High current production of Shewanella oneidensis with electrospun carbon nanofiber anodes is directly linked to biofilm formation

We study the current production of Shewanella oneidensis MR-1 with electrospun carbon fiber anode materials and analyze the effect of the anode morphology on the micro-, meso- and macroscale. The materials feature fiber diameters in the range between 108 nm and 623 nm, resulting in distinct macropore sizes and surface roughness factors. A maximum current density of (255 ± 71) μA cm-2 was obtained with 286 nm fiber diameter material. Additionally, micro- and mesoporosity were introduced by CO2 and steam activation which did not improve the current production significantly. The current production is directly linked to the biofilm dry mass under anaerobic and micro-aerobic conditions. Our findings suggest that either the surface area or the pore size related to the fiber diameter determines the attractiveness of the anodes as habitat and therefore biofilm formation and current production. Reanalysis of previous works supports these findings.


Shewanella oneidensis MR-1 and its applications in Bioelectrochemical Systems
Due to recent attempts to find alternatives for the fossil-based economy, Bioelectrochemical systems (BES) moved into the focus of scientific interest. 5 In BES, oxidative and reductive reactions are catalyzed by, e.g., electroactive bacteria. Electrodes are used to establish an electrical connection between bacteria cells and an external electrical circuit. The most prominent applications P R E P R I N T 1.3. Anode materials tailored for the operation with Shewanella oneidensis MR-1 Compared to Geobacter sulfurreducens, MR-1 produces lower current densities, but the current production can be improved by choosing a suitable anode material [20]. Numerous studies with mixed-community biofilms have been conducted 55 on anode materials [21,22,23,24], while little systematic work with MR-1 is reported in the literature. Kipf et al. [25] identified a commercial steamactivated knitted carbon fiber material with a current density of 24 µA cm −2 at a potential of −200 mV (sat. Calomel; 41 mVvs NHE ) as superior to other carbonbased materials. Patil et al. [26] studied the current production of electrospun 60 carbon fiber materials partly modified with activated carbon and graphite particles. Increased current production was observed with increasing specific BET surface area. Pötschke et al. [27] investigated the carbon composition and fiber breaks. They identified a combination of highly conductive graphitic carbon, and free filament ends as key to high current densities. In the present study, we systematically investigate the effect of the electrode the surface area, and the pore size on the anodic current production and biofilm formation of MR-1 under anaerobic and micro-aerobic conditions. The controlled fabrication process of the materials allows us to separate effects of the macro, meso-and micro-scale morphology on the current production. The surface chemistry is presumably 70 identical, allowing us to separate the effects of the material morphology and the surface chemistry on the current production.
We tested four materials with distinct fiber diameters that feature surface roughness factors (SRF s) between 286 and 1004 and mean macropore diameters between 0.4 µm and 1.6 µm. Additionally, the electrospun carbon fiber surface 75 was modified by steam and CO 2 activation to increase the surface area further and introduce micro-and mesopores, which have been hypothesized to enhance MET [25,26]. We also reevaluate previous systematic anode material studies [25,26,27] based on the premise that mainly the macrostructure of the anode determines the current production and biofilm formation of MR-1. The commercial reference 80 materials were investigated previously [28,5,25,20] and serve as a benchmark in this work.

Preparation of tailored electrospun and commercial reference anodes
Four electrospun carbon fiber mat materials with average fiber diameters of 85 ∼108 nm (ES100), ∼286 nm (ES300), ∼400 nm (ES400), and ∼623 nm (ES600) were fabricated by electrospinning of PAN with concentrations between 6 wt% and 13 wt% in N,N-dimethylacetamide and subsequent carbonization at 1100°C as previously reported [29]. Activation was conducted for 4 h with CO 2 at 832°C and with steam at 757°C. Two commercial carbon materials, knitted steam 90 activated carbon C-Tex 13 (Mast Carbon International Ltd., United Kingdom) and graphite felt GFD 2 (SGL Carbon SE, Germany), were used as reference materials [28,25,20] for comparison. SEM photographs of these materials are provided in Fig. S1.1. The anodes were cut into pieces of 2 cm 2 × 2 cm 2 and P R E P R I N T mounted in polycarbonate frames exposing 2.25 cm 2 to the growth medium ( Fig.  S1.2). To remove gas bubbles from the porous anode structures, the anode material was wetted with isopropyl alcohol, rinsed with DI water, and placed in vacuum for approximately 10 minutes before the experiments.

Anode material characterization
The electrochemically active surface (ECAS ) of the anodes was determined 100 via the measurement of the double-layer capacitance using a Solartron Analytical potentiostat (1470E, Farnborough, United Kingdom) as reported previously [29]. Nitrogen adsorption was conducted at 77 K using a Belsorp-Max instrument (BEL Japan Inc., Japan) for the electrospun materials and a Sorptomatic 1990 (CE Instruments, Italy) for the commercial reference materials. Scanning electron 105 microscopy (SEM) was performed using a DSM 962 (Carl Zeiss Microscopy GmbH, Germany) and an SX 100 (Cameca, France). The macropore sizes were estimated from SEM photographs using the software ImageJ and DiameterJ [30] as described in Section S4 in the supplementary information. The porosity ϕ was calculated based on the anode thickness in the mounting frames t according Herein, ρ A is the material's area density, and ρ is the density of PAN-derived carbon (1.75 g cm −3 , [31]). The internal surface area was assessed by the ECAS method [29]. From the weight-specific ECAS and S BET surface area, and the area density ρ A , the Surface Roughness Factor (SRF ) was calculated according to SRF ECAS = ECAS · ρ A and SRF BET = S BET · ρ A . The 115 SRF, defined by the ratio of the internal surface area and the anode's projected surface area, allows the direct comparison of the materials with different area densities. Similarly, the absolute micro-and mesopore volume V Micro and V Meso , determined from N 2 adsorption isotherms (t-plot and BJH analysis respectively), was calculated from the weight-specific value ν for the projected anode area: 120 V = ν · ρ A · 2.25 cm 2 . The macropores of the electrospun materials and GFD 2 were assessed by digital image processing and 3D modeling as described in Section S4 in the supplementary information. With this method, a 2D SEM image is analyzed for its apparent porosity and pore diameter. In a second step, the 2D apparent pore diameter is converted into a 3D pore diameter by 125 comparison with a 3D model of the fiber material morphology. This unique method was chosen because mercury intrusion porosimetry underestimates the pore sizes of compressible materials [32,33].

Bioelectrochemical characterization
MR-1 cells were prepared following the procedure described by Kipf et al. 130 [25]. In short, MR-1 cells were first cultivated aerobically in Lysogeny broth (LB medium) and then transferred to a growth medium containing 50 mM Na-D/L-lactate and 100 mM fumarate. The cells were washed with washing buffer (growth medium lacking lactate and fumarate) thrice before inoculation of the bioelectrochemical reactor ( Fig. S1.3). The amount of inoculum was 135 adjusted to an initial optical density (OD 600 ) of ∼0.05. Before inoculation, the bioelectrochemical reactor was sterilized by steam autoclavation at 121°C for P R E P R I N T 20 min. The bioelectrochemical reactor contained 1 L anode medium with 50 mM initial concentration of Na-D/L-lactate as an electron donor and no electron acceptor other than 6 anodes. The anodes were polarized at −41 mV against a sat.

140
Calomel reference (SCE, KE11, Sensortechnik Meinsberg, Germany) using PGU-MOD 500 mA potentiostats (IPS Elektroniklabor GmbH & Co. KG, Münster, Germany). To minimize non-coulombic currents, the anodes were polarized at least 12 h before the inoculation. A platinum mesh acted as a counter electrode. The current production was recorded for 14 days. In the present publication, 145 potentials are additionally provided vs. Normal Hydrogen Electrode (NHE) at 25°C for better comparison. This way, different temperature dependencies of the redox couples are accounted for. The reactor is operated at 30°C and its headspace is continuously purged with humidified N 2 at 125 mL min −1 for anaerobic conditions (<20 ppm). Trace amounts of O 2 (200 ppm -400 ppm 150 corresponding to 1.4 µg L −1 -2.8 µg L −1 dissolved oxygen) were introduced into the reactor by guiding the gas flow through an O 2 permeable silicone tube. Oxygen concentrations were determined with a Fibox 3 LCD trace oxygen meter (PreSens Precision Sensing GmbH, Germany). The medium was stirred at 300 rpm and sampled every other day to monitor pH, cell density, lactate 155 consumption, and flavin concentration.

Medium monitoring
Flavin concentration in the medium was measured with a Nanodrop 3300 spectrometer (emission at 525 nm, excitation 470 nm) using riboflavin (RF) as a standard. RF exhibits similar fluorescence signals as FMN and FAD [34]. 160 Therefore, the flavin concentrations reported comprise FMN, FAD, and RF. The CE was calculated based on the lactate consumption under microaerobic conditions and based on the acetate accumulation under anaerobic conditions assuming partial oxidation of lactate to acetate with 4 electrons per molecule [35]. The initial and final lactate and acetate concentrations were determined 165 by high-performance liquid chromatography (Elite La Chrome, Hitachi, Japan) equipped with an Aminex HPX-87H column (Bio-Rad, Germany) and RI-detector (RefraktoMax 521, IDEX Health & Science, Japan). The planktonic dry weight m in the medium was determined by {m} mg = 716 · {V } Liter · {OD 600 }. The conversion factor was determined experimentally with the dry weight of a cell 170 filter cake from medium with known volume and OD 600 .

Biofilm analysis
The biofilm dry weight was quantified after 14 days of operation as follows. The anodes were placed in 1 mL of lysis buffer (20 mM Tris/HCl, 16.5 mM Triton X-100, 137 mM NaCl, protease inhibitor cocktail for bacteria 25 µL mL −1 , Carl Roth, Germany) at 4°C overnight. The lysate's protein content was then analyzed using a biuret test according to the instructions (Roti-quant universal, Carl Roth, Germany). Dry mass was calculated by comparison to cell lysate with known dry weight. For SEM imaging, the biofilm was fixated with 2.5 % glutaraldehyde in anode medium lacking lactate for 24 h at 4°C. The samples were then dehydrated P R E P R I N T with a series of 10 %, 30 %, 50 %, 70 %, 80 %, 90 %, and 100 % ethanol in washing buffer for 10 min each and air-dried at room temperature.

Anode material properties
The pore sizes and surface area of the anode materials were characterized 185 on the micro (<2 nm), meso (>2 nm to <50 nm), and macro (>50 nm) scale by nitrogen adsorption, electrochemically accessible surface (ECAS ), and SEM imaging. Scanning electron microscopy photographs of the materials are provided in Fig. S1.1. The self-made electrospun materials ES100 to ES600 feature ultrathin fibers with mean fiber diameters between ∼100 nm and ∼600 nm. The

Material
Diameter dF The steam activated C-Tex 13 exhibits the highest SRF ECAS , SRF BET , and micropore volume V Micro among the investigated materials. Interestingly, the graphitic GFD 2 also features a substantial micropore volume but a lower SRF ECAS of 27. Graphite contains lesser oxygen-containing groups [36] that 205 increase the double layer capacitance [37] and thus the ECAS. The high mesopore volumes V Meso of 3.39 × 10 −3 cm 3 (C-Tex 13) and 1.95 × 10 −3 cm 3 (GFD 2) are mainly due to the higher area densities of 182 g m −2 and 165 g m −2 . The mean P R E P R I N T macropore diameter d Macro of GFD 2, 31 µm, is about a factor of 20 higher than the electrospun materials with similar morphology. The macropore structure of 210 C-Tex 13 can only be assessed semi-quantitatively due to its knitted structure: The grooves along the fibers introduce porosity in the submicron range, and interfiber distances vary from direct contact to several fiber diameters within one fiber bundle. Maximum distances between the fiber bundles are in the range of ∼200 µm. The electrospun materials' ECAS values do not monotonically increase with decreasing fiber diameter, possibly due to the fiber cross-section's deviation from a perfectly round shape [29] or differing surface roughness. The same holds for the SRF ECAS due to similar area densities. However, S BET and SRF BET increase monotonically with decreasing fiber diameter. CO 2 and steam activation increase the SRF ECAS of ES300 by 30 and 55 and SRF BET by a 220 factor of 12 and 24, respectively. The increased micro-and mesopore volume of the activated electrospun materials show that the increased SRF s are linked to micro-and mesopore formation during activation. Because of the low weight loss during activation, the fiber diameter and macropores remain practically unaffected by CO 2 and steam activation.

Electrical resistivity of the anode materials
The sheet resistances for the electrospun materials given in Erben et al. [29] allow the estimation of the ohmic potential drop across the electrode according to Madjarov et al. [38]. For the electrospun materials and GFD 2, with sheet resistances between ∼1.8 Ω and ∼2 Ω, the ohmic drop for the current densities of 230 MR-1 obtained in this work is estimated to be less than 1 mV. C-Tex 13 has a higher sheet resistance of 144 Ω, resulting in an ohmic drop of less than 40 mV at the maximum current densities measured with the materials. Therefore, we do not expect that the different electrical resistances affect the current production. A trend to higher current densities with decreasing fiber diameter is apparent under anaerobic conditions. Welch corrected t-tests did not reveal a significant improvement of the current production of activated ES300 materials ES300-CO 2 (p = .14, Welch corrected t-test) and ES300-H 2 O (p = .15) compared to the nonactivated ES300 material (Fig. S2.5). Under micro-aerobic conditions, similar current productions of the electrospun materials and C-Tex 13 are observed. The current production of GFD 2 is increased 13-fold under micro-aerobic conditions compared to anaerobic conditions. B) The maximum current density normalized to ECAS reveals that the current production is not directly linked to the internal surface area.

P R E P R I N T
The six different anode materials were characterized in four technical replicates to account for the experimental conditions' variability. The values reported in this section are averages with sample standard deviation (n = 4). As one anode material of a kind was used for biofilm imaging by SEM, three values were considered in the biofilm dry weight analysis in Section 3.4. The maximum current densities presented in Fig. 1 A were extracted from chronoamperometry data recorded over the experimental period of 2 weeks (data not shown). Under anaerobic conditions (Fig. 1A, red), a maximum current density of (255 ± 71) µA cm −2 was recorded with the ES300 material. This current density is about 1.4-fold higher than the highest value of 186 µA cm −2 reported up to now [27]. A trend towards higher current densities with decreasing fiber diameter is apparent (Fig. 1 A, red). Through the fiber diameter d F and the porosity ϕ, SRF and macropore diameter d Macro are inseparably linked in fibrous materials on the macroscale (a detailed derivation of the analytical model can be found in Section S4.1 of the supplementary information): As predicted by the analytical model, SRF ECAS , SRF BET , and d Macro intercorrelate strongly for the non-activated electrospun materials (absolute correlation coefficients between 0.82 and 0.96, Tab. S3.2). Additionally, we find that the mesopores volume V Meso is correlated to the SRF and d Macro (absolute correlation coefficients between 0.79 and 0.96). As expected, the maximum current 240 density can be correlated reasonably well with all of the above material properties (Fig. S2.4). Therefore, the increased current production of thin fiber material cannot be attributed to the increased SRF, reduced macropore diameter, or the increasing mesopore volume. Steam and CO 2 activation introduce microand mesopores, while the macropore structure remains unchanged. The surface 245 roughness factor increases due to the surface within the micro-and mesopores, but the additional surface does not significantly improve current production ( Fig.  S2.5). This allows us to rule out micro-and mesopore's positive effects on the current production and the related pore surface. We will further discuss the role of macropore size and SRF for the current production again in Section 3.7. Under 250 micro-aerobic conditions ( Fig. 1 A,  the current production can be observed. Although oxygen serves as a competing electron acceptor to the anode, it enhances current production [25] and bacterial growth and biofilm formation (see Section 3.4). Little systematic research on the influence of the anode macro morphology on the current production has been conducted up to now. Diverse characterization techniques make it impossible 260 to compare results directly between different studies. The current production of the commercial reference materials C-Tex 13 and GFD 2 was characterized previously using a quasi-galvanostatic protocol by Kipf et al. [25]. Analysis

P R E P R I N T
of the experimental setup in retrospect revealed that trace amounts of oxygen were present in the reactor. Therefore, micro-aerobic conditions in this work 265 are comparable to the anaerobic conditions in Kipf et al. [25]. Interestingly, the commercial reference materials C-Tex 13 and GFD 2 exhibit about 10-fold higher current densities, with (229 ± 82) µA cm −2 and (112 ± 36) µA cm −2 respectively, than previously reported [25]. The 10-fold higher current production is most likely a result of the different characterization methods. In the present 270 study, even e.g. trace amounts of oxygen change the difference in the reference materials' current production from 9-fold under micro-aerobic to 2-fold under micro-aerobic conditions. This underlines the necessity of a systematic and well-controlled approach to material characterization. In the following, we reevaluate data from previous publications based on the premise that the anode 275 materials' macrostructure determines its current production with MR-1 taking into account the possible effects of oxygen contamination. Tab. S5.3 summarizes the data of Kipf et al. [25] and includes additional morphological characteristics. We find that knitted, and woven activated carbon fabrics' current production is significantly higher than the carbon felts and carbon paper (p ¡ .001, Welch 280 corrected t-test). The current production of felt materials, made of grooved activated carbon fibers (V1-Felt) and smooth graphitized fibers (GFD 2), does not differ significantly (p = .13). This confirms our finding that micro-and mesopores do not enhance the current production considerably. Within the group of knitted and woven fabric anode materials, C-Tex 27 produces a significantly 285 lower current than C-Tex 13 (p = .0018). Judging from the SEM images, the densely aligned fibers of C-Tex 27 and woven structure cause less accessible pore space in and between the fiber bundles compared to C-Tex 13, consisting of knitted twisted fiber bundles. C-Tex 20 features a higher area density and thickness than C-Tex 13, while the fiber diameter is the same. The higher 290 current production of C-Tex 20 compared to C-Tex 13 (p = .069) can therefore be attributed to the increased area-specific volume due to the higher thickness of the electrode material, similar to four layers of C-Tex 13 compared to a single layer [25]. It is noteworthy that the current production does not scale linearly with the number of layers, possibly to mass transport limitations. Patil et al.

295
[26] investigated the current production of electrospun PAN-based carbon fibers without (PAN) and with additional activated carbon (PAN-AC) and graphite (PAN-GR) particles (SEM images and material properties see Tab. S5.4). The higher current production of the materials with particles was attributed to their higher surface area. Besides higher BET surface area and micropore volume of PAN-AC and PAN-GR, the materials with particles and a higher surface area also exhibited smaller fiber diameters. Similar to our results, the increasing current production with decreasing fiber diameter could result from the macrostructure with a larger accessible surface area and smaller macropores. In a study by Pötschke et al. [27], the effect of the chemical carbon fiber composition and fiber  The bacterial dry weight was quantified after two weeks of operation (Fig.  2 A). Under micro-aerobic conditions, a higher bacterial dry mass was found for all anode materials compared to anaerobic conditions. Under micro-aerobic 320 conditions, similar amounts of bacterial dry mass with large variability between 38 % and 73 % were found on the electrospun materials and the commercial reference materials. Under anaerobic conditions, however, increasing bacterial dry mass with decreasing fiber diameter can be observed for the electrospun materials (Fig. 2 A, plain red). The similarity to the maximum current densities 325 in Fig. 1 A is apparent. Therefore, the final current densities (after two weeks of operation) were normalized to the bacterial dry weight (Fig. 2 B). Most interestingly, the electrospun materials show similar values as would be expected P R E P R I N T for a proportional relationship between the number of current producing cells and current production for anaerobic and micro-aerobic conditions. The values of C-Tex 13 under both operating conditions, and GFD 2 under micro-aerobic conditions, do not deviate significantly from the average value of (65.2 ± 7.0) µA mg −1 for the electrospun materials. (65.2 ± 7.0) µA mg −1 corresponds to about 47 fA per cell (calculated using a dry weight of 716 pg/cell determined with the dry weight of an MR-1 filter cake from an anaerobic culture with known cell den-335 sity). This value is well in line with values reported by Lu et al. [39] for MR-1 single-cell current production of 165 fA under aerobic conditions to 36 fA at 0.42 mg L −1 dissolved oxygen. In this study, the anode was poised at ∼200 mV (sat. Ag/AgCl, 400 mV vs. NHE), and the cells were quantified by confocal laser scanning microscopy. With real single-cell probing using optical tweezers, 340 a current of 50 fA at the same potential as this work (200 mV vs. NHE) was determined [40]. The lower value of GFD 2 could be explained by interstitial cells that populate the pore space [41] and do not contribute to the current production due to lacking electrical contact to the anode. This effect would be less pronounced under micro-aerobic conditions. Interestingly, the current production of single Geobacter sulfurreducens cells of (93 ± 33) fA [42] is in the same range as single MR-1 cells.

Morphology of the MR-1 biofilms
SEM images of the biofilm grown under anaerobic conditions on the electrospun anode materials (Fig. 3) reveal biofilm coverage independent of the 350 fiber diameter at the anode front and a trend towards higher biofilm coverage with decreasing fiber diameter at the anode back. The biofilms reaching further into the electrospun anode material and increased bacterial dry weight suggests that thin fiber materials provide a more attractive habitat for MR-1 cells than thick fiber materials. However, under micro-aerobic conditions, a different trend 355 is visible: Higher biofilm coverage on the anode front compared to the anode back is observed. The likely explanation for this observation is oxygen tension towards the anode front due to the abundance of trace amounts of oxygen in the medium [43]. It is important to note that the SEM images do not show the biofilm structure in operation. During sample preparation, the drying process 360 leads to a collapse of the porous biofilm onto the anode surface and shrinkage of the cells themselves.  P R E P R I N T 3.7. The role of pore size and surface for the current generation and biofilm formation As described above, current production is directly linked to the bacterial dry weight attached to the anode. Under anaerobic conditions, the dry weight increases with decreasing fiber diameter of the non-activated electrospun materials. Steam and CO 2 activation, which introduce micro-and mesopores and increase the surface area, does not further increase biofilm attachment. Therefore, we con-405 clude that the anode materials' macrostructure determines bacterial attachment to the anode and thus current production. However, as established in Section 3.3 SRF, porosity, macropore diameter, and fiber diameter are inseparably linked in fibrous materials on the macroscale. Thus, the higher current production of the electrospun materials with decreasing fiber diameter, which is directly 410 related to the biomass attached to the anode, cannot be attributed directly to either the increased SRF or the reduced macropore size. The literature does not provide an in-depth analysis of macropore size and the related surface regarding current production and biofilm formation. In the following, we discuss microscopic processes within biofilms and electrodes that allow attributing the 415 high current production of macroporous electrospun materials to either pore size or surface. Electrospun fiber materials provide high SRF s for electron discharge and mean pore diameters in the range of 0.4 µm to 1.6 µm (Tab. 1) allow the bacteria to penetrate the anode pore space and attach to the internal surface. Chong et al. [47] concluded that pore sizes in the order of a millimeter are 420 preferential for the current production with bacteria that form dense biofilms such as the Geobacter spp. and sludge-derived mixed communities. Smaller pores lead to (partial) pore-clogging that hinders the mass transport of nutrients and metabolic products and eventually limits the achievable current densities. MR-1 forms loose biofilms that do not block the pores and do not cover the 425 entire fiber surface (Fig. 3). Hence, the surface area is unlikely to limit bacterial attachment and current production. This raises the question of whether the confinement of cells in proximity to the anode surface or a more attractive chemical microenvironment within the anode pore space leads to increased biofilm formation with decreasing fiber diameter. A recent study by Pirbadian et al. [16] revealed that the cell membrane hyperpolarization, a measure for metabolic activity, decreases rapidly with the distance from the anode under anaerobic conditions. The membrane hyperpolarization drops within (8.9 ± 12.4) µm to 1/e of the value in contact with the electrode. Upon addition of 5 µM riboflavin, this value increases to (21.8 ± 4.3) µm. Biofilms that were grown on flat graphite 435 reach a similar thickness [14]. Therefore, pores, larger than a few µm, will be occupied by the biofilm with partially reduced average metabolic activity and current production per cell (panel A in Fig. 6). The low current per dry weight values for GFD 2 under anaerobic conditions (little flavin accumulation) and the increase under micro-aerobic conditions (flavin accumulation in the range of 440 1 µM) can thus be attributed to the large pores of GFD 2. Three mechanisms could explain the decreasing metabolic activity with increasing distance from the anode surface: consecutive redox-cycling of adjacent cells [17], MET through self-secreted flavins [7,9], and a dynamic biofilm structure caused by motile P R E P R I N T cells that increase the apparent surface coverage with cells, enabling more cells to participate in EET [17]. Therefore, we cannot attribute the high current densities of electrospun materials to either of the above EET mechanisms. The mean pore diameters of the electrospun materials of 0.4 µm to 1.6 µm are in the same range as the MR-1 cell size (2 µm to 3 µm in length and 0.4 µm to 0.7 µm in diameter, [48]). Therefore, it is not surprising that the current per dry 450 weight for all electrospun materials is similar, considering that the cells must be in close contact with the anode surface in small pores. The average value of 47 fA per cell, we determined in Section 3.4, may represent the metabolic limit of MR-1 cells under this study's experimental conditions. The total cell volume in the anode is about 8.5 mm 3 (estimated from the volume of a single 455 cell 0.7 µm 3 ), the maximum current density (255 µA cm −2 , and 47 fA). This is about 8 % of the pore space of the electrospun materials. Therefore, the available pore space is not limiting biofilm formation. A possible explanation for the low cell density within the pore space could be a self-inhibiting effect caused by gradients of metabolic products: i.e., protons and the related local acidification 460 inside the anode (panel B in Fig. 6). Presumably, redox-cycling of flavins occurs between the anode and the MR-1 cells (panel C in Fig. 6). Increasing numbers of MR-1 cells inside the anode shift the equilibrium of oxidized and reduced flavins towards the reduced form. Therefore, the positive chemotaxis of MR-1 cells to oxidized riboflavin [43] is decreased by the presence of MR-1 cells. In turn, 465 both smaller pore diameter and larger surface area increase the regeneration of oxidized flavins. Thus, a gradient of oxidized flavins can be sustained with larger numbers of cells within the pore space. Small pore diameter (shorter diffusion distances) and the surface (more reaction sites) would enhance this mechanism. Okamoto et al. [44] proposed a different role of pores concerning 470 flavins: dead-end pores fully covered by cells could lead to a local increase of flavin concentration in the dead-end pores, enhancing the MET (panel D in Fig. 6). Also, small pores would decrease diffusion distances. This proposed effect requires dead-end pores with sizes within close boundaries: The lower limit is set by the molecular size of flavins (1.3 nm 2 to 1.5 nm 2 , [49]), and the 475 cell footprint gives the upper limit in the order of 1 µm 2 . We cannot exclude this possibility, as the present study's activated materials mainly feature micropores and open macropores. In conclusion, we can categorize the above processes in surface-related processes (cellular attachment, DET, and oxidation of flavins) and transport processes (flavins, electrons in the biofilm, and bacteria). As MR-1 480 biofilms do not clog pores, transport processes can be expected to be accelerated by small macropores. Materials with small macropores also provide a larger surface close to the surrounding bulk medium and allow steeper gradients that enhance mass transport of nutrients, oxygen, and metabolic products.

Implications for further research 485
Our material ES300 with open 0.78 µm macropores exhibits a maximum current density of (255 ± 71) µA cm −2 , which corresponds to a 1.4-fold increase compared to the previous literature [27]. In addition, this extraordinarily high current density is accompanied by a coulombic efficiency (CE) of (96. 6 Figure 6: Illustration of microscopic processes in the biofilm and their dependence on pore size, surface area, and the number of cells inside the anode. A) Metabolic activity as a function of the distance to the closest anode surface (free adaption of Pirbadian et al. [16]. B) Acidification in the anode as a function of distance from the bulk medium and cell density. C) Flavin oxidation state as a function of pore size and cell density inside the anode. D) Flavin accumulation in dead-end pores, according to Okamoto et al. [44].
that exceeds the previously reported values considerably (see Section 3.6). Both, 490 high CE and current density are required for commercial applications with a high product yield and production rate. As shown by our experimental results, the high current production is linearly related to the bacterial dry weight found on the anodes. The electrode surface is only partially covered with cells, suggesting that the surface area does not directly limit cell attachment. Also, the pore 495 space of the electrospun material is only filled to about 8 % with cells. The underlying mechanisms, that determine the attractiveness of the anodic habitat for MR-1 are not fully understood, and require further research. Further efforts to improve current densities with MR-1 should therefore focus on enhanced biofilm formation. Possible strategies include the chemical modification of the 500 carbon surface, tailoring of pore size and shape. We discussed processes that affect the chemical microenvironment inside the anode material and might cause the improved biofilm formation with decreasing fiber diameter: • Cellular attachment, DET, and oxidation of flavins are enhanced by large anode surface areas.

505
• Mediator transport by diffusion and electron transport in the biofilm profit from small macropores.
• High porosity and interconnected pores are required for the transport of nutrient and metabolic products between the bulk medium and the anode's interior.

510
Besides the material development, the chemical microenvironment inside the anode could be improved through media optimization, e.g. increase in buffer capacity. Special care needs to be taken during material characterization to ensure anaerobic conditions. Even trace amounts of oxygen lower the CE P R E P R I N T considerably and (partially) mask the macrostructure's effect on the current 515 production. Planktonic cell growth and flavin accumulation are indicators for oxygen contamination.

Conclusions
The current production of Shewanella oneidensis MR-1 with tailored electrospun anode materials with fiber diameters between 108 nm to 623 nm was characterized. Our highest current production of (255 ± 71) µA cm −2 with the 286 nm fiber diameter material exceeds the highest value reported in the literature 1.4-fold. The high current density is accompanied by a coulombic efficiency of (96.6 ± 1.8) %. Additional activation did not improve the current production significantly. We observe a linear relationship between current production and 525 bacterial dry weight and increased penetration depth of the biofilm. We conclude that the anode material's attractiveness for MR-1 cells is most likely determined by both the macropore size and the accessible surface area.

Acknowledgements
We are grateful for the financial support from the German Ministry of 530 Education and Research (BMBF) under the program 03SF0496A. We thank Dr. Guillaume Alexis Adrien Pillot for fruitful discussions and critical feedback during the writing process.