High-power hybrid biofuel cells using layer-by-layer assembled glucose oxidase-coated metallic cotton fibers

Electrical communication between an enzyme and an electrode is one of the most important factors in determining the performance of biofuel cells. Here, we introduce a glucose oxidase-coated metallic cotton fiber-based hybrid biofuel cell with efficient electrical communication between the anodic enzyme and the conductive support. Gold nanoparticles are layer-by-layer assembled with small organic linkers onto cotton fibers to form metallic cotton fibers with extremely high conductivity (>2.1×104 S cm−1), and are used as an enzyme-free cathode as well as a conductive support for the enzymatic anode. For preparation of the anode, the glucose oxidase is sequentially layer-by-layer-assembled with the same linkers onto the metallic cotton fibers. The resulting biofuel cells exhibit a remarkable power density of 3.7 mW cm−2, significantly outperforming conventional biofuel cells. Our strategy to promote charge transfer through electrodes can provide an important tool to improve the performance of biofuel cells.


Response 2:
We believe that the reviewer confused the work of another group with our group's previous works. We should mention that our previous research has nothing to do with the paper reported by Prof. Yongchai Kwon and colleagues [J. Power Source, 286, 197−203 (2015)]. However, for resolving the concerns raised by reviewer #1, we explain the difference between Prof. Yongchai Kwon's results and our current study from the viewpoint of device structure and interfacial chemistry.
In the paper reported by Prof. Yongchai Kwon, they described that the (GOx/PEI) n multilayers were LbL-assembled onto the 20 nm-thick bare multiwall carbon nanotube (MWCNT) without any chemical treatment, and additionally, these (GOx/PEI) n multilayer-coated MWCNTs and Nafion were drop-casted onto glass carbon electrodes (GCEs) for electrochemical measurements. That is, the biofuel cells reported by Prof. Kwon and colleagues were based on flat film electrode [reference to redacted figure] instead of implantable thread or fiber-type electrodes shown in our study (see Figure R1). Furthermore, they did not provide any information about the total thickness of the solution-casted (GOx/PEI) 2 multilayer-coated MWCNT films although they reported that the thickness of (GOx/PEI) 2 multilayers was ~2.5 nm. Additionally, from the perspective of the LbL assembly, it is unreasonable that the polyelectrolyte multilayers composed of bulky anionic GOx and cationic PEI [molecular weight (M w ) of PEI used in their study ~ 750,000] are uniformly coated onto 20 nm-thick bare MWCNTs without any chemical treatment. Therefore, we have much difficulty in accepting their results that the best power performance of biofuel cell was obtained from (GOx/PEI) 2 -MWCNT-based electrodes, and additionally, the thickness of (GOx/PEI) 2 multilayer was approximately 2.5 nm. The detailed explanations about our concerns are as follows: When the LbL assembly of the (GOx/PEI) n multilayers was performed onto uncharged bare MWCNT, many processing steps (i.e., the deposition of polymer, the sedimentation of polymer-coated MWCNTs, the removal process of supernatant solution, the washing processes for removal of weakly adsorbed polymers, and then the ultrasonication for dispersion of polymer-coated MWCNTs) were repeatedly required for the coating of polymeric multilayers such as (GOx/PEI) n multilayers. However, these 20 nm-thick MWCNTs are seriously agglomerated and sedimented even after the first deposition of bulky PEI layer with high M w because of polymeric bridging between neighboring MWCNTs. It should be noted that in the case of small colloids with a diameter below 100 nm (i.e., anionic polystyrene or silica colloids), electrostatic LbL assembly using polyelectrolyte linkers is very difficult. Additionally, in the case of using the uncharged bare MWCNTs instead of anionic MWCNT with the COO − functional groups, these phenomena are generally accelerated during repeated LbL 4 assembly process [these aggregation phenomena are well explained in the previous paper reported by Frank Caruso, Macromolecules, 35, 9780−9787 (2002)]. Therefore, it is reasonable to analogize that the periodic zeta-potential changes obtained from the alternating deposition of PEI and GOx onto MWCNTs in their paper are mainly induced by the presence of excess PEI and GOx onto MWCNTs.
To clarify this issue, we have prepared the GOx-based multilayer using PEI [i.e., (GOx/PEI) 30 ] with M w of ~750,000 used in Yongchai Kwon's group [J. Power Source, 286, 197, (2015)] under the same preparation condition, and compared to the (GOx/TREN) 30 multilayers in our system (see Figure R2). As a result, the (PEI/GOx) 30 multilayers displayed the film thickness of ~318 ± 7 nm, while the (GOx/TREN) 30 multilayers showed the thickness of ~13 ± 5 nm, which were measured by crosssectional FE-SEM images. In other words, the average thickness of the enzyme layer (i.e., GOx/each linker) deposited per 1 bilayer can be estimated to be significantly different values of ~10.6 nm for GOx/PEI and ~0.43 nm for GOx/TREN layer. Although we used the Au-coated Si wafer as a substrate to deposit the enzyme layer, the obtained results clearly indicate that the polymeric aggregation phenomenon can be accelerated when using the bulky PEI linker instead of the small-molecule TREN. Figure R2. Cross-sectional FE-SEM images for film thickness of a, (GOx/PEI) 30 and b, (GOx/TREN) 30 multilayers deposited onto the Au-coated Si wafer. Furthermore, the respective PEI layers sandwiched between adjacent GOx layers can seriously limit the electron transport and tunneling from GOx to electrode despite the use of low bilayer number (i.e., 2 or 3 bilayered GOx/PEI) because of their bulky size (M w of PEI ~ 750,000) and insulating characteristic.
On the other hand, we highlight that the bilayer-controlled GOx/TREN multilayers shown in our study are very similar to the thickness-controlled single GOx layer due to the use of TREN (M w ~ 146) as a small organic linker. This possibility can be confirmed by our experimental results on the linkerdependent current density performance for (GOx/linker) m , and conductive Au-based layer (Au NP/ linker) n , (Figure 4b, c and Supplementary Figure 20 in the revised manuscript and Supplementary Information). In this case, PEI-based electrodes showed poor electrochemical performance compared to TREN-based electrodes in both the anode and cathode. These results strongly suggest that the use of bulky polymer linker such as PEI for the preparation of GOx-based multilayers makes a fatal effect on the electron transport between the GOx multilayer and the metal electrode.
Response 3: We appreciate reviewer #1's comment. As the reviewer pointed out, the thickness of GOx/TREN multilayers is a critical factor for electron transfer and catalytic performance because the electron communication between GOx and electrode is based on electron tunneling mechanism. Particularly, in the case of LbL assembly using bulky GOx, the multilayer thickness can exceed the tunneling thickness. Therefore, controlling the adsorption amount and thickness of GOx is most important with the small organic linkers (i.e., TREN) for efficient electron transfer.
To this end, we delicately controlled the thickness (or loading amount) of GOx/TREN multilayers using solution concentration-controlled electrostatic LbL assembly of GOx (5 mg ml −1 ) and TREN (1 mg ml −1 ) solution at human-body compatible pH 7.4. First, it should be noted that the pK a (pH with a degree of ionization of 50 %) of carboxylic acid groups within GOx is approximately 4.5, and therefore GOx at pH 7.4 is highly negatively charged. On the other hand, in the case of TREN with pK a ~ 10, TREN at pH 7.4 is highly positively charged. These highly charged GOx and TREN generate the extremely low loading amount per layer because of long-range electrostatic repulsion between the same charged GOx as well as the same charged TREN. In the case of measuring the film thickness using Au-coated Si substrates similar to Au NP-coated cotton fibers (i.e., metallic cotton fiber, MCF), the thicknesses of (GOx/TREN) 20 and (GOx/TREN) 30 multilayers were measured to be ~9 ± 3 nm and ~13 ± 5 nm, respectively despite deposition of 20 or 30 bilayers (see Figure R3). Considering the dimension of GOx (6.0 x 5.2 x 7.7 nm 3 ) [Sensors, 8, 5637-5648 (2008).], these thicknesses of the multilayers indicate that GOx enzymes are adsorbed on the substrates as a single layer for 20 bilayer-film and as a bi-layer of 30 bilayerfilm. Thus, the progressive increase of the anodic current up to 20 bilayers can be attributed to the gradual adsorption of GOx enzymes as the single layer form on the substrate (Figure 3c). On the other hand, the minor increase of the anodic current from 20 to 30 bilayers may be due to the adsorption of the GOx enzymes as the double-layer form that can interfere the electron communication between the GOx and the Au NPs (Figure 3c).  It should also be noted that the thickness (or loading amount) of GOx/TREN multilayers onto the outer surface of MCFs can be notably different from the multilayer thickness onto the interior of MCFs. As already mentioned in our manuscript, it is not easy for bulky GOx to be deeply incorporated into the interior of highly porous MCFs due to the blocking-up phenomena. Although we could not quantitatively investigate the thickness of GOx/TREN multilayers coated onto the inside cotton fibers, it is reasonable to analogize that the thicknesses of (GOx/TREN) 20 and (GOx/TREN) 30 multilayers deposited onto the interior of MCFs may be thinner (presumably, corresponding to thickness of 2-3 nm for electron tunneling) than the outermost surface of metallic cotton fibers, and additionally facilitate electron tunneling from GOx within porous metal frame to metallic electrodes.
Furthermore, we highlight that the equivalent series resistance (ESR) value of highly porous MCFs is much lower than that of (GOx/TREN) m multilayer-coated nonporous and flat substrates with the same bilayer number (n) (i.e., Au NP-coated Si substrates). For example, in the case of the (GOx/TREN) 20 /MCF showing almost saturated anodic current density and the (GOx/TREN) 20 /nonporous substrate, their ESR values at 1 kHz were measured to be ~38.6 and ~62.6 Ω, respectively (see Figure R4), and resultantly the anodic current densities of (GOx/TREN) 20 multilayer-coated MCF electrode outperformed those of (GOx/TREN) 20 multilayer-coated nonporous electrodes (see Supplementary Figure 18 in the revised Supplementary Information). Figure R4. Nyquist plots of the (GOx/TREN) 20 multilayers coated onto MCF (red solid circle) and nonporous plate substrate (i.e., Au-coated Si wafer substrate, blue solid square). In this case, ESR values of each electrode were measured to be ~38.6 Ω for 20-GOx/MCF and ~62.6 Ω for 20-GOx/Au-coated Si wafer electrodes, respectively. Figure 18. Electrochemical performance of flat substrate-based anodes. a, Anodic current density curves. b, The normalized current density levels of cotton substrate-and Si wafer-based anodes measured at +0.6 V as a function of m. All the measurements were performed at a scan rate of 5 mV s −1 in a PBS containing 300 mmol l −1 glucose under ambient conditions. Here, the normalized anodic current density performance is ~0.01 mA cm −2 for the Si wafer substrate at 300 mmol −1 glucose, compared to the value of the cotton substrate, ~33.6 mA cm −2 (m = 20). 9 We also observed that the electron transfer kinetics at the interface of the electrode are significantly influenced by a kind of outermost layer of GOx/TREN multilayers (see Figure R5). More specifically, OH-functionalized glucose has a high affinity (mainly, hydrogen-bonding interaction) with aminefunctionalized TREN than negatively charged GOx (due to the presence of carboxylate ion (COO − ) groups of GOx). In line with this high affinity, glucose could be easily diffused into outermost TRENcoated multilayer films, and therefore the outermost TREN-coated electrode exhibited much lower electron transfer resistance than outermost GOx-coated electrodes in spite of the increase of layer number. These similar phenomena were also demonstrated by Willner group [Langmuir, 17, 1110[Langmuir, 17, -1118[Langmuir, 17, (2001]. They have shown that the electron transfer resistance at the interface between the electrode and the electrolyte can be controlled by the electrostatic attraction or repulsion between the charged outermost layer and the redox label in the electrolyte [i.e., FeI(CN) 6 3−/4− ]. Considering that all the biofuel cells shown in our study are based on the outermost TREN-coated electrodes as well as highly porous electrodes, the electron transfer resistances of metallic cotton fiber-based electrodes can be significantly lowered compared to those of other nonporous electrodes. Figure R5. Outermost layer-dependent interfacial electron transfer kinetics (tested on the Au-coated Si wafer substrate). a, Nyquist plots of (GOx/TREN) m multilayers as a function of bilayer number (m). b, Representative schematics of (GOx/TREN) m multilayers having different surface charge as a function of outermost layer. c, Specific data sheet of (GOx/TREN) m multilayers.

Supplementary
In summary, the electron transport mechanism of the (GOx/TREN) m multilayer-coated MCFs without electron mediator basically follows the electron tunneling mechanism between GOx and metal electrode. Although the bilayer number (m) of (GOx/TREN) m multilayers was increased up to 10 20 and 30, the thickness of (GOx/TREN) m multilayers deposited onto the outer surface of MCFs was measured to be ~9 ± 3 nm (for 20 bilayers) and ~13 ± 5 nm (for 30 bilayers). Considering the dimension of GOx (6.0 x 5.2 x 7.7 nm 3 ), these thickness range of the multilayers indicate that GOx enzymes are adsorbed on the substrates as single or double layers. However, in the case of the GOx/TREN multilayers onto the interior of highly porous MCFs, it is considered that their thicknesses will be much thinner with going from the outer surface to the center region of MCFs because the repetitive deposition of bulky GOx layers can block-up the porous structure of MCFs. Therefore, the overall anodic current density of MCF-based anode can be gradually increased up to the thickness of GOx/TREN multilayers allowing efficient electron tunneling. Additionally, in view of the structural frame, highly porous MCFs with the large surface area can significantly decrease electron transfer resistance compared to other nonporous electrodes, which can consequently boost up the power performance of metallic cotton-based biofuel cells.

On the revised Supplementary Information:
Supplementary Figure 11. Cross-sectional FE-SEM images for film thickness of a, (GOx/TREN) 20 and b, (GOx/TREN) 30 multilayers deposited onto the Au-coated Si wafer. Figure 14. Nyquist plots of the (GOx/TREN) 20 multilayers coated onto MCF (red solid circle) and nonporous plate substrate (i.e., Au-coated Si wafer substrate, blue solid square). In this case, ESR values of each electrode were measured to be ~38.6 Ω for 20-GOx/MCF and ~62.6 Ω for 20-GOx/Au-coated Si wafer electrodes, respectively. Figure 15. Outermost layer-dependent interfacial electron transfer kinetics (tested on the Au-coated Si wafer substrate). a, Nyquist plots of (GOx/TREN) m multilayers as a function of bilayer number (m). b, Representative schematics of (GOx/TREN) m multilayers having different surface charge as a function of outermost layer. c, Specific data sheet of (GOx/TREN) m multilayers.

Supplementary
On page 9 in the revised manuscript: "Additionally, the formed (GOx/TREN) m multilayers, which are sequentially deposited with small-molecule linkers instead of conventional bulky polyelectrolytes 35−37 , exhibited the ultrathin film thicknesses of ~9 ± 3 nm for 20 bilayers and ~13 ± 5 nm for 30 bilayers (Supplementary Fig. 11). Considering the dimension of GOx (6.0 x 5.2 x 7.7 nm 3 ) 48 , these multilayer thicknesses indicate that (GOx/TREN) 20 and (GOx/TREN) 30 are adsorbed on the substrates like a single GOx layer and double GOx layers, respectively. As a result, the adsorption conformation of thin and uniform (GOx/TREN) m multilayers can facilitate the electron transfer through the enzyme layers." On page 9 in the revised manuscript: "This bilayer (m)-dependent anodic performance can be explained by different adsorption behavior of the GOx layer on the inner and outer surface of the MCFs. In other words, in the case of the GOx/TREN multilayers onto the interior of highly porous MCFs, it is considered that their thicknesses will be much thinner with going from the outer surface to the center region of MCFs because the repetitive deposition of bulky GOx layers can block up the porous structure of MCFs. Therefore, the overall anodic current density of MCF-based anode can be gradually increased up to the thickness of GOx/TREN multilayers allowing efficient electron tunneling." On page 10 in the revised manuscript: "However, the ESR value of the MCF-based anode electrode is much lower than that of the nonporous substrate (i.e., Au-coated Si wafer)-based anode electrode (Supplementary Fig. 14) although the loading amount of (GOx/TREN) 30 multilayers onto MCF electrode is approximately 185 times higher than that of (GOx/TREN) 30 multilayers onto nonporous electrode (i.e., ~869.5 μg cm −2 for MCF-based anode and ~4.7 μg cm −2 for nonporous substrate-based anode at 30 bilayers), which indicate the effective electrode conformation of the MCF-based anode for facile electrochemical kinetics." On page 10 in the revised manuscript: "We also observed that the electron transfer kinetics at the interface of the electrode are significantly influenced by a kind of outermost layer of GOx/TREN multilayers (Supplementary Fig. 15). More specifically, OH-functionalized glucose has a high affinity (mainly, hydrogen-bonding interaction) with amine-functionalized TREN than negatively charged GOx at pH 7.4 [due to the presence of carboxylate ion (COO − ) groups of GOx]. In line with this high affinity, glucose could be easily diffused into the outermost TREN-coated multilayer films, and therefore the outermost TREN-coated electrode exhibited much lower electron transfer resistance than the outermost GOx-coated electrodes in spite of the increase of layer number. These similar phenomena were also demonstrated by Willner group 49 . Considering that all the biofuel cells shown in our study are based on the outermost TREN-coated electrodes as well as the highly porous electrodes, the electron transfer resistances of MCF-based electrodes can be significantly lowered compared to those of other nonporous electrodes." On references:
Additionally, unlike our micro-sized implantable fiber-type MCF-BFCs, they used carbon fiber sheet (1 cm x 1 cm, 750-μm thickness) for anode and cathode electrode. These large-sized BFC electrodes have difficulties in terms of the practical implantable devices.
On the other hand, our BFC is membrane-less system and does not use any redox mediators for BFC system as already mentioned in the introduction part of our original manuscript. In general, a membrane-less BFCs have advantages such as small size for implantation into living organisms, longterm stability, and low cost, compared to the membrane-based BFCs. Additionally, to our knowledge, the resulting output power density of 3.7 mW cm −2 in our membrane-less and mediator-less BFC is the best record obtained from the direct power measurement of current flow through an external resistor, which is approximately 36 times higher than the previously-reported best stationary power density (~0.1 mW cm −2 ) of membrane-less and mediator-less BFC (tested under oxygen saturated condition, pH 3) by Prof. Yongchai Kwon's group [Sci. Rep., 6:30128 (2016)].

Reviewer #2
Overall comments to the Author: The authors report here the development of a hybrid biofuel cells that employes a glucose oxidase anode, made of metallic cotton fibres, and a cathode made of the same electrode. Overall the manuscript is interesting and very well presented. Particularly appreciated are the characterisation made on the electrodes, both material characterisation and electrochemical characterisation, which provides a very comprehensive and systematic study. As such, this Reviewer suggests publications, after major revision.

Comment 1. One aspect the authors should focus on and enhance is the part related to the fuel cell operation, as most of the manuscript focuses on the development and characterisation of the electrodes only. the results should be better discussed with reference on existing literature and particularly mediator-free enzymatic fuel cells.
Response 1: We appreciate the reviewer #2's valuable comment on improving the manuscript. According to the reviewer #2's comments, we have modified the main text and included the suggested recent three references regarding the mediator-free enzymatic fuel cells (see Supplementary References 15-17). As checked in the Supplementary References 15-17, mediator-free enzymatic fuel cells generally provided relatively low power output generally, ~1 mW cm −2 or smaller than that value, measured by an external variable resistor), compared to the mediated electron transfer BFCs. We have additionally included the suggested reference regarding the best previously-reported stationary power density of membrane-less and mediator-less BFC (tested under oxygen saturated condition, pH 3) by Prof. Yongchai Kwon's group [Sci. Rep., 6, 30128 (2016)

For more clarifying this issue raised by Reviewer #2, we have additionally added the new discussion especially for the mediator-free enzymatic fuel cells in the modified Supplementary Table 1, and new text related to the fuel cell operation were included in the revised manuscript and Supplementary Information as follows.
********************************************************************************** Supplementary References 15, 16, 17  On page 15 in the revised manuscript: "This high-power output (3.7 mW cm −2 ) obtained from our membrane-free and mediator-free MCF-BFC is in stark contrast with those measured from other mediator-free enzymatic fuel cells (i.e., in the case of being measured by an external variable resistor, the power output of mediator-free enzymatic fuel cells is generally less than ~1 mW cm −2 ). For examples, the power output shown in our system is nearly three times higher than that of the previous DET-based compressed CNT electrode 5 (1.25 mW cm −2 , measured by polarization tests (i.e., CV measurements) without an external variable resistor), and even higher than that of the reported osmium complex-based biscrolled carbon nanotube yarn MET-BFC 7 (2.18 mW cm −2 ). Consequently, we demonstrated that our highly porous metallic cotton fiber-based BFC (MCF-BFC) fabricated by small molecule linker-induced LbL assembly of the active catalysts can significantly realize the DET rate between the enzymes and the conductive supports."

On the revised Supplementary Information (with regard to the suggested references for mediatorfree enzymatic fuel cells,
On page 9 in the revised manuscript: "This bilayer (m)-dependent anodic performance can be explained by different deposition behavior of the GOx layer on the inner and outer surface of the MCFs. In other words, in the case of the GOx/TREN multilayers onto the interior of highly porous MCFs, it is considered that their thicknesses will be much thinner with going from the outer surface to the center region of MCFs because the repetitive deposition of bulky GOx layers can block up the porous structure of MCFs. Therefore, the overall anodic current density of MCF-based anode electrode can be gradually increased up to the thickness of GOx/TREN multilayers allowing efficient electron tunneling."

On page 13 in the revised manuscript (for the enhancement related to the anode performance):
"Additionally, we expect that the anodic performance can be further controlled and enhanced with increasing the electrical conductivity of MCF for electrical communication and the surface area of MCF for efficient enzyme loading (by a further increase (n > 20) of bilayer number of (TOA-Au NP/TREN) n multilayers onto cotton fibers)." Figure 13): "In this case, the 30-GOx/20-MCF anode maintained ~97.5 % of its initial current density after continuous operation for 1 hour (herein, the initial current density of the anode electrode was obtained after pre-training for 1 day in 300 mmol l −1 glucose containing PBS buffer condition), indicating the excellent electrode stability caused by the precisely controlled deposition of GOx using TREN (Supplementary Fig. 13)."  Figure 14): "However, the ESR value of the MCF-based anode electrode is much lower than that of the nonporous substrate (i.e., Au-coated Si wafer)-based anode electrode (Supplementary Fig. 14) although the loading amount of (GOx/TREN) 30 multilayers onto MCF electrode is approximately 185 times higher than that of (GOx/TREN) 30 multilayers onto nonporous electrode (i.e., ~860 μg cm −2 for MCF-based anode and ~4.65 μg cm −2 for nonporous substrate-based anode at 30 bilayers), which indicate the effective electrode conformation of the MCF-based anode for facile electrochemical kinetics."

On the revised Supplementary Information:
Supplementary Figure 14. Nyquist plots of the (GOx/TREN) 20 multilayers coated onto MCF (red solid circle) and nonporous plate substrate (i.e., Au-coated Si wafer substrate, blue solid square). In this case, ESR values of each electrode were measured to be ~38.6 Ω for 20-GOx/MCF and ~62.6 Ω for 20-GOx/Au-coated Si wafer electrodes, respectively. Figure  15): "We also observed that the electron transfer kinetics at the interface of the electrode are significantly influenced by a kind of outermost layer of GOx/TREN multilayers (Supplementary Fig.  15). More specifically, OH-functionalized glucose has a high affinity (mainly, hydrogen-bonding interaction) with amine-functionalized TREN than negatively charged GOx at pH 7.4 [due to the presence of carboxylate ion (COO − ) groups of GOx]. In line with this high affinity, glucose could be easily diffused into the outermost TREN-coated multilayer films, and therefore the outermost TRENcoated electrode exhibited much lower electron transfer resistance than the outermost GOx-coated electrodes in spite of the increase of layer number. These similar phenomena were also demonstrated by Willner group 49 . Considering that all the biofuel cells shown in our study are based on the outermost TREN-coated electrodes as well as the highly porous electrodes, the electron transfer 20 resistances of MCF-based electrodes can be significantly lowered compared to those of other nonporous electrodes."

Comment 2. It is noted that the graphs related to the electrochemical characterisation of the electrodes and of the fuel cells, have no error bars. Have the authors done any replicates of the experiments? what is the reproducibility of the results?
Response 2: We appreciate the reviewer #2's comment. In the present work, all of our reported results were obtained from reproducible average values of three or five replicated samples in each experiment conditions.
Response 3: We appreciate the reviewer #2's comment. We address each in turn below.

3-1. With regard to loading amount of the enzyme (GOx).
Actually, it is not easy to measure the exact loading amount of GOx and TREN separately because the LbL-assembled GOx/TREN multilayers were distributed evenly throughout the interior and outer surface of the cellulose fiber [see Figure 4a (right side) and Supplementary Figure 22c,d in the revised manuscript and Supplementary Information]. However, given that TREN molecule has a very low M w (~146 g mol −1 ) compared to bulky GOx with M w of ~160 kDa, it is assumed that the loading amount of GOx in the multilayers is dominant.
In the case of 30-bilayered GOx/TREN multilayers, the thickness was measured to be ~13 ± 5 nm (please see Figure R7). We also quantitatively measured the loading amount of the GOx/TREN multilayers coated onto the MCF using an analytical balance (XP205 model, Mettler Toledo, resolution of 0.01 mg). As a result, the loading amount of (GOx/TREN) 30 multilayers was measured to 25 be approximately 860 μg cm −2 (~408 ± 35 μg), which was ~185 times higher than that of the nonporous substrate (i.e., Au-coated Si wafer)-based one (~4.65 μg cm −2 ) at the same 30 bilayers.
On the revised manuscript:

Figure R7
. Cross-sectional FE-SEM images for film thickness of (GOx/TREN) 30 multilayers deposited onto the Au-coated Si wafer.

3-2. With regard to reproducibility and stability of the electrodes.
Our BFC electrodes were prepared by electrostatic LbL assembly of the negatively charged GOx (5 mg ml −1 in PBS for 10 min) and the positively charged TREN (1 mg ml −1 in PBS for 10 min), at pH 7.4 PBS solution. All the experiments were carried out at least three or five times using independent sample electrodes under same experimental conditions, and the results were selected the reasonable average values (i.e., standard deviation) for each analysis (more details can be found in Experimental Section in the main manuscript). In order to clarify the reproducibility and stability of our enzymatic anode electrodes, we have tested time-dependent anodic current density change [I/I 0 (%)] for 1 hour (Here, the initial current density of the anode electrode was obtained after pre-training for 1 day in 300 mmol l −1 glucose containing PBS buffer condition) for (GOx/TREN) 30 /20-MCF (see Figure R9) (here, the results obtained indicate the average values from three different electrodes in each case). As shown in Figure R8, the (GOx/TREN) 30 /20-MCF showed the excellent operation stability of ~97.5% of initial current density with continuous operation. This high operation stability of TRENbased BFCs can be explained by the role of amine-functionalized molecule linker (i.e., TREN). In this case, TREN plays important roles in not only reducing the inter-distance and contact resistance between the introduced materials (i.e., GOx/GOx, GOx/Au NP, and Au NP/Au NP) but also providing the stable bonding of each materials by electrostatic and/or covalent bonds, resulting in the desirable operation stability and output performances. This is quite different from the conventional BFC electrode conformation with bulky and thick single enzyme layer, resulting in the low activation efficiency despite the high mass loading of GOx due to their insulating nature and self-aggregation. Figure R8. Stability of anodic current density (%) for (GOx/TREN) 30 /20-MCF anode electrodes.

3-3. With regard to the leaching issue over long-term operation.
According to the long-term stability test shown in the above-mentioned Figure R9, we confirmed that there was a leaching problem of the enzyme in the enzymatic anode electrode. However, the anodic current density measured for 1 hr was stably operated with a slight decrease from the initial status 27 with a sufficient pre-training step (~1 day). For more confirmation, the complete BFC cell prepared by connecting the anode (i.e., 30-GOx/20-MCF) and cathode (i.e., 120-MCF) electrode was continuously operated in 10 mmol l −1 glucose buffer solution for prolonged operation time. These results are shown in Figure R9. However, to more clarify the issues raised by reviewer #2, we have additionally added the discussion and Figure R8 and R9 to our revised manuscript and Supplementary Information as follows.

Comment 4. Was the fuel cell performance assessed only via polarisation tests or was the voltage versus time recorded for long period of times?
Response 4: We appreciate the reviewer #2's comment. In the present study, the power densities were obtained by operating the BFC through an external variable resistor to control the cell voltage and measure the current, instead of transient power density via polarization tests. These power densities were continuously measured for a long period of time, 35 days.

On the caption of Supplementary
Response 5: We appreciate the reviewer #2's comment. The peak power densities in Supplementary  Figure 22c in the original version (please see Figure R10 and Supplementary Figure 26d in the revised Supplementary Information). All power density values were determined by measuring the current flowing through an external variable resistor in the range of 1 kΩ ~ 10 MΩ to control cell voltage. Additionally, if the power density of MCF-based biofuel cells shown in our study was measured using polarization tests, the output power density could be increased up to 8.7 mW cm −2 due to the generation of capacitive and other parasite currents, which was also tested with continuous operation in PBS buffer (see Figure R11). However, given that this ultrahigh power density value was not stationary power possibly used for practical applications but transient power, these power output results obtained from polarization tests were not included in our manuscript. We also strongly agree with the referees' comments that biofuel cells should be continuously operated, and this is the essential condition for practical applications.  To more clarify the issues raised by reviewer #2, we have additionally added the Figure R11 to our Supplementary Information as follows. **********************************************************************************