Supplementary Information for Thermally activated charge transport in microbial protein nanowires

The bacterium Geobacter sulfurreducens requires the expression of conductive protein filaments or pili to respire extracellular electron acceptors such as iron oxides and uranium and to wire electroactive biofilms, but the contribution of the protein fiber to charge transport has remained elusive. Here we demonstrate efficient long-range charge transport along individual pili purified free of metal and redox organic cofactors at rates high enough to satisfy the respiratory rates of the cell. Carrier characteristics were within the orders reported for organic semiconductors (mobility) and inorganic nanowires (concentration), and resistivity was within the lower ranges reported for moderately doped silicon nanowires. However, the pilus conductance and the carrier mobility decreased when one of the tyrosines of the predicted axial multistep hopping path was replaced with an alanine. Furthermore, low temperature scanning tunneling microscopy demonstrated the thermal dependence of the differential conductance at the low voltages that operate in biological systems. The results thus provide evidence for thermally activated multistep hopping as the mechanism that allows Geobacter pili to function as protein nanowires between the cell and extracellular electron acceptors.

monolaterally, maximizing interactions with insoluble electron acceptors such as Fe(III) oxides, which are often dispersed and otherwise difficult to reach 14 .
Unlike other bacterial type IV pilins, the structure of the Geobacter pilin is divergent and predicted to be optimized for ET 19 . The pilin retains, for example, the conserved α -helix and amino acids required for assembly but has a short, carboxy-terminal (C-t) random coil instead of the β -stranded head of other pilins 19 . The Geobacter pilin also contains aromatic residues (tyrosines and phenylalanines) and charged amino acids predicted to distribute the charges and density of states along the peptide to favor charge transport 19 . These predictions are in agreement with the conductivity measured on cell-associated and mechanically-sheared pili 14,20,21 and the reduced in vitro 22 and in vivo 23 conductivity of pili carrying mutations in aromatic amino acids. Two incongruent models have been proposed to explain how the pili could transport charges. A metallic model was proposed to explain the metallic-like temperature-dependence of piliated, electrochemically active biofilms and preparations of pili and other proteins, including c-cytochromes, dried on gold electrodes 24 . However, in situ electrical conductivity measurements on living biofilms grown on an electrode demonstrated the thermal dependence of incoherent redox conductivity 25 . Similarly, a structural model of the pilus refined via molecular dynamics (MD) identified paths of aromatic residues for transversal and axial multistep hopping 23 . However, although some inter-aromatic distances were optimal for π stacking (~3.5 Å), the geometry of the contacts was never of the sandwich type needed for metallic conductance 24 . Furthermore, the aromatic contacts never formed at the same time, as would have been expected in a metal wire 23 .
In the present study, we isolated pili free of metals and organic redox cofactors and used scanning probe methods to evaluate the contribution of the protein fiber to charge transport. We demonstrate efficient charge transport along individual pili and the involvement of aromatic contacts of the predicted axial, multistep hopping path 23 . We also show a temperature effect on the pilus differential tunneling transversal conductance that is consistent with an incoherent hopping model at biological voltages. The implications of these findings for the role of these conductive appendages as biological protein nanowires and their applications in nanoelectronics are discussed.

Results
Charge transport along individual pili. Individual pili were deposited across the edge of gold electrodes nanofabricated onto an electrically insulating SiO 2 substrate and the tip of a conductive-probe atomic force microscope (CP-AFM) was positioned on regions of the pilus at various distances from the electrode edge to form the contacts for a two-point electronic transport measurement (Fig. 1A). The current along the pilus was then measured while ramping a bias voltage of ±1 V between the electrode and the conductive tip to acquire current-voltage (I-V) plots. Figure 1B illustrates the I-V responses for a pilus probed at distances between 50 to 900 nm from the gold edge and the linear fits (R 2 = 0.98-0.99 for all) obtained within a ± 0.6 V voltage range. Additional I-V curves are shown in Fig. S1. As controls, we periodically measured the response from the bare gold electrode and the lack of current detected when the tip was positioned on the insulating silica substrate (Fig. 1B, inset). To account for technical and biological variability, we performed transport measurements along 13 pilus fibers purified from four biological cultures and estimated the electrical resistance of the pilus from the linear portion of each of the I-V plots.
From the linear resistance-distance correlation (Fig. 1C) we estimated an average resistance value of approximately 730 MΩ along a 1-μm long pilus, with a negligible contribution from the contact resistance (30 ± 8 MΩ; Fig. S2). This resistance value corresponds to an average electron transport rate (~9 × 10 8 electrons per second at 100 mV) two orders of magnitude higher than the cellular rate of Fe(III) oxide respiration (~9 × 10 6 electrons per cell per second; Fig. S3). The measured pilus conductivity ranged from 1.4 to 4.3 S/cm, depending on whether the diameter of the immobilized pilus (2 nm height measured by AFM, Fig. S4) or the fiber core modeled in solution (3.5 nm) 23 was considered. These values are within the orders reported for chemically-fixed membrane protrusions of S. oneidensis containing c-cytochrome complexes (1 S/cm) 16 , which are also high enough to satisfy the respiratory needs of the cell 26 .
Contribution of redox cofactors to charge transport. The range of resistivity (0.23-0.70 Ω.cm) calculated for the immobilized and solvated pilus (2 and 3.5 nm, respectively) is within the lowest ranges reported for moderately doped Ag and Cu nanowires 27 . However, inductively coupled plasma-atomic emission spectroscopy (ICP-AES) did not detect any metals bound to the pili. The ICP-AES analyses did detect Ca 2+ (1.7 ± 0.9 atoms per pilin subunit), a cation that is typically detected in pili purified from other bacteria 28 due to its high affinity for the pilus surface negative charges 29 . Absorption and fluorescence spectroscopy did not detect organic redox cofactors either (Fig. 2). The pili's absorption spectrum ( Fig. 2A) displayed, for example, strong emission below 230 nm and at ~270 nm (peptide bonds and tyrosines, respectively) 30 but had no peaks in the visible region (at ~360 and ~450 nm) where flavins absorb 31 (Fig. S5). The Soret band (406 nm) and broad band (528 nm) of oxidized c-cytochromes such as OmcS 32 , which has been proposed to associate to the pilus and contribute to ET 33 , was not present either. Similarly, the pili's fluorescence emission spectrum contained the excitation peaks of tyrosine and its ionized form, tyrosinate (ca. 300 and 340 nm, respectively) 34 , but did not have peaks in the spectral region of quinone and flavin emission (Fig. S5) 35 .
Contribution of aromatic residues to charge transport. The MD pilus model revealed a potential axial multistep hopping path involving aromatic residues (tyrosines and phenylalanines) 23 . A snapshot from the MD simulations showing a predicted aromatic path is shown in Fig. 3A, as a reference. Fast rates of charge transport are predicted to require the formation of intramolecular (Y27-F24) and intermolecular (F24-F54) contacts (inter-aromatic distances less than 5 Å) 23 . Consistent with this prediction, replacing Y27 with an alanine (Y27A) maintained inter-aromatic distances (5.7 to 10.4 Å) similar to the wild type (WT) pilus ( Fig. 3B), which are optimal for charge hopping 36 , but reduced the number of contacts in the Y27A pilus (Fig. 3C). The Y27A aromatic contacts were also more transient because they formed between F54 and a tyrosine (Y57) of the pilin's C-t random coil, which fluctuates greatly, and between F1 and Y32 or F51, which are located further apart (8-10 Å) (Fig. 3B). As a result, whereas each WT pilin forms, on average, two or more contacts, Y27A pilins only form 1 to 1.5 contacts, thereby reducing the average number of contacts by almost half throughout the MD simulations (Fig. 3D). Not surprisingly, the electrical resistance (~4 GΩ; 1 μm pilus) of the Y27A pilus increased 5-fold ( Fig. 1C and Fig.  S6). Furthermore, the pilus conductivity decreased from 4.3 to 0.77 S/cm and its resistivity increased from 0.23 to 1.3 Ω.cm (2-nm diameter).
The conformational arrangements in the Y27A pilus also exposed more positively charged side chains on the pilus surface (Fig. 3E). These regions are smaller and neutralized in the WT pilus, an arrangement predicted to minimize charge trapping 23 . Thus, the exponential increases in Y27A pilus resistance with distance ( Fig. 1C) are likely to result from the cumulative scattering effects resulting from the filling of the surface charge traps.
Carrier mobility and concentration. We screened the I-V curves acquired for the WT and Y27A pili to identify those that exhibited nonlinear features at high voltages that could result from space charge limited transport (SCLT), as described in the Supplementary Information. This behavior is dramatically enhanced in thin wires because their large surface-to-volume ratio effectively lowers the carrier concentration at the point of contact with the electrode and a layer of depleted charge forms in the contact region that can be injected at a sufficiently high voltage (the crossover voltage or V C ) 37 . As a result, the ohmic (I ~ V) regime transitions into an electronic transport regime in which I ~ V 2 38 . Figure 4 shows representative I-V curves for the WT and the Y27A pili showing this type of transition. The power-law fit above the V C follows the equation I = αV b , where the exponent b is close to 2, as expected for SCLT curves 38 . From the WT scale factor (a = 1.7 × 10 −9 A/V 2 ) and the length of probing (1.6 μm) we estimated a carrier mobility of 3.2 × 10 −2 cm 2 /Vs (see Supplementary Information). This value is small compared to most metal and semiconductor nanowires, including carbon nanotubes, whose mobilities are typically in the tens or hundreds of cm 2 /Vs 39 , but is within the ranges reported for some organic semiconductors such as the hole mobilities of alkyl-substituted polythiophenes (~0.1 cm 2 /Vs) 40 and the electron mobilities of poly(fluorene)-based polymers (10 −3 -10 −2 cm 2 /Vs) 41 .
From the fitting parameters (scale factor, 4.8 × 10 −10 A/V 2 ; distance, 1.1 μm) of the Y27A I-V curve (Fig. 4) we estimated a carrier mobility (6.3 × 10 −3 cm 2 /Vs) one order of magnitude lower than the WT, as expected for a pilus with reduced aromatic contacts. By contrast, the effective carrier concentration, calculated as the number of conducting electrons (or holes) per unit volume, was similar in the WT (2 × 10 20 cm −3 ) and Y27A (2.4 × 10 20 cm −3 ) pili. This value is high compared to most nanowire systems and organic semiconductors, even after extensive doping. For example, InAs nanowires, which also exhibit SCLT, have carrier concentrations on the order of 10 16    Temperature-dependence of pilus electron energy spectrum. We capitalized on the higher spatial resolution of scanning tunneling microscopy (STM) compared to CP-AFM to gain insights into the electronic structure of the WT pili and produce electron energy spectra of pili as a function of temperature. STM images of individual pili deposited on highly oriented pyrolytic graphite (HOPG) showed conducting filaments at room and cryogenic (77 K) temperatures (Fig. 5A,G, respectively), but more periodic molecular sub-structures were resolved at room temperature (Fig. 5A). These features correspond to regions on the pili that supply more tunneling current due to an increase in the local electronic density of states. The room temperature STM scans resolved, for example, 3-4 nm periodic topographic substructures interspersed with deeper 14 nm features (Fig. S7), which match well with the periodic grooves and ridges of type IV pili 43 . Furthermore, from the apparent pilus height (0.8-1.2 nm) and width (8 nm) (Fig. S7), we extrapolated the tip shape-induced convolution and estimated an actual pilus diameter of 4-5 nm, as reported for cell-associated pili by STM 21 and for the solvated pilus fiber core with its C-t random coil (4.7 nm) 23 . By contrast, pili purified from the bacterium Pseudomonas aeruginosa strain K (PAK) could only be imaged at large sample voltages (~3 V) and low tunneling current set points (Fig. 5D). At these high voltages, tip instabilities often arise due to the large electric field between the tip and the sample, which causes distortions and noise in the imaged data.
I-V measurements taken at fixed locations on the pilus filaments are consistent with conducting behavior of the Geobacter pili in the ± 1 V voltage range and the insulating behavior of the PAK pili in the ± 2 V range (Fig. 5B,E, respectively). Furthermore, the plot of tunneling conductance, dI/dV, versus tip-sample bias voltage, V, revealed clear electronic states at low voltages in the Geobacter pili, never reaching zero conductance (Fig. 5C) but a large (~4 V) band gap in the PAK pili (Fig. 5F), which is characteristic of insulators. A band gap feature with zero differential conductance was also revealed in dI/dV curves of the Geobacter pili when probed at cryogenic (77 K) temperatures (Fig. 5H). As a control, we routinely checked the quality of the conductive tip by alternating dI/dV measurements on the pili with tunneling spectroscopy on the graphite substrate several nanometers away from the pili. The width and exact voltage location of the edges of the gap varied with the region of the pilus probed (± 0.1-0.2 V), likely due to variations in the density of states at different locations on the pilus, but the gap feature was consistently observed in all. This contrasts with the room temperature dI/dV of the pili acquired in the same voltage range, which lacks the gap feature (Fig. 5I). The energy gap at 77 K is on the order of hundreds of millivolts; as Boltzmann constant times temperature is only 25 mV at room temperature, simple thermal broadening cannot account for the absence of the gap at room temperature. Rather, the low temperature STM data is consistent with thermally induced redox conductivity.

Discussion
The results demonstrate that the pilus protein fiber transports charges, consistent with the predicted role of pili as protein nanowires between the cell and external electron acceptors such as Fe(III) oxides, 14 uranium 44 and matrix-associated c-cytochromes in electrochemically active biofilms 23 . Using the average potential (100 mV) that exists between the menaquinone electron carriers of the cell's inner membrane and the extracellular Fe(III) oxides that serve as electron acceptor 45 , we calculated average electron transport rates along a 1 μm-long pilus (~9 × 10 8 electrons per second) high enough to support the respiratory rates of the cell during the reduction of Fe(III) oxides (~9 × 10 6 electrons per cell per second; Fig. S4). The pili probed by CP-AFM were not chemically fixed and retained the topographic and electronic features (Fig. 5 and Fig. S7) of cell-associated pili 21 . However, the adsorption of the pili on the surface prevented any pilus dynamic motions that are predicted to promote electronic coupling and charge transport 23 . This suggests that the in vivo pilus charge transport rates could be within the orders of those measured by CP-AFM in vitro. As Fe(III) oxides are dispersed in the environment and difficult to access, cells maximize the chances of finding the electron acceptor and disposing respiratory electrons by assembling several pili per cell 14 .
The ability of pili to transport charges over micrometer distances challenges current understanding of protein-mediated electron transfer. The carrier mobility of the WT pilus (3.2 × 10 −2 cm 2 /Vs) extracted from the discontinuous I-V plot (Fig. 4), which gives a measure for how easily the carriers (electrons or holes) move through the pilus, is too low to fit a coherent band conduction mechanism 46 . Furthermore, MD simulations did not reveal aromatic π -stacking 23 , as proposed in the metallic model of pilus conductance 24 . Even with π -π stacking, carrier mobilities far greater than 1 cm 2 /Vs would have been required to describe pilus charge transport according to band theory 46 . The experimental and modeling data support instead an incoherent multistep hopping mechanism, whereby aromatic residues cluster and form the contacts needed to conduct charges along the pilus. Consistent with this model, removing the aromatic ring of Y27 (Y27A pilus), a key relay amino acid in the axial multistep hopping pathway 23 , reduced the aromatic contacts in approximately half (Fig. 3) and decreased the carrier mobility of the pilus (6.3 × 10 −3 cm 2 /Vs). The mutation also enlarged the regions of positive potential on the pilus surface (Fig. 3E). As the traps in the Y27A fill, they accumulate charges and scattering increases 40 . Scattering effects become more prominent with distance, as more trapped states accumulate, thereby increasing the electrical resistance exponentially as a function of distance (Fig. 1C). Charge trapping is further supported by the discontinuous plots of the Y27A pilus, which showed correlations I ~ V b above the crossover voltage, where the exponent b was greater than 2 (~2.5 in the representative plot shown in Fig. 4), as in materials with trapped states 47 . The presence of traps reduces the space-charge-limited current and modifies the shape of the I-V plot from an ideal quadratic to a higher power dependence of current on voltage 48 . By contrast, the WT pilus, which has small surface regions of positive potential neutralized to minimize charge trapping 23 , produced discontinuous plots with the pure quadratic scaling of current with voltage above the V C (Fig. 4) that is typical of SCLT 37 .
The Y27A mutation did not substantially affect the structure of the pilus (Fig. 3) or the number of conducting charges per volume unit (carrier concentration), which was ~10 20 cm −3 , as in the WT and within the orders (10 21 cm −3 ) reported for cell-associated pili after charge injection with a conductive tip and visualization of the propagated charge by electrostatic force microscopy 20 . High carrier concentrations in nanowires are the result, at least partially, of their reduced diameter 38 , which in the pili results from the replacement of the C-terminal globular head of the pilin with a short random coil 23 . Thus, the carrier characteristics of the pili appear to combine features of the organic nature of the conductive material (low mobility) and the nanowire dimensions (high carrier concentration). Yet despite the low carrier mobility, the high carrier concentration delivers a charge transport rate high enough to provide an efficient respiratory strategy.
The diameter of the Geobacter pilus estimated from the STM data (4-5 nm) matched well the diameter of the favored MD conformer of the WT pilus in solution (4.7 nm) 23 , which accounts for the volume occupied by the flexible C-t random coil. The detection of tunneling current in the C-t random coils of the pilus fiber is not unexpected because this is the region of the pilin where the terminal tyrosine of the transversal pathway (Y57) is located 23 . This is also the region where anionic ligands are located that bind positively charged electron acceptors, such as the uranyl cation, and position them close to Y57 to facilitate that last step in ET 23 . The helical assembly of the pili also positions the exposed C-t segments helically and confers on the pilus characteristic topographic periodicities, as detected by STM (Fig. S7). Consistent with its unique structural and electronic features, the Geobacter pili were transversally conductive in the ±1 V voltage range by STM, and produced differential tunneling conductance plots without a full energy gap at room temperature. By contrast, PAK pili controls were insulating in the ±2 V range (Fig. 5). Furthermore, the cryogenic STM measurements demonstrated the thermal activation of the differential transversal conductance of the Geobacter pili at the low (±100-200 mV) voltages that operate in biological systems 45 (Fig. 5), which is again consistent with incoherent redox conductivity.
The ability of a peptide assembly such as Geobacter pili to transport electrons at μm distances is especially significant for nanotechnological applications. Recombinant techniques can be applied to mass-produce pilin derivatives, which can be self-assembled in a cell-free environment using chemical methods 49 . Such approaches could be coupled to genetic engineering to produce protein nanowires with custom functionalities, including chemical tags for selective and specific integration into nanoelectronic devices. This contrasts with fabrication methods for inorganic nanowires, which involve high temperatures, toxic solvents, vacuums, and specialized equipment 50 and the limited methods available for their functionalization 51 . Protein nanowires also circumvent major concerns regarding cyto-and genotoxicity that limit commercial applications of carbon, metal, and metal-oxide based nanomaterials 52 , making them suitable for the development of biodegradable and biocompatible nanoelectronic devices. Of special significance is the natural ability of Geobacter pili to bind and reductively precipitate cationic metal contaminants such as uranium 44 , a process that requires specific amino acid ligands on the pilus surface 23 and could be harnessed to develop sensors and deployable devices for bioremediation.

Methods
Bacterial strains and culture conditions. G. sulfurreducens strain PCA and a Y27A mutant strain (constructed as described in the Supplementary Information) were used throughout the study. Cells were routinely grown anaerobically at 30 °C in NB medium 53 with 30 mM acetate (electron donor) and 40 mM fumarate (electron acceptor) (NBAF). When indicated, cells were grown at 30 °C in fresh water (FW) medium 44 with 30 mM acetate and 40 mM fumarate (FWAF), transferred to a fresh FWAF culture, and grown for 72 h at 25 °C to induce pili expression, as described elsewhere 14,44 . The headspace of the culture vessels was always N 2 :CO 2 (80:20).
Pili purification. WT and Y27A pili were isolated as reported elsewhere 44 , except that all the buffers used during the purification contained 1 mM ethylenediaminetetraacetic acid (EDTA) to remove metal impurities and all drying steps were carried out with a constant flow of filter-sterilized N 2 gas to prevent contamination. The purified pili were incubated in a 2:1 (v/v) chloroform-methanol solution for 2 h at 4 °C to extract any cell-derived quinone contaminants in the chloroform phase 54 and the methanol phase containing the pili was evaporated with a constant flow of filter-sterilized N 2 gas. The dry pili samples were stored at − 20 °C for short-term use or flash frozen in liquid nitrogen and stored at − 80 °C for long-term use. Pili were also purified from P. aeruginosa strain K (PAK), as described in the Supplementary Information. Protein concentration in pili samples was determined with the bichinchoninic acid (BCA) assay (Pierce ® , Thermo Scientific), as previously described 44 . Spectroscopy analyses. Three replicate pili samples (20 to 30 μg of pili protein) in 10 mM CHES buffer (pH 9.5, 1 mM EDTA) and buffer controls (no protein) were dispensed in acid-treated glass vessels and subjected to ICP-AES quantitative elemental analysis using a Thermo Jarrell-Ash Enviro 36 Inductively Coupled Argon Plasm (Chemical Analysis Laboratory, University of Georgia, Athens). The number of pilin subunits in each sample was computed for the pili protein values using the predicted pilin's mass of 6,568.51 Da (or 1.09 × 10 −14 μg). The concentration (in ppm) of elements was used to extrapolate the number of elemental atoms per pilin.
Purified pili were also suspended in a 1:1 (v/v) solution of 10 mM Tris buffer (pH 7) and isopropanol and their spectral properties were characterized by UV-VIS and fluorescence spectroscopy in reference to l-tyrosine (500 μM), riboflavin (100 μM) and menaquinone (100 μM) standard solutions in Tris:isopropanol (1:1, v/v). Absorption spectra were collected with a Cary100 UV-Vis spectrometer (Varian) set to 2 nm bandpass. Fluorescence spectra were acquired with a QuantaMaster spectrometer (Photon Technology International), with 270 nm excitation and 4 nm bandpass, and corrected with the fluorescence emission baseline from a Tris:isopropanol solution. All spectra were collected at room temperature, in quartz cuvettes with a 1 cm path length (Spectrocell Inc.).

Conductivity analyses with Scanning Probe Microscopy.
Pili were suspended in ddH 2 O and deposited on patterned gold electrodes nanofabricated onto a silicon chip for CP-AFM analysis, as described in the Supplementary Information. The AFM was first operated in tapping mode to identify pili lying across the gold edge. The CP-AFM tip was then positioned on regions of the pilus lying on the silicon substrate at various distances from the gold edge and current-voltage (I-V) curves were collected at each position (3 nN force, 1 Hz rate) while applying a bias voltage sweep (±1 V). I-V curves were also generated periodically on neighboring regions on the gold electrode and silicon substrate as positive and negative controls, respectively. Methods to calculate resistance values, transport rates, resistivity, conductivity, and carrier characteristics (mobility and concentration) of the pili from the I-V curves and cellular rates of Fe(III) oxide reduction for comparative analyses are described in the Supplementary Information.
Scanning Tunneling Microscopy (STM) was used to probe the topography and electronic properties of Geobacter and PAK pili after deposition on freshly cleaved HOPG. Information about sample deposition and STM imaging and spectroscopy at room and cryogenic (77 K) temperatures is available in the Supplementary Information. MD simulations. All the simulations were performed with the GROMACS software suite 55-57 , using the AMBER99SB force field parameterization 58 . The WT snaphots plot of number of contacts were extracted from the published pilus MD simulations 23 . The Y27A mutation was introduced into the geometry-optimized WT pilus and optimized in a 1-ns MD run, as previously described 23 . The concentration of Na and Cl ions used to neutralize the net charge of the Y27A fiber (− 36 e) was as in the WT (0.1 M). The atomic coordinates of the Y27A pilus model are provided as a supplemental file (GPIL-Y27A.pdb).
The VMD program 59 was used to generate maps of electrostatic potential and aromatic density. The latter averaged the atomic locations of the aromatic residues over the 70 ns MD trajectories. Three volumetric slices (xz plane; offset, 0.30, 0.50 and 0.70) of the aromatic density map were extracted for each pilus and overlaid using the Photoshop CS3 image analysis software to generate 3D-projections of the average pilus' aromatic density distribution. The number of aromatic contacts in the modeled WT and Y27A pili was estimated throughout the 70-ns MD simulations using the routine g_select program of the GROMACS package 60 . The criterion to regard two aromatic residues as a contact was based on the proximity of at least one pair of carbon atoms at a distance of 5 Å or less regardless of the configuration of the dimers.