Secondary structure determines electron transport in peptides

47 Proteins play a key role in biological electron transport, but the structure-function 48 relationships governing the electronic properties of peptides are not fully understood. 49 Despite recent progress, understanding the link between peptide conformational 50 flexibility, hierarchical structures, and electron transport pathways has been challenging. 51 Here, we use single-molecule experiments, molecular dynamics (MD) simulations, non-52 equilibrium Green’s function -density functional theory (NEGF-DFT) calculations, and 53 unsupervised machine learning to understand the role of primary amino acid sequence 54 and secondary structure on charge transport in peptides . Our results reveal a two-s tate 55 molecular conductance behavior for peptides across several different amino acid 56 sequences. MD simulations and Gaussian mixture modeling are used to show that this 57 two-state molecular conductance behavior arises due to the conformational flexibility of 58 peptide backbones, with a high-conductance state arising due to a more defined 59 secondary structure (beta turn) and a low-conductance state occurring for extended 60 peptide structures. Conformer selection for the peptide structures is rationalized using 61 principal component analysis (PCA) of intramolecular hydrogen bonding distances 62 along peptide backbones. Molecular conformations from MD simulations are used to 63 model charge transport in NEGF-DFT calculations, and the results are in reasonably 64 good agreement with experiments. Projected density of states (PDOS) calculations and 65 molecular orbital visualizations are further used to understand the role of amino acid 66 side chains on transport. Overall, our results show that secondary structure plays a key 67 role in electron transport in peptides, which provides new avenues for understanding the 68 electronic properties of longer peptides or proteins. 69 70 71 Significance Statement 72 Electron transport in proteins serves as a biological power line that fuels cellular 73 activities such as respiration and photosynthesis. Within cells, proteins act as conduits, 74 shuttling electrons through a series of reactions and pathways to generate proton 75 gradients and to fuel ATP synthesis. Despite recent progress, the mechanisms 76 underlying the flow of energy in protein complexes are not fully understood. Here, we 77 study electron transport in peptides at the single-molecule level by combining 78 experiments and molecular modeling. Our results reveal two distinct molecular sub-79 populations underlying electron transport that arise due to the flexib ility of peptide 80 backbones and the ability to fold into compact structures. This work provides a basis for 81 understanding energy flow in larger proteins or biomolecular assemblies. 82 83 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint 3

84 Electron transport in proteins is essential for maintaining fundamental life processes 85 such as respiration and photosynthesis 1. In recent years, a wide range of experiments 86 and theoretical studies has focused on understanding electron transfer in biological 87 systems 2–4, ranging from redox events in metalloproteins 5,6 and redox-active cofactors 88 7,8 to metal-reducing bacteria 9 . Recent work has shown that proteinaceous nanowire 89 filaments of metal-reducing bacteria such as Geobacter sulfurreducens exhibit 90 remarkable abilities for long-distance electron transport on the micron scale 10,11. During 91 such redox-mediated electron transport events, intervening residues between redox 92 centers are thought to provide a conductive matrix for electron transport 12. However, 93 proteins exhibit complex secondary structures due to intramolecular hydrogen (H)-94 bonding interactions within the underlying conductive protein matrix . Despite recent 95 progress, understanding how secondary structure formation in peptides and proteins 96 affects electron transport is not yet fully understood. 97 98 Electron transport in molecules can occur by different mechanisms such as single-step 99 (coherent) tunneling, multi-step (incoherent) hopping, resonant tunneling, or flickering 100 resonant tunneling 13–15. The dominant mechanism for nanoscale charge transport in 101 short peptide sequences has been reported as non-resonant coherent tunneling 3,4,16–22, 102 where conductance decays exponentially with molecular length. However, electron 103 transport in long peptide or protein sequences also occurs by hopping 7,23, where 104 conductance decreases inversely with distance. The environment around a protein 105 affects the driving force for the electron transfer reaction and the reorganization energy, 106 in accordance with Marcus theory 24. Molecular conformation and intramolecular H-107 bonding that arise due to the protein sequence and environment are pivotal for 108 controlling biological electron transport over long distances 25. Prior work has focused on 109 understanding electron transport in helical peptides 26–29 using bulk conductivity , 110 electrochemistry, thin-film conductivity, or electronic measurements on assembled 111 peptide monolayers 7,19,27. However, key knowledge gaps remain in understanding how 112 other types of secondary structures in peptides and proteins affect electron transport in 113 biological systems. Elucidating electron transport at the single-molecule level holds the 114 potential to provide valuable new insights into the electronic properties of more complex 115 peptide or protein structures. 116 117 Single-molecule techniques offer the ability to characterize conformation-dependent 118 electron transport in the absence of intermolecular interactions in monolayers or bulk-119 scale measurements. In recent years, single-molecule conductance measurements for 120 peptides have primarily focused on short peptide sequences containing up to two or 121 three amino acids 22,30 or chemically functionalized peptides to facilitate metal electrode 122 contact31. However, peptide backbones are generally more flexible compared to -123 conjugated carbon backbones commonly used in synthetic organic electronic materials, 124 and this enhanced backbone flexibility could give rise to conformation-dependent 125 electron transport pathways in oligopeptides. By using the scanning tunneling 126 microscope break junction (STM- BJ) technique, the phenomenon of electron tunneling 127 while pulling 32,33 single molecules has been studied . In addition, it has been reported 128 that a special arrangement of hydrogen bonds 34,35 could give rise to conducting 129 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint 4 pathways in non-conjugated peptide backbones. From this view, single-molecule 130

offer intriguing routes to understand the role of amino acid sequence and the 131 effect of secondary structure on molecular charge transport in peptides, thereby adding 132 new insights into electronic phenomena and their correlation to structure in biomolecular 133 systems. 134 135 The ability to combine molecular simulation with single-molecule electronics 136 experiments provides a powerful approach to understand biophysical processes. The 137 rich conformational space of biomolecules 36 such as peptides 37 can be explored using 138 molecular dynamics (MD) simulations. Biomolecular simulation offers a predictive tool 139 for structural biology due to the high spatial and temporal resolutions and the 140 extensively tested and validated force fields 38,39. MD simulations have been used to 141 understand the influence of atomic structure on the electronic properties of synthetic 142 organic materials by modeling the structural dynamics of molecular junctions 40–42. 143 However, classical force fields are limited in their description of molecular junctions that 144 involve transition metal atoms such as gold. Incorporating Au atoms into classical MD 145 simulations requires either a physical ly rigorous but computationally demanding 146 quantum mechanical (QM) description of gold and its interaction with the surrounding 147 system, or an approximate but more computationally feasible model of interactions with 148 Au atoms. Examples of the latter include representing gold atoms as dummy particles 149 restricted to only interact with specified anchor atoms through harmonic potentials 41 and 150 utilization of reactive force fields to model bond formation and disruption 43. Molecular 151 conformations generated by MD can be used in computationally efficient QM 152 calculations for improved comparison between theory and experimental results. 153 154 In this work, we investigate the role of amino acid sequence and secondary structure on 155 the electronic properties of peptides using a combination of experiments and 156 computational modeling. A key feature of our work lies in using MD simulations to 157 understand the conformational dynamics of molecular junctions in single-molecule 158 charge transport experiments. A scanning tunneling microscope break junction (STM-159 BJ) technique is used to experimentally characterize the molecular charge transport 160 properties of oligopeptide s. Our results reveal a two-state conductance behavior for 161 peptide sequences contain ing 4 or 5 amino acids . Our results further indicate that 162 longer amino acid sequences can show enhanced conductance values for the extended 163 state due to the presence of aromatic or constrained amino acid side chains. Gaussian 164 mixture modeling (GMM) and MD simulations are used to show that this two-state 165 molecular conductance behavior arises due to the conformational flexibility of the 166 peptide backbone. Classical MD simulations with custom potentials for implicitly 167 representing gold are used to understand the molecular basis for conformation-168 dependent electron transport in peptides. Characteristic conformers for each peptide 169 sequence are selected from MD simulations and quantitatively analyzed using principal 170 component analysis (PCA) to understand the role of hydrogen bonding (H-bonding) 171 interactions along the peptide backbone . Interestingly, results from PCA show that 172 specific H-bonding distances between peptide backbone atoms significantly contribute 173 to the structural variation observed in MD simulations. Molecular conformations from 174 MD simulations are then used in non -equilibrium Green’s function -density functional 175 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint 5 theory (NEGF -DFT) calculations to understand the role of molecular conformation on 176 charge transport. Projected density of states (PDOS) calculations and molecular orbital 177 visualization are further carried out to understand the role of amino acid side chains and 178 the underlying transport mechanisms. Our results reveal that an extended peptide 179 sequence gives rise to a low conductance state, whereas a folded conformation (beta 180 turn) gives rise to a high conductance state. Overall, our work highlights the importance 181 of molecular conformation and secondary structure on the electron transport behavior of 182 peptides. 183 184

and Discussion 185 Single-molecule conductance measurements and chemical characterization 186 Tetra- and pentapeptides were designed with different amino acid sequences to 187 understand the role of non-polar aliphatic R groups , aromatic R groups, or sterically 188 constrained R groups on electron transport ( Figures 1a,b,c and Supplementary 189 Figures 1- 10). The N- and C-terminal residues of the tetra- and pentapeptides were 190 selected as methionine, which contains a methyl sulfide (-S- CH3) group that readily 191 binds to gold 44, thereby providing robust electrical contacts to metal electrodes in STM-192 BJ. All STM-BJ measurements on peptides were carried out in water (peptide 193 concentration <1 mM). 194 195 Circular dichroism (CD) spectra were first obtained for all tetra- and pentapeptides in 196 water at room temperature under identical solvent conditions used in STM-BJ 197 experiments (Supplementary Figures 11-15). CD spectra clearly indicate the presence 198 of H-bonding interactions for all tetra- and pentapeptides and show spectral features 199 expected for 3 10 helices, such as a maximum or minimum around ~200-210 nm and a 200 shoulder or small peak around ~220 nm 45,46. CD spectral features for 3 10 helices are 201 qualitatively different than the spectral features observed for alpha helices, beta sheets, 202 or random coils 47. Based on results from CD experiments, the proline and alanine-203 based peptide sequences show minima in CD spectra around 200 nm, which is 204 consistent with a tendency to adopt right-handed 310 helices. On the other hand, peptide 205 sequences containing glycine, tyrosine, and tryptophan show peaks in CD spectra 206 around 200 nm, which is consistent with left- handed 310 helices. Overall, these results 207 clearly indicate the presence of H-bonding interactions amongst the tetra- and 208 pentapeptides characterized in single-molecule electronics experiments. 209 210 We began by characteriz ing the electronic properties of peptides containing non-polar 211 aliphatic R groups. The molecular conductance of oligopeptides was determined using a 212 custom-built STM-BJ instrument ( Figure 1d ), as described in prior work 48,49. Our 213 experiments revealed the presence of two distinct conductance populations, as shown 214 in characteristic single-molecule conductance traces ( Figure 1e). We hypothesized that 215 the high and low conductance states could arise due to a folded , compact conformation 216 and an extended peptide conformation, respectively . Characteristic single-molecule 217 conductance traces for all tetra- and pentapeptides ( Figures 2a,b) indicate that the two 218 conductance states occur in the same individual traces rather than in two separate 219 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint 6 molecular sub-populations. This behavior suggests that dynamic conformational 220 changes during molecular pulling events give rise to multiple conductance states. 221 222 One-dimensional and two-dimensional molecular conductance histograms were 223 generated for the tetra- and pentapeptides across ensembles of >5000 single molecules 224 (Figures 2c,d,e,f and Supplementary Figures 16-17 ). A bimodal conductance 225 distribution is observed for all oligopeptide sequences across the entire range of applied 226 biases (100 mV - 400 mV) studied in this work ( Supplementary Figure 18 ). Bimodal 227 conductance distributions can arise due to conformationally distinct molecular sub-228 populations (static heterogeneity) or due to conformation-dependent conductance 229 during molecular pulling (dynamic heterogeneity). To investigate the origins of this 230 behavior, we determined the most probable conductance of the low and high 231 conductance states from a Lorentzian fit to the conductance data 50 ( Supplementary 232 Tables 1-2). The high-conductance peak (~10-2.80 - 10-2.90 G0) occurs at nearly the same 233 value for all the oligopeptide sequences. The low conductance peak value shows a 234 small dependence on the backbone sequence and side chain composition. In addition, 235 the molecular displacement corresponding to the low conductance peak is significantly 236 larger than the displacement for the high conductance peak. Based on these results, we 237 hypothesized that the low conductance peak arises due to an extended peptide 238 configuration, whereas the high conductance peak is related to a folded or more 239 compact peptide conformation. 240 241 There are some subtle differences in the low conductance state for the tetra- and 242 pentapeptides studied in this work , which suggests that amino acid side chain identity 243 plays a role in transport . For the tetrapeptides, the low conductance state of MGGM is 244 ~0.2-0.3 log G0 lower compared to all other sequences (MAAM, MYYM, MWWM, and 245 MPPM). These results show that changing the amino acid side chain from hydrogen to 246 a methyl, aromatic , or a constrained side chain leads to an enhancement in 247 conductance. For the pentapeptides, the conductance values for MGGGM and MAAAM 248 are approximately half an order- of-magnitude smaller compared to MYYYM, MWWWM, 249 and MPPPM. The higher conductance values for MYYYM and MWWWM indicate that 250 aromatic side chains can lead to enhanced conductance values . MPPPM has a higher 251 conductance compared to the glycine or alanine-based sequences, as proline provides 252 a constrained side chain that reduces the conformational flexibility and increases the 253 rigidity of the molecule . The high er conductance values observed for the extended 254 conformations for peptides containing tyrosine, tryptophan, and proline sequences are 255 also corroborated by NEGF-DFT simulations ( Figure 5 e,f), as discussed below. Based 256 on these results, STM-BJ experiments reveal several intriguing findings regarding the 257 role of amino acid side chains on oligopeptide charge transport. 258 259 Single-molecule data can be quantitatively analyzed using unsupervised learning 260 algorithms to classify molecular charge transport behavior into characteristic groups and 261 to identify underlying structure-property relationships 31,51–54. Here, we use silhou ette 262 clustering55 (Supplementary Figure 19 ) to determine the optimal number of clusters 263 for data sets corresponding to molecular ensembles for each peptide sequence. 264 Silhouette clustering indicates that the optimal number of clusters for all tetra- and 265 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint 7 pentapeptides is two. Gaussian mixture modeling (GMM) is further used to analyze the 266 two different clusters identified by Silhouette clustering ( Supplementary Figures 20-267 21). 268 269

from GMM show that Cluster 1 accounts for 85-95% of the single-molecule 270 traces and shows both characteristic conductance populations appearing together in the 271 same molecular traces. Cluster 2 accounts for only 5-15% of the data and represents 272 traces in which no molecule is detected or only background signal is observed. If the 273 bimodal distribution arose due to stable, conformationally distinct molecular sub-274 populations (static heterogeneity ), then the two characteristic conductance populations 275 would segregate into different clusters. However, our results show that the bimodal 276 conductance populations appear sequentially in single-molecule traces for all tetra- and 277 pentapeptides, which strongly supports conformation-dependent charge transport 278 behavior in peptide backbones (dynamic heterogeneity). 279 MD simulations 280 To understand the role of molecular conformation on charge transport, we performed 281 MD simulations for all tetra- and pentapeptides ( Figures 1a,b,c) in explicit solvent with 282 a series of custom potentials to implicitly represent interactions between peptides and 283 gold electrodes. These custom potentials and their resulting collective variable 284 distributions are shown in Figures 3a,b,c and Supplementary Figure 22. 285 286 The projection of the end- to-end distance (sulfur anchor- to-anchor distance on terminal 287 methionines) of the peptide along the experimental pulling axis was harmonically 288 restrained to a range of values (6 Å, 9 Å, and 12 Å), allowing the peptide to adopt an 289 ensemble of conformations. The conformations observed in MD simulations are not 290 significantly affected by a change in applied voltage ( Supplementary Figure 23), which 291 is consistent with single-molecule conductance experiments. Ramachandran free 292 energy plots 56 ( Supplementary Figures 24-25) were determined for the non-terminal 293 residues for all tetra- and pentapeptides. These results indicate that all sequences can 294 form left-handed or right-handed helices, except for those based on proline , consistent 295 with CD measurements. 296 297

from MD simulations show that backbone hydrogen bonds, which play a key 298 role in defining the secondary structure of the peptide 57, form with remarkable 299 consistency during the 6 Å end- to-end holding of all peptide sequences considered in 300 this work ( Figures 3d,e and Supplementary Figures 26-27). However, H-bonding 301 interactions are completely abolished when the end- to-end distance is restrained to a 302 distance of 12 Å. For the tetra- and pentapeptides considered here, a canonical 303 secondary structure forms at small end-to-end distances, indicative of a beta turn. A 304 beta turn is defined by an H-bond between the carbonyl oxygen of residue i and the 305 amide hydrogen of residue i+3 57. In the tetrapeptides, a 1→4 H -bond is observed, 306 whereas for the penta peptides, a 2→5 H -bond is consistently observed. Two 307 conformers are selected from the 6 Å and 12 Å holding stages ( Figures 3f,g) of each 308 peptide from the peak of the probability distributions of 1→4 H-bond and 2→5 H-bond 309 distances for tetra- and pentapeptides, respectively. It is known that consecutive beta 310 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint 8 turns in a longer peptide sequence give rise to 3 10 helices57. From this view, our work 311 suggests that helical elements play a key role in the charge transport behavior of 312 biomolecules with defined secondary structures. 313 314 We next performed a linear dimensionality reduction on the MD trajectories to quantify 315 how individual interatomic distances contribute to the peptide conformational landscape. 316 The main objective of this analysis is to identify conserved structural differences across 317 all peptides of interest between various end- to-end holding stages ( Figure 4 and 318 Supplementary Figures 28-31). Peptides are represented using a Euclidean distance 319 matrix of the common molecular subgraph shared between all sequences. Using this 320 approach, each peptide’s structural ensemble is projected onto a shared basis. In 321 addition, the i→i+3 H-bonding distances selected as a basis for conformer extraction 322 are well captured in the first two principal components, indicating that these distances 323 contribute significantly to the variance in molecular structure compared to other 324 interatomic distances . Regions of conformational space corresponding to small i→i+3 325 distances are shown to depopulate with increasing inter-anchor displacement across all 326 peptide sequences. Based on these results, MD coupled with unsupervised machine 327 learning (ML)-based data analysis clearly elucidates the key structural features for 328 characterizing tetra- and pentapeptides in molecular junctions, revealing the most 329 probable peptide conformations. The most probable conformations are then used in 330 computationally efficient NEGF-DFT calculations to understand the role of molecular 331 conformation on the electron transport properties of peptides. 332 NEGF-DFT calculations 333 To understand the role of molecular conformation on charge transport in peptide 334 backbones, NEGF-DFT calculations are performed using the most probable simulated 335 MD conformations. NEGF-DFT simulations are carried out for extended and turn 336 conformations for each peptide sequence ( Figures 5a,b ) using the TranSiesta and 337 Tbtrans package (Methods). The transmission probabilities as a function of energy 338 indicate stark differences between the turn and extended peptide conformations 339 (Figures 5c ,d and Supplementary Figures 32-34). The conductance at zero bias 340 differs significantly between the extended and the turn state of the tetra- and 341 pentapeptides ( Figures 5e,f ). Our results show reasonable qualitative agreement 342 between experiments and NEGF-DFT simulations ( Supplementary Table 3-6). Results 343 from the combined approach of using MD simulations with NEGF-DFT simulations 344 support the hypothesis that the low conductance population arises from an extended 345 peptide conformation, whereas the high conductance population is related to a more 346 defined secondary structure (beta turn) in the peptide. Figures 5e,f also corroborate the 347 role of amino acid side chains that was observed in experiments on tetra-and 348 pentapeptides. The glycine-based tetrapeptide sequence has a lower value of the 349 transmission probability near the Fermi level for the extended conformation compared to 350 all other sequences. For the pentapeptides, MGGGM and MAAAM show similar 351 conductance values in the extended state , albeit lower than MYYYM, MWWWM and 352 MPPPM. Overall, NEGF-DFT results qualitatively agree with single-molecule charge 353 transport experiments and provide insights into the role of side chains on oligopeptide 354 charge transport. 355 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint 9 356 Site specific PDOS calculations were carried out for all tetra-and pentapeptides 357 (Supplementary Figure 35 ) in the extended peptide conformation to understand the 358 role of amino acid side chain s on electron transport. For all tetrapeptides, PDOS 359 calculations were performed for two carbon atoms along the backbone in the energy 360 range of -5 to 5 eV ( Supplementary Figures 36 a,b). At the Fermi energy level for the 361 tetrapeptides (~ -2.20 eV), the PDOS has a relatively low value ( Table 7) due to the 362 non-conjugated peptide backbone. These results imp ly that the backbone orbitals in 363 peptides generally yield small er conductance values near the Fermi level compared to 364 fully -conjugated systems. To compare these results for the various tetrapeptides, the 365 behavior near the Fermi energy level was investigated ( Supplementary Figures 36 366 c,d). Around the Fermi energy level, higher PDOS values are observed for sequences 367 containing tyrosine and tryptophan. For all pentapeptides, PDOS calculations were also 368 performed for three carbon atoms along the backbone in the energy range of -5 to 5 eV 369 (Supplementary Figures 37 a,b,c). The DFT-derived Fermi energy level for the various 370 pentapeptides is around -2.20 eV, with molecular LUMO levels being above the Fermi 371 level by at least 200 meV, though it is important to note that standard DFT generally 372 underestimates the molecular HOMO-LUMO gap. At the Fermi energy level, the value 373 of PDOS approach es zero ( Table 8 ), similar to the case of tetrapeptides . A similar 374 analysis was performed for the pentapeptides near the Fermi energy level 375 (Supplementary Figures 37 d,e,f ), showing larger PDOS values for sequences 376 containing tyrosine and tryptophan for the 1st and 2nd carbon atoms along the backbone. 377 For the 3 rd carbon atom, a relatively high PDOS value is observed for MGGGM , 378 MYYYM, and MWWWM. However, the transmission probability for MGGGM is 379 significantly lower compared to MYYYM and MWWWM . Taken together, these results 380 show that the orbitals of the aromatic side chains tend to mix more readily with the 381 backbone orbitals compared to other amino acids, which leads to enhancement in 382 conductance values. 383 PDOS calculations were also carried out for all carbon and hydrogen atoms 384 (Supplementary Figure 38) for MGGM, MYYM, MGGGM, and MYYYM in the extended 385 peptide conformation . Our results indicate significantly higher PDOS values for the 386 sequences containing tyrosine compared to glycine. Overall, these results indicate that 387 oligopeptide sequences with aromatic side chains have more contribution from the 388 backbone orbitals to the overall electronic density and hence molecular conductance. 389 Molecular orbitals were plotted using Siesta 58 and visualized using Vesta 59 390 (Supplementary Figures 39- 42). Here, HOMO, HOMO-1, LUMO and LUMO +1 are 391 plotted for the glycine and tyrosine-based tetra- and pentapeptides using an isosurface 392 value of 0.025. These results illustrate relatively weak coupling between the molecules 393 and electrodes, which is consistent with the transmission function results observed for 394 the oligopeptides, in agreement with the proposed tunneling mechanism. These results 395 further suggest the absence of - stacking interactions between the tyrosine 396 sidechains. Overall, these results are consistent with non-resonant tunneling rather than 397 resonant tunneling or flickering resonant transport mechanisms for electron transport . 398 Prior work by Xiao et al. 22 characterized electron transport in short peptide sequences 399 such as cysteamine-glycine-glycine -cysteine and cysteine-glycine -cysteine, with results 400 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint 10 showing an exponential decay in conductance as a function of molecular length , 401 consistent with single-step tunneling as the dominant transport mechanism. The tetra-402 and pentapeptides based on glycine studied here are of similar length, with the primary 403 difference of methionine as the N- and C-termini amino acids in place of cysteine or 404 cysteamine. Based on these results, and the relatively short distance of transport 405 observed in our molecular junctions (< 1.4 nm60), our results are fully consistent with off-406 resonant coherent tunneling for the oligopeptides studied in this work. 407 408 During the STM-BJ pulling experiments, we observe two conductance states related to 409 two distinct molecular conformations. To further understand the role of H-bonding on 410 transport pathways, we used a bond counting methodology based on the tunneling 411 pathway model61,62. In general, there is a conductance decay associated with through-412 bond, through-space, and through -H-bond electron transport 63. Tunneling is generally 413 more efficient for through-bond compared to through-space transport due to the lower 414 potential barrier 64. As a rule of thumb, it can be assumed that the conductance decay 415 through an H-bond is twice as large compared to the decay through a covalent bond 64. 416 Supplementary Figure 43 indicates that if transport were to occur entirely through -417 bond, then the pathway would be approximately three bonds longer with an order of 418 magnitude smaller decay compared to the case of electron transport through H-bonds65. 419 420 To further understand the importance of H- bonds on transport, we performed control 421 experiments for STM -BJ using 1,16-hexadecanedithiol ( Supplementary Figure 44) in 422 1,2,4-tricholorbenzene. 1,16-hexadecanedithiol has a similar contour length as the 423 peptides studied in this work but with a flexible alkane chain backbone and no possibility 424 of intramolecular H-bonding. Our results show that two conductance populations are 425 observed for the peptides (at ~10 -2.8 G/G0 and 10 -4.2 G/G0), but no significant 426 conductance peaks are observed for the flexible alkane backbones , though a faint 427 population is observed between ~10-1 -10-2 G/G0, which arises due to the use of different 428 anchors and strong binding between the - SH terminal anchor groups and the gold 429 electrode66. Overall, these results show a two order- of-magnitude increase in molecular 430 conductance for a peptide compared to an alkane chain with similar contour length. We 431 further compared these results to prior work in the literature. Inkpen et al. 67 studied 432 charge transport in alkane chains such as C 12(SH)2 and C 12(SMe)2, and only a single 433 conductance population was observed below ~ 10-5 G/G0 for the C 12 sequences. It 434 should be noted that t he average conductance values reported for the C 12 sequences 435 are approximately one order- of-magnitude lower compared to the low conductance 436 state of the 17- or 19-mer oligopeptide sequences studied in this work . Taken together, 437 these results show that the electron transport behavior of alkane chains is significantly 438 different than peptides due to intramolecular backbone H-bonding. 439 440 In this work, we use a combination of single-molecule conductance experiments, MD 441 simulations, and NEGF-DFT calculations to investigate the charge transport properties 442 of a series of different peptide sequences. Our results unequivocally reveal the 443 structure-function relationships governing the observed electron transport in peptides, 444 highlighting the importance of secondary structure on charge transport in biomolecules . 445 Unsupervised learning is used to analyze single-molecule conductance data, showing 446 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint 11 that peptides exhibit a bimodal conductance distribution with a low and high 447 conductance population arising from distinct conformational states of peptide 448 backbones. A key feature of our work lies in using MD simulations to sample and 449 characterize the conformational space of the peptides and to identify conformations to 450 be used in electron transport (NEGF-DFT) calculations . Moving forward, our work could 451 provide new avenues to understand the interplay between molecular charge transport 452 and secondary structure in more complex peptide sequences with mixed amino acids 453 and/or longer peptides. Proteins are candidate materials for fabricating functional 454 molecular electronic devices due to biocompatibility, anti-fouling properties 68, and 455 tunable redox activity due to aromatic amino acids 69. From this view, our work can 456 provide further insights into understanding the role of higher order assembled structures 457 on biological charge transport, which can be used to inform the design self-assembled 458 bioelectronic materials. 459 460

461 462 Oligopeptide sequences 463 All oligopeptide sequences were purchased from GenScript (Piscataway, NJ). Mass 464 spectrometry data for these sequences are provided in the Supplementary Information 465 (Supplementary Figures 1-10). 466 467 Single-molecule conductance measurements 468 Single-molecule conductance measurements were performed using a custom-built 469 scanning tunneling microscope break junction (STM-BJ) 48,49,66. Gold STM tips were 470 prepared using 0.25 mm Au wire (99.998%, Alfa Aesar). STM-BJ experiments were 471 carried out in Milli-Q water (Specific resistance of 18.2 MΩ·cm @ 25 °C ). Due to the 472 polarity of the solvent, STM tips were coated with an Apiezon wax to prevent Faradaic 473 currents from masking characteristic molecular features 70. Gold su bstrates for the 474 measurements were prepared by evaporating 120 nm of gold on polished AFM metal 475 discs (Ted Pella). Peptide concentrations ( 5000 traces) are 477 generated for all molecules without data selection. Silhouette clustering and Gaussian 478 mixture modelling (GMM) were further used to analyze the bimodal conductance 479 distribution (Supplementary Information). 480 481 MD simulations 482 Molecular dynamics (MD) simulations were performed to generate conformational 483 ensembles for the tetra- and pentapeptide molecular junctions at three anchor 484 displacements (referred to as stages 6 Å, 9 Å, and 12 Å ). For each peptide, 16 initial 485 structures were prepared using the PeptideBuilder python package 71. Phi and psi 486 backbone dihedrals of each of the 16 structures were randomized independently. Each 487 backbone dihedral angle of non -proline residues was initialized to a random value 488 between -180 and 180 degrees, whereas the p hi angle of proline was initialized to a 489 random value between -80 and -50 degrees. Hydrogens were added to the peptides 490 with the VMD plugin PSFGEN 72 using the NTER and CTER terminal patches to create 491 positively and negatively charged N- and C-termini, respectively. Peptide structures 492 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint 12 were then solvated in a cubic box of TIP3P water of side length 38 Å using the VMD 493 SOLVATE plugin72. The solvated systems were then subjected to MD simulations with 494 the CHARMM36m protein force field 38,39 using OpenMM 7.7.0 73. Dynamics were 495 integrated using the LangevinMiddleIntegrator 74 with friction coefficient of 1 ps -1, 496 temperature of 300 K, and a timestep of 4 fs. Hydrogen mass repartitioning was not 497 utilized. Bonds involving hydrogen atoms, and all bonds and angles involving water 498 were constrained 74. Nonbonded interactions were computed with a cutoff of 12 Å with 499 smooth switching starting at 10 Å. Electrostatic interactions were evaluated using 500 particle mesh Ewald 75 (PME) summation with error tolerance of 0.0005. Each replicate 501 was simulated for 200 ns for each of three holding stages, for a total aggregate 502 simulation time of 96.0 μs (10 peptides × 3 stages × 16 replicates × 200 ns ). The 503 conformational ensemble of each peptide is shown to converge after 200 ns of 504 simulation per replicate per holding stage ( Supplementary Figures 45-46). Holding 505 stages were enforced using a series of custom external potentials, applied using 506 OpenMM’s custom force classes, described below. The last 190 ns of each simulation 507 was used for subsequent analysis. 508 509 A series of custom potentials were implemented to implicitly represent interactions 510 between the peptide and gold particles. Three potentials were defined: (1) a potential to 511 restrain the distance between the anchors of the molecular junction along the pulling 512 axis to 6 Å, 9 Å, or 12 Å (representing the restraints imposed by connections to the gold 513 electrodes); (2) a per-atom charge-dependent potential along the pulling axis 514 accounting for electric field forces arising from a voltage-biased junction; and, (3) a 515 potential that orients methionine’s thioether moiety such that the average position of 516 each sulfur’s lone pairs are oriented towards the (implicitly represented) gold electrodes 517 along the pulling axis. These potentials are described in detail in the next section and 518 depicted in Supplementary Figure 17. 519 520 After MD simulations, characteristic conformations of each peptide were determined 521 from their aggregate MD trajectories. For each peptide, two conformations were 522 selected from their 6 Å and 12 Å holding-stage simulations at the peaks of their 523 respective hydrogen-bond distance distribution histograms. The H-bond distance 524 distributions used as the basis for conformation selection for the tetra- and 525 pentapeptides were the 1 ⟶4 and 2 ⟶5 distances, respectively. All free energy plots 526 (Figures 4b ,c and Supplementary Figures 21,23 ) were prepared using PyEMMA 527 2.5.1176. 528 529 Custom potential for implicit gold peptide interactions 530 A key challenge for simulating single-molecule pulling processes is large difference 531 between the pulling rates used in experiments and those accessible by MD simulations . 532 Typical experimental pulling rates are on the order of Angstroms per millisecond (1 Å 533 per 5 ms in present study), whereas single-trajectory MD simulations (at most) typically 534 reach ms timescales, e.g., with the use of bespoke hardware 77 or massively distributed 535 computing schemes78. In addition, the need for multiple independent simulation replicas 536 to claim ensemble convergence and statistical certainty of key observables further 537 restricts simulations to sub-experimental timescales. However, because the 538 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint 13 experimental pulling rate is also slow relative to characteristic relaxation timescales of 539 small peptides, we assume that all molecular conformations accessible at a given end-540 to-end distance are sampled during each step of the experimental pulling process. In 541 other words, experimental pulling occurs as an equilibrium process. Rather than 542 performing costly simulations of the entire pulling process, it is more computationally 543 feasible to simulate the molecular junction at various holding (end- to-end distance) 544 stages representing the different separation distances arising during the pulling 545 experiments. 546 547 Using this approach, we perform ed a series of independent simulations where we 548 restrained the end- to-end (sulfur-sulfur) distance along the pulling axis to one of three 549 distances spanning the range of end-to-end distances (6 Å, 9 Å, or 12 Å). We define the 550 pulling axis as the z-axis in our simulations. Schematic illustrations for each potentia l 551 are shown in Supplementary Figure 17. The functional form of the potential utilized to 552 enforce this restraint is given in Equation 1: 553 554 𝑈1 = 1 2 𝑘1[(𝑧𝑆2 − 𝑧𝑠1)− 𝑧0]2, (1) 555 where the coefficient k1 is the force constant of the harmonic potential, zS1 and zS2 are 556 the z-coordinate of the sulfur atoms of the N-terminal and C-terminal methionine 557 residues respectively, and z0 is the equilibrium distance for the given stage. We use a 558 value of 1 kcal/mol/Å 2 for k1, and we utilize three independent holding stages with z0 559 equal to either 6 Å, 9 Å, or 12 Å. This force constant was selected such that the 560 resulting distributions of zS2 – zS1 distances have slight overlap ( Supplementary 561 Figures 17a,d). 562 563 By restraining the z-displacement between the sulfur atoms, rather than the distance, 564 the movement of each sulfur atom is effectively restrained to one of two parallel planes 565 which implicitly represent two parallel planes of gold electrode. 566 567 A potential is introduce d to represent an applied electric field due to the voltage 568 difference across the two electrodes. The functional form is given in Equation 2: 569 570 𝑈2 = ∑ −𝑞𝑖𝐸𝑧𝑖 𝑁𝑎𝑡𝑜𝑚𝑠 𝑖=1 = ∑ −𝑞𝑖 ( 𝑉 𝑧0 + 2𝑙𝑆−𝐴𝑢 )𝑧𝑖 𝑁𝑎𝑡𝑜𝑚𝑠 𝑖=1 (2) 571 where Natoms is the total number of atoms in each system including solvent, qi is the 572 charge of atom i, zi is the z-coordinate of atom i, z0 is the equilibrium end- to-end 573 distance (displacement along z) for a holding stage, and lS-Au is the length of the sulfur-574 gold bond. 575 576 We further introduce a potential to orient each sulfur atom’s lone pairs in either the 577 positive or negative z-direction, such that a feasible dative bond may occur between the 578 sulfur and a fictitious gold particle. This is a key step in ensuring that any conformation 579 generated by MD simulations can be placed into a gold-gold junction for subsequent 580 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint 14 NEGF-DFT calcu lations. Because electron lone pairs are not explicitly represented in 581 atomistic MD simulations, we define surrogate vectors that involve each sulfur’s 582 adjacently bonded carbon atoms to act as a proxy for the direction of the electron lone 583 pairs (Supplementary Figure 17b). We impose a restraint directly on the dot product of 584 each surrogate vector with the pulling axis. The functional form of this potential is shown 585 in Equation 3: 586 587 𝑈3 = ∑ 𝑘3 [|𝑟 ⃗𝑆𝑖 − ( 𝑟 ⃗𝐶𝐺𝑖 − 𝑟 ⃗𝐶𝐸𝑖 2 )| ∙ 𝑧 ⃗] 2 𝑖=1 = ∑ 𝑘3[|𝑝 ⃗𝑖| ∙ 𝑧 ⃗](−1)𝑖 2 𝑖=1 588 where 𝑟 ⃗𝑆𝑖 represents the three-dimensional Cartesian coordinates of the sulfur atom of 589 interest, with S 1 and S 2 subscripts indicating the identity of the sulfur atoms in the N-590 terminal and C-terminal methionine residues, respectively, 𝑟 ⃗𝐶𝐺𝑖 and 𝑟 ⃗𝐶𝐸𝑖 are Cartesian 591 coordinates of the adjacent carbon atoms covalently bonded to each sulfur of interest, 592 and 𝑧 ⃗ is the unit vector in the direction of the z-axis. Vertical lines denote vector 593 normalization. The final term in the equation determines the sign of the potential (and 594 thus the direction of the surrogate vector) allowing for one sulfur’s lone pair to be 595 oriented in the positive z-direction while the other is oriented oppositely in the negative 596 z-direction. The value of k3 is taken as 10 kcal/mol, resulting in a strong potential that 597 tightly secures the orientation of sulfur lone pairs towards the implicitly represented gold 598 electrodes (Supplementary Figures 17c,e). 599 600 Principal component analysis of MD trajectories 601 The resulting MD trajectory data was subjected to dimensionality reduction by means of 602 principal components analysis (PCA). PCA was performed separately for the tetra- and 603 pentapeptides simulations. For the tetrapeptides, the Cartesian coordinates of the 604 peptide backbone heavy atoms were extracted. The Euclidean distance matrix upper 605 triangle was computed for these 17 shared backbone atoms, resulting in a 136-606 dimensional vector representation for each trajectory frame. These vector 607 representations, concatenated across all sequences and holding stages and each 608 interatomic distance, were standardized with Z-score normalization. Finally, the first two 609 principal components were calculated with PCA-whitening using the scikit-learn python 610 package79. PCA of the pentapeptide trajectories was performed following that of the 611 tetrapeptides, with the exception that the shared molecular subgraph of the 612 pentapeptides was instead composed of 21 backbone heavy atoms, resulting in a 210-613 dimensional vector representation for each MD trajectory frame. All other steps were 614 performed identically. 615 616 NEGF-DFT calculations 617 NEGF-DFT calculations are performed with a DFT based non- equilibrium Green’s 618 function (NEGF) approach using the TranSiesta and Tbtrans package 58,80,81. The 619 electrodes contain 8 layers of 16 gold atoms along with a pyramid of 10 Au atoms . 620 Sulfur atoms in the oligopeptide were made to interact with the gold atoms using a 621 trimer binding motif, as described in literature 30. Geometry relaxation of the sequences 622 were performed using generalized gradient approximation-Perdew-Burke -Ernzerhof 623 (GGA-PBE) functional82 using the TranSiesta package 58. SZP basis sets were used for 624 (3) .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint 15 all the gold atoms. DZP basis sets were used for carbon, hydrogen, oxygen, sulfur, and 625 nitrogen. Electrode calculations were carried out with a 4 × 4 × 50 k-mesh. The 626 geometry relaxation was carried out using a 4 × 4 × 1 k-mesh, which was performed till 627 all the forces were < 0.05 eV/Å. After the junction was relaxed, the transport calculations 628 were carried out using the TranSiesta package 80,81 with the same functionals, basis 629 sets, pseudopotential, and k-mesh as the geometry relaxation. Tbtrans 81 was used to 630 carry out the NEGF calculations and to obtain electron transmission as a function of 631 energy (relative to the fermi energy level). NEGF calculations were carried out from - 5 632 eV to 5 eV with 0.05 eV energy increments . The transmission plots are shifted with 633 respect to the Fermi energy values of each peptide. The difference between charged 634 and uncharged species for the MAAM turn configuration (this is a trial sequence, and 635 not the same sequence obtained from the MD simulations using PCA) has been 636 reported in Supplementary Fig. 27 . There are similar qualitative agreements between 637 charged (zwitterionic) and uncharged species. 638 639 PDOS calculations were carried out for the peptides in the molecular junctions from - 5 640 eV to 5 eV using Siesta 58. The PDOS calculations were carried out using two Au 641 pyramids and two Au layers, repeated periodically. The PDOS calculations are carried 642 out and plotted over a suitable energy range such that the Fermi energy level of eac h 643 peptide falls within the interval. For the PDOS calculations, the plane-wave orbitals are 644 projected into atomic orbitals, and the resulting projection coefficients and atomic orbital 645 overlaps that correspond to a given value of the energy in the plot are multiplied 646 together and summed over for each atom of interest. Orbital visualizations were carried 647 out for the molecule with one gold atom on each side using Siesta 58. The orbitals were 648 visualized using Vesta59 to plot HOMO, HOMO-1, LUMO, and LUMO+1 energy levels. 649 650 Corresponding author 651 Further information and requests for resources should be directed to and will be fulfilled 652 by the lead contact Charles M. Schroeder ([email protected]). 653 654 Data Availability 655 Solvent-stripped molecular dynamics trajectories are available at: 656 https://doi.org/10.5281/zenodo.7843691 657 All other data are available from the corresponding author upon request. 658 659 Code Availability 660 STM-BJ data were acquired using a custom instrument controlled by custom software 661 (Igor Pro, Wavemetrics). Codes for MD simulations and analysis are available at: 662 github.com/moeenmeigooni/peptide-conductance 663 664 665 666 667 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint 16 668 669 670 671 672 673

674 1. Stillman, M. Biological Inorganic Chemistry. Structure and Reactivity. Edited by 675 Ivano Bertini, Harry B. Gray, Edward I. Stiefel and Joan S. Valentine. Angew. Chem. 676 Intl. Ed. 46, 8741–8742 (2007). 677 2. Winkler, J. R. & Gray, H. B. Long-range electron tunneling. J. Am. Chem. Soc, 136, 678 2930–2939 (2014). 679 3. Amdursky, N. et al. Electronic Transport via Proteins. Adv. Mat. 26, 7142–7161 680 (2014). 681 4. Gray, H. B. & Winkler, J. R. Electron tunneling through proteins. Qtly. Rev. of 682 Biophysics 36, 341–372 (2003). 683 5. Fereiro, J. A. et al. Tunneling explains efficient electron transport via protein 684 junctions. Proc. Natl Acad. Sci. USA 115, 4577–4583 (2018). 685 6. Nocera, D. G., Winkler, J. R., Yocom, K. M., Bordignon, E. & Gray, H. B. Kinetics of 686 Intramolecular Electron Transfer from Ru11 to FeIH in Ruthenium-Modified 687 Cytochrome c. J. Am. Chem. Soc 106, 5145-5150 (1984). 688 7. Shipps, C. et al. Intrinsic electronic conductivity of individual atomically resolved 689 amyloid crystals reveals micrometer-long hole hopping via tyrosines. Proc. Natl 690 Acad. Sci. USA 118, e2014139118 (2021) 691 8. Záliš, Stanislav, et al. Photoinduced hole hopping through tryptophans in 692 proteins. Proc Natl Acad Sci USA 118, e2024627118 (2021) 693 9. Wang, F. et al. Structure of Microbial Nanowires Reveals Stacked Hemes that 694 Transport Electrons over Micrometers. Cell 177, 361-369 (2019). 695 10. Ru, X., Zhang, P. & Beratan, D. N. Assessing Possible Mechanisms of Micrometer-696 Scale Electron Transfer in Heme-Free Geobacter sulfurreducens Pili. J. Phys. 697 Chem. B 123, 5035–5047 (2019). 698 11. Dahl, P. J. et al. A 300-fold conductivity increase in microbial cytochrome nanowires 699 due to temperature-induced restructuring of hydrogen bonding networks. Sci. Adv. 8, 700 eabm7193 (2022). 701 12. Williamson, H. R., Dow, B. A. & Davidson, V. L. Mechanisms for control of biological 702 electron transfer reactions. Bioorg. chem. 57, 213–221 (2014). 703 13. Hines, T. et al. Transition from tunneling to hopping in single molecular junctions by 704 measuring length and temperature dependence. J. Am. Chem. Soc. 132, 11658–705 11664 (2010). 706 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint 17 14. Zhang, Y. et al. Biological charge transfer via flickering resonance. PNAS 111, 707 10049-10054 (2014). 708 15. Li, Songsong, et al. Transition between nonresonant and resonant charge transport 709 in molecular junctions. Nano letters 21, 8340-8347 (2021). 710 16. Cordes, M. & Giese, B. Electron transfer in peptides and proteins. Chem. Soc. Rev. 711 38, 892–901 (2009). 712 17. Juhaniewicz, J., Pawlowski, J. & Sek, S. Electron Transport Mediated by Peptides 713 Immobilized on Surfaces. Israel J. Chem. 55, 645–660 (2015). 714 18. 16. Scullion, L. et al. Large conductance changes in peptide single molecule 715 junctions controlled by pH. J. Phys. Chem. C 115, 8361–8368 (2011). 716 19. Baghbanzadeh, M. et al. Charge Tunneling along Short Oligoglycine Chains. Angew. 717 Chem. 127, 14956–14960 (2015). 718 20. Juhaniewicz, J. & Sek, S. Peptide molecular junctions: Distance dependent electron 719 transmission through oligoprolines. Bioelectrochemistry 87, 21–27 (2012). 720 21. Guo, Cunlan, et al. Tuning electronic transport via hepta-alanine peptides junction by 721 tryptophan doping. PNAS 113, 10785-10790 (2016). 722 22. Xiao, Bingqian Xu, and Tao. Conductance titration of single-peptide 723 molecules. JACS 126 5370-5371 (2004). 724 23. Malak, R. A., Gao, Z., Wishart, J. F. & Isied, S. S. Long-range electron transfer 725 across peptide bridges: The transition from electron superexchange to hopping. J. 726 Am. Chem. Soc. 126, 13888–13889 (2004). 727 24. Marcus, R. A., Sutin, N. & Amos, A. Electron transfers in chemistry and biology. 728 Biochimica et Biophysica Acta 811, 265-322 (1985). 729 25. Beratan, David N., J. N. Betts, and J. N. Onuchic. Protein electron transfer rates set 730 by the bridging secondary and tertiary structure. Science 252, 1285-1288 (1991). 731 26. Amdursky, N. Electron Transfer across Helical Peptides. ChemPlusChem 80, 1075–732 1095 (2015). 733 27. Sepunaru, L. et al. Electronic transport via homopeptides: The role of side chains 734 and secondary structure. J. Am. Chem. Soc. 137, 9617–9626 (2015). 735 28. Mandal, H. S. & Kraatz, H. B. Electron transfer mechanism in helical peptides. J. 736 Phys. Chem. Lett. 3, 709–713 (2012). 737 29. Horsley, J. R., Yu, J., Moore, K. E., Shapter, J. G. & Abell, A. D. Unraveling the 738 interplay of backbone rigidity and electron rich side-chains on electron transfer in 739 peptides: The realization of tunable molecular wires. J. Am. Chem. Soc. 136, 740 12479–12488 (2014). 741 30. Brisendine, J. M. et al. Probing Charge Transport through Peptide Bonds. J. Phys. 742 Chem. Lett. 9, 763–767 (2018). 743 31. Stefani, D. et al. Conformation-dependent charge transport through short peptides. 744 Nanoscale 13, 3002-3009 (2021). 745 32. Lin, Jianping, and David N. Beratan. Tunneling while pulling: the dependence of 746 tunneling current on end-to-end distance in a flexible molecule. J. Phys. Chem 747 A 108, 5655-5661 (2004). 748 33. Schneebeli, Severin T., et al. Single-molecule conductance through multiple π− π-749 stacked benzene rings determined with direct electrode-to-benzene ring 750 connections. JACS 133, 2136-2139 (2011). 751 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint 18 34. Zhang, Bintian, et al. Role of contacts in long-range protein conductance. PNAS 752 116, 5886-5891 (2019). 753 35. Kretchmer, Joshua S., et al. Fluctuating hydrogen-bond networks govern anomalous 754 electron transfer kinetics in a blue copper protein. PNAS 115, 6129-6134 (2018). 755 36. Ciudad, S. et al. Aβ(1-42) tetramer and octamer structures reveal edge conductivity 756 pores as a mechanism for membrane damage. Nat. Commun. 11, 3014-3028 757 (2020). 758 37. Biswas, M., Lickert, B. & Stock, G. Metadynamics Enhanced Markov Modeling of 759 Protein Dynamics. J. Phys. Chem. B 122, 5508–5514 (2018). 760 38. Brooks, B. R. et al. CHARMM: The biomolecular simulation program. J. Comput. 761 Chem. 30, 1545–1614 (2009). 762 39. Huang, J. et al. CHARMM36m: An improved force field for folded and intrinsically 763 disordered proteins. Nat. Methods 14, 71–73 (2016). 764 40. Wang, H. & Leng, Y. Gold/Benzenedithiolate/Gold Molecular Junction: A Driven 765 Dynamics Simulation on Structural Evolution and Breaking Force under Pulling. J. 766 Phys. Chem. C 119, 15216–15223 (2015). 767 41. Mejía, L., Renaud, N. & Franco, I. Signatures of Conformational Dynamics and 768 Electrode-Molecule Interactions in the Conductance Profile during Pulling of Single-769 Molecule Junctions. J. Phys. Chem. Lett. 9, 745–750 (2018). 770 42. Strange, M., Lopez-Acevedo, O. & Häkkinen, H. Oligomeric gold-thiolate units define 771 the properties of the molecular junction between gold and benzene dithiols. J. Phys. 772 Chem. Lett. 1, 1528–1532 (2010). 773 43. Li, Z. & Franco, I. Molecular Electronics: Toward the Atomistic Modeling of 774 Conductance Histograms. J. Phys. Chem. C 123, 9693–9701 (2019). 775 44. Batra, A. et al. Tuning rectification in single-molecular diodes. Nano Lett. 13, 6233–776 6237 (2013). 777 45. Kumar, P., Paterson, N. G., Clayden, J., & Woolfson, D. N., De novo design of 778 discrete, stable 310-helix peptide assemblies. Nature, 607(7918), 387-392 (2022). 779 46. Brown, R. A., Marcelli, T., De Poli, M., Solà, J., & Clayden, J. (2012). Induction of 780 unexpected left‐handed helicity by an N‐terminal L‐amino acid in an otherwise 781 achiral peptide chain. Angewandte Chemie, 124(6), 1424-1428 (2012). 782 47. Wei, Y., Thyparambil, A. A., & Latour, R. A. Protein helical structure determination 783 using CD spectroscopy for solutions with strong background absorbance from 190 to 784 230 nm. (BBA)-Proteins and Proteomics, 1844(12), 2331-2337 (2014). 785 48. Li, S. et al. Charge Transport and Quantum Interference Effects in Oxazole-786 Terminated Conjugated Oligomers. J. Am. Chem. Soc. 141, 16079–16084 (2019). 787 49. Li, B. et al. Intrachain Charge Transport through Conjugated Donor-Acceptor 788 Oligomers. ACS Appl. Electron. Mater. 1, 7–12 (2019). 789 50. Hybertsen, M. S. et al. Amine-linked single-molecule circuits: Systematic trends 790 across molecular families. J. of Phys. Condens. Matter. 20, 374115 (2008). 791 51. Bamberger, N. D., Ivie, J. A., Parida, K. N., McGrath, D. v. & Monti, O. L. A. 792 Unsupervised Segmentation-Based Machine Learning as an Advanced Analysis 793 Tool for Single Molecule Break Junction Data. J. Phys. Chem. C 124, 18302–18315 794 (2020). 795 52. Lin, L. et al. Spectral clustering to analyze the hidden events in single-molecule 796 break junctions. J. Phys. Chem. C 125, 3623–3630 (2021). 797 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint 19 53. Liu, B., Murayama, S., Komoto, Y., Tsutsui, M. & Taniguchi, M. Dissecting Time-798 Evolved Conductance Behavior of Single Molecule Junctions by Nonparametric 799 Machine Learning. J. Phys. Chem. Lett. 11, 6567–6572 (2020). 800 54. Cabosart, D. et al. A reference-free clustering method for the analysis of molecular 801 break-junction measurements. Appl. Phys. Lett. 114, 143102 (2019). 802 55. Rousseeuw, P. J. Silhouettes: a graphical aid to the interpretation and validation of 803 cluster analysis. J. Compu. and App. Math. 20, 53-65 (1987). 804 56. Hollingsworth, Scott A. and Karplus, P. Andrew. "A fresh look at the Ramachandran 805 plot and the occurrence of standard structures in proteins" Biomolecular Concepts 1, 806 271-283 (2010) 807 57. Kabsch, W. & Sander, C. Dictionary of protein secondary structure: Pattern 808 recognition of hydrogen‐bonded and geometrical features. Biopolymers 22, 2577–809 2637 (1983). 810 58. Soler, J. M. et al. The SIESTA method for ab initio order-N materials simulation. J. 811 Phys.: Condens. Matter. 14, 2745-2781 (2002). 812 59. Momma, K. & Izumi, F. VESTA 3 for three-dimensional visualization of crystal, 813 volumetric and morphology data. J Appl Crystallogr 44, 1272–1276 (2011). 814 60. Page, Christopher C., et al. Natural engineering principles of electron tunnelling in 815 biological oxidation–reduction. Nature 402, 47-52 (1999). 816 61. Beratan, D. N., Onuchic, J. N. & Hopfield, J. J. Electron tunneling through covalent 817 and noncovalent pathways in proteins. J Chem. Phys. 86, 4488–4498 (1987). 818 62. Wuttke, Deborah S., et al. Electron-tunneling pathways in cytochrome 819 c. Science 256, 1007-1009 (1992). 820 63. Onuchic, J. N. & Beratan, D. N. A predictive theoretical model for electron tunneling 821 pathways in proteins. J Chem. Phys. 92, 722–733 (1990). 822 64. Beratan, D. N., Onuchic, J. N., Winkler, J. R. & Gray, H. B. Electron-Tunneling 823 Pathways in Proteins. Science 258, 1740–1741 (1992). 824 65. Betts, J. N., Beratan, D. N., & Onuchic, J. N. Mapping electron tunneling pathways: 825 an algorithm that finds the" minimum length"/maximum coupling pathway between 826 electron donors and acceptors in proteins. JACS, 114(11), 4043-4046 (1992). 827 66. Inkpen, M. S. et al. Non-chemisorbed gold–sulfur binding prevails in self-assembled 828 monolayers. Nat. Chem. 11, 351–358 (2019). 829 67. Venkataraman, Latha, et al. Single-molecule circuits with well-defined molecular 830 conductance. Nano letters 6, 458-462 (2006). 831 68. Hostert, J. D. et al. Self-Assembly and Rearrangement of a Polyproline II Helix 832 Peptide on Gold. Langmuir 37, 6115–6122 (2021). 833 69. Creasey, R. C. G. et al. Biomimetic Peptide Nanowires Designed for Conductivity. 834 ACS Omega 4, 1748–1756 (2019). 835 70. Nagahara, L. A., Thundat, T. & Lindsay, S. M. Preparation and characterization of 836 STM tips for electrochemical studies. Rev. Sci. Instr. 60, 3128–3130 (1989). 837 71. Tien, M. Z., Sydykova, D. K., Meyer, A. G., & Wilke, C. O. PeptideBuilder: A simple 838 Python library to generate model peptides. PeerJ 1, e80 (2013). 839 72. Humphrey, W., Dalke, A., & Schulten, K. VMD: visual molecular dynamics. J. mol. 840 graphics 14, 33-38 (1996). 841 73. Eastman, P. et al. OpenMM 7: Rapid development of high performance algorithms 842 for molecular dynamics. PLoS Comput. Biol. 13, e1005659 (2017). 843 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint 20 74. Zhang, Z., Liu, X., Yan, K., Tuckerman, M. E. & Liu, J. Unified Efficient Thermostat 844 Scheme for the Canonical Ensemble with Holonomic or Isokinetic Constraints via 845 Molecular Dynamics. J. Phys. Chem. A 123, 6056–6079 (2019). 846 75. Darden, T., York, D. & Pedersen, L. Particle mesh Ewald: An N·log(N) method for 847 Ewald sums in large systems. J. Chem. Phys. 98, 10089–10092 (1993). 848 76. Scherer, M. K. et al. PyEMMA 2: A Software Package for Estimation, Validation, 849 and Analysis of Markov Models. J. Chem. Theory Comput. 11, 5525–5542 (2015). 850 77. David shaw, by E. et al. Anton, a Special-Purpose Machine for Molecular Dynamics 851 Simulation. Commun. of the ACM 51, 91-97 (2008). 852 78. Shirts, M. & Pande, V. S. Screen savers of the world unite. Science 290 1903–1904 853 (2000). 854 79. Pedregosa, Fabian, et al. "Scikit-learn: Machine learning in Python. The J. MLR 12, 855 2825-2830 (2011) 856 80. Brandbyge, M., Mozos, J. L., Ordejón, P., Taylor, J. & Stokbro, K. Density-functional 857

for nonequilibrium electron transport. Phys. Rev. B 65, 1654011–16540117 858 (2002). 859 81. Papior, N., Lorente, N., Frederiksen, T., García, A. & Brandbyge, M. Improvements 860 on non-equilibrium and transport Green function techniques: The next-generation 861 TRANSIESTA. Comput. Phys. Commun. 212, 8–24 (2017). 862 82. Perdew, J. P., Burke, K. & Ernzerhof, M. Generalized Gradient Approximation Made 863 Simple. Phys. Rev. lett. 77, 3865-3869 (1996). 864 865 866 867 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint 21 Figure Captions 868 Figure 1: Schematic of experimental setup and chemical structures of peptides studied 869 in this work. Structures of tetra- and pentapeptide with (a) nonpolar aliphatic, (b) 870 aromatic, or (c) sterically constrained R-groups. (d) Schematic of a single-molecule 871 junction containing peptide with sequence Met- Ala-Ala-Met (MAAM) using 872 conformations from MD simulations. (e) Characteristic single-molecule trace for a 873 peptide with sequence MAAM revealing two distinct conductance populations. 874 Figure 2: Scanning tunneling microsco pe-break junction (STM-BJ) measurements of 875 oligopeptides at 250 mV applied bias. (a), (b) Characteristic single-molecule traces for 876 tetra- and pentapeptides. (c), (d) 1D conductance histograms for the tetra- and 877 pentapeptides. (e), (f) 2D conductance histograms for MGGM and MGGGM. 878 Figure 3: Molecular dynamics (MD) simulations methodology and results. (a) Inter-879 anchor displacement potentials at 6 Å, 9 Å, and 12 Å holding stages defined in Eq. 1 . 880 (b) Schematic of applied electric field defined in Eq. 2. (c) Sulfur-orienting potential 881 defined in Eq. 3. (d), (e) Violin plots showing backbone H-bonding distance distribution 882 for tetra- and pentapeptides indicating elimination of intramolecular H-bonds at larger 883 displacements (12 Å). Triangles indicate peaks in the molecular extension distribution at 884 which peptide conformers are selected for NEGF-DFT calculations . (f), (g) Snapshots 885 for tetra- and pentapeptide conformations at small (blue, 6 Å) and large (green, 12 Å ) 886 displacements. 887 Figure 4: Principal component analysis (PCA) results showing the effect of end- to-end 888 stretching on peptide conformation and molecular descriptors. (a) Structure of MAAM 889 turn conformation from MD simulations shown in a single-molecule junction. (b), (e) 890 Principal component projections for M AAM turn conformational landscapes denoted 891 with respect to energy and H-bonding distance at a 6 Å holding stage . (d) Structure of 892 MAAM extended conformation from MD simulations shown in a single-molecule 893 junction. (c), (f) Principal component projections for M AAM extended conformational 894 landscapes colored with respect to energy and H-bonding distance a 12 Å holding 895 stage. Regions of conformational space corresponding to low i → i+3 backbone H-bond 896 distances are shown to deplete with increasing inter-anchor displacements as denoted 897 by the black dotted circle. Blue and green crosses correspond to MAAM turn and 898 extended conformations, respectively. 899 Figure 5: Non-equilibrium Green’s function -density functional theory (NEGF-DFT) 900 calculations for electron transport. (a), (b) Schematic of molecular junctions showing 901 gold metal electrodes and MGGM turn and extended conformations for NEGF-DFT 902 calculations. (c), (d) Transmission probability as a function of energy (relative to the 903 Fermi energy level) for MGGM and MGGGM, showing drastic differences in 904 transmission probability at E - E F = 0 for the turn (blue) and the extended (green) 905 configurations. (e), (f) Zero bias conductance for tetra- and pentapeptides indicating 906 large differences in transmission probabilities between the two conformational states. 907 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint 22 Acknowledgments 908 This work was supported by the U.S. Department of Energy, Office of Science, Basic 909 Energy Sciences under Award No. DE-SC0022035 for M. Meigooni, H.Y., J.L., E.T. , 910 X.L, and C.M.S. and the National Science Foundation under Award 2227399 for R.S. 911 and C.M.S. 912 Author contributions 913 R.S., M. Meigooni, and C.M.S. conceived this study. R.S. performed STM-BJ 914 experiments and NEGF-DFT calculations. M. Meigooni performed MD simulations and 915 PCA analysis. H.Y. assisted with control experiments and J.L. assisted with GMM 916 modeling. X.L assisted with the CD meas urements. M. Mosquera and N.E.J. assisted 917 with the NEGF-DFT calculations. E.T. and C.M.S. supervised the research. The 918 manuscript was written by R.S., M. Meigooni, E.T. and C.M.S. with contributions from all 919 authors. 920 Competing Interests 921 The authors declare no competing interests. 922 Additional Information 923 Supplementary information contains supplementary figures, supplementary tables, and 924 supplementary text. 925 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint a b c d e .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint a b c d e f .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint a c b d e f g .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint a b c d e f .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint a b c d e f .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 20, 2024. ; https://doi.org/10.1101/2024.02.18.578245doi: bioRxiv preprint