 A and E 769 A (DKE Ͼ AKA)]
and ␤lac-TT ␣12 mutations [residues D 618 K, D 621 K, and D 622 K (DIIDD Ͼ KIIKK) and F 612 A, W 615 A, and F 623 A (FWF Ͼ AAA)] were designed using Agilent or NEB Web-based tools. Briefly, templates and primers were mixed and amplified per the specifications in the Q5 kit instructions (New England Biolabs). Amplified product was incubated with KLD (kinase, ligase, and DpnI) before transformation of the entire reaction into E. coli TG1.
Cell culture protocols. Neuro-2a cells (ATCC CCL-131), from a Mus musculus neuroblastoma, were cultured as described previously (12) with the exception that coverslips were coated with poly-D-lysine (Sigma-Aldrich) followed by assay 1 day after plating at ϳ70% confluence. E18 rat cortices from Sprague Dawley rats (BrainBits, LLC) were triturated to single cells as described by the supplier and plated in NBActiv4 (BrainBits, LLC) (45,000 cells/well) on glass-bottom total internal reflection (TIRF) plates (Mat-Tek). TIRF plates were precoated with 20 g/ml poly-D-lysine (Sigma-Aldrich) overnight, followed by 3 g/ml mouse laminin for 3 h, and equilibrated with neurobasal medium for 30 min before plating cells in NBActiv4. Neurons were cultured for 7 to 12 days with a half-fresh media change, using NBActiv1 (BrainBits, LLC) on days 4 and 7 postplating.
Trypan blue uptake assay (pore formation) of ␤lac-TT variants in Neuro-2a cells. Trypan blue uptake was performed as previously described (26). Briefly, cells were plated as described above and loaded with 10 g/well of GT1b. Cells were washed with cooled low-K ϩ buffer (15 mM HEPES, 145 mM NaCl, 5.6 mM KCl, 2.2 mM CaCl 2 , 0.5 mM MgCl 2 , pH 7.4) and incubated on ice for 10 min. ␤lac-TT or scanning deletion variants or cis-loop variants (40 or 80 nM) were suspended in cooled low-K ϩ buffer and incubated with cells for 20 min on ice, buffer was replaced with prewarmed low-K ϩ buffer to reach pH 5.5 with 5 mM sodium citrate for 15 min, and the reaction mixture was incubated at 37°C (42). Cells were aspirated, 300 l of 0.2% trypan blue was added for 1.5 min, and the reaction mixture was aspirated and fixed with 4% paraformaldehyde (PFA). The clostridial pore-forming toxin tetanolysin (List Biological Laboratories, CA) was used as a positive control for pore formation. Several negative controls, including ganglioside-loaded cells incubated at pH 7 or pH 5.5 with ␤lac-HCC/T as a measure of translocationindependent trypan blue uptake, were used to assess pore formation-independent trypan blue uptake. ␤lac-TT was incubated at pH 7 to confirm pH-dependent trypan uptake.
Entry and intracellular trafficking of ␤lac-TT variants into Neuro-2a cells. Plated Neuro-2a cells were loaded with 10 g/well of sonicated ganglioside GT1b (Matreya) in minimal essential medium (MEM) containing 0.5% fetal bovine serum (FBS) for 4 h at 37°C. Cells were washed with prewarmed Dulbecco's phosphate-buffered saline (DPBS) and incubated with 40 nM ␤lac-TT or variants in low-K ϩ buffer for 20 min, at which time cells were washed and processed for immunofluorescence (described below). Cells were located as SV2C-positive stained cells, and Z-stacks were taken at 0.4-m steps followed by cropping individual cells and blind deconvolution to estimate the point spread function (PSF) with 15 iterations.
Binding of TT variants on primary rat cortical neurons. ␤lac-TT and variants were incubated (10 or 40 nM) with precooled primary rat cortical neurons in low-K ϩ buffer for 30 min on ice as described previously (12). Cells were washed with ice-cold DPBS and processed for immunofluorescence.
LC translocation assay in primary rat cortical neurons. Rat cortical neurons were incubated in equilibrated neurobasal media containing (10 or 40 nM) single-chain ␤lac-TT or variants for 30 min at 37°C. Neurons were washed with Hanks' balanced salt solution lacking Ca 2ϩ and Mg 2ϩ (HBSS Ϫ/Ϫ ) (Life Technologies) and cooled to room temperature (RT) for 10 min followed by loading of media with 2 M CCF2-AM (Life Technologies) and 1 mM probenecid (Life Technologies), an anion transport inhibitor, in HBSS Ϫ/Ϫ for 30 min. Neurons were washed and processed for immunofluorescence. The inhibitor-treated translocation assay was performed as described above, with the following preincubation modifications. Bafilomycin A1 (400 nM), a vesicular ATPase inhibitor (Sigma-Aldrich), was preincubated with neurons in neurobasal media for 30 min at 37°C before aspiration and application of 40 nM single-chain ␤lac-TT in neurobasal media containing inhibitor at the indicated concentration. Then, auranofin (500 nM), a thioredoxin reductase inhibitor (Sigma), was preincubated with neurons for 30 min at 37°C before its removal and addition of 40 nM single-chain ␤lac-TT in neurobasal media containing inhibitor at the indicated concentration. After being maintained 30 min with ␤lac-TT at 37°C, cells were washed and loaded with CCF2 as described above followed by fixation.
Data analysis. Secondary Alexa-conjugated antibody-only controls were used to subtract autofluorescence and background fluorescence emitted by cells, yielding the net average intensity of fluorescence. FLAG or HA epitopes were normalized to a cellular marker such as synaptophysin. For protein binding and inhibitor-treated neurons, neurons were localized with guinea pig anti-synaptophysin and rabbit anti-NeuN, respectively. For the measurement of LC translocation following inhibitor treatment, clusters of 3 to 6 neurons were located as NeuN-positive cells with a 60ϫ objective and assessed for cleaved CCF2 substrate above the no-protein control. To compare the translocation efficiencies of cis-loop variants, micrographs were acquired at ϫ20 magnification and neurons containing cleaved CCF2 were scored as positive/negative for ␤lac-LC translocation. At least 10 fields (containing 20 to 30 neurons) were analyzed per experiment, and at least 3 independent experiments were performed. For entry of cis-loop variants into Neuro-2a cells, Z-stacks were acquired and deconvolved. Pearson's correlation coefficient was obtained using Nikon AR analysis software as described previously (12).
Molecular simulations of the beltless TT(HCN). Gangliosides were removed from the crystal structure file of TT (PDB: 5N0B), and the structure was truncated to residues 564 to 870 [TT(HCN)], using PyMol. A coarse-grain structure with an elastic network was generated using the martinize.py script (version 2.2) with secondary-structure assignment provided by the Define Secondary Structure of Proteins (DSSP) program (version 2.0.4). Lipid and amino acid parameters were taken from Martini Lipidome and Martini (version 2.2). Simulations were performed as follows. The force field used for generating the coarse-grain structure was martini22, with an elastic bond force constant (-ef) of 500, an elastic bond upper distance bound (-eu) of 0.9, and an elastic bond lower bound (-el) of 0.5. The lipid bilayer system was generated using the insane.py script with a 20-, 20-, 30-nm box (xyz) with 48% phosphatidylcholine, 31% phosphatidylethanolamine, 9% phosphatidylserine, 5% sphingomyelin, and 3% phosphatidylinositol to mimic the mixtures of lipids found in neuronal vesicles. Using GROMACS 2018 (released 10 January 2018), the energy of the system was minimized for 2,000 steps with a 0.02-ps step size (40 ps) using the steep integrator. The system then underwent temperature equilibration from 0 to 300 K using v-rescale temperature coupling for 10,000 steps with a 0.01-ps step size (100 ps) and the md (leap-frog) integrator. Then, the system was equilibrated to 1 bar using the Berendsen barostat with semi-isotropic pressure coupling for 50,000 steps with a 0.01-ps step size (500ps) and the md (leap-frog) integrator. The system was finally equilibrated to 1 bar using the Parrinello-Rahman barostat with semi-isotropic pressure coupling for 50,000 steps with a 0.01-ps step size (500 ps) and the md (leap-frog) integrator. Simulations were carried out on MCW Research Computing Cluster Tesla using Tesla K40 and K80 graphics processing units (GPUs). Trajectories were converted and analyzed using VMD (version 1.9.3).
Statistics. Student's unpaired, two-tailed t test was utilized to determine if two data sets were significantly different where appropriate.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.