Versatile templates from cellulose nano ﬁ brils for photosynthetic microbial biofuel production † – 5835 | 5825 Journal of

Versatile templates were fabricated using plant-derived nanomaterials, TEMPO-oxidized cellulose nano ﬁ brils (TEMPO CNF) for the e ﬃ cient and sustainable production of biofuels from cyanobacteria and green algae. We used three di ﬀ erent approaches to immobilize the model ﬁ lamentous cyanobacteria or green algae to the TEMPO CNF matrix. These approaches involved the fabrication of: (A) a pure TEMPO CNF hydrogel; (B) a Ca 2+ -stabilized TEMPO CNF hydrogel; and (C) a solid TEMPO CNF ﬁ lm, which was crosslinked with polyvinyl alcohol (PVA). The di ﬀ erent immobilization approaches resulted in matrices with enhanced water stability performance. In all cases, the photosynthetic activity and H 2 photoproduction capacity of cyanobacteria and algae entrapped in TEMPO CNF were comparable to a conventional alginate-based matrix. Green algae entrapped in Ca 2+ -stabilized TEMPO CNF hydrogels showed even greater rates of H 2 production than control alginate-entrapped algae under the more challenging submerged cultivation condition. Importantly, cyanobacterial ﬁ laments entrapped within dried TEMPO CNF ﬁ lms showed full recovery once rewetted, and they continued e ﬃ cient H 2 production. The immobilization mechanism was passive entrapment, which was directly evidenced using surface sensitive quartz crystal microbalance with dissipation monitoring (QCM-D). The results obtained demonstrate a high compatibility between CNF and photosynthetic microbes. This opens new possibilities for developing a novel technology platform based on CNF templates with tailored pore-size and controllable surface charges that target sustainable chemical production by oxygenic photosynthetic microorganisms.


Introduction
Cyanobacteria and green algae are widely distributed photosynthetic microorganisms, which are capable of harnessing solar energy and converting it into chemical energy by assimilating atmospheric CO 2 . They are considered to be suitable feedstocks for blue bioreneries and attractive biocatalysts for the production of biofuels such as biodiesel, isoprene, alcohols, ethylene and other valuable industrial compounds. [1][2][3] Theoretical photosynthetic light conversion efficiencies to carbon-based products of approximately 4.6% have been reported, 4 which would satisfy the feasibility of industrial applications. Besides carbon-based biofuels, cyanobacteria and green algae can also produce molecular hydrogen (H 2 ), which has the highest energy content of reported biofuels and a zero CO 2 release index. 5 These qualities make H 2 an ideal energy carrier for the sustainable bioeconomy of the future. The photosynthetic conversion efficiency to H 2 is much higher than to other carbon-based biofuels. For example, nitrogenase-driven H 2 photoproduction in cyanobacteria has a maximum efficiency of around 6-7%, 6 while direct water biophotolysis in green algae can drive H 2 production with a potential efficiency of 10-12%. 7,8 In reality, however, only a small fraction of these values has been achieved, even in laboratory scale photobioreactors. This is due to various metabolic and production process hurdles, such as the low light utilization efficiency of suspension cultures, sensitivity of photosynthetic organisms to environmental factors, the competition for photosynthetic reductant from wasteful metabolic pathways and the high energy demand of cell cultivation, harvesting and maintenance of photobioreactors.
Thin-layer immobilization is a novel technology recently proposed to ensure uniform light distribution to high-density phototrophic cultures xed within a controllable volume, and for increasing light-to-product conversion efficiency. 9,10 Besides better light utilization (as compared to suspensions), the entrapment of the dense cultures within the immobilization matrix limits cell division and engages more efficient distribution of photosynthetic reductant to the production of desired end-products, instead of biomass accumulation.
Immobilization also protects the cells from harsh environmental conditions, eases the maintenance of photobioreactors and dramatically prolongs the cultivation period. 11,12 The combination of these factors improves the volumetric and areal productivity of photobioreactors with immobilized photosynthesizing cells. Indeed, Gosse et al. 11,13 have recently demonstrated that the entrapment of the phototrophic bacterium, Rhodopseudomonas palustris within latex coatings extends the duration of H 2 photoproduction leading to increased H 2production yields compared to suspension cultures. A similar effect has been demonstrated for H 2 -producing green algae 14,15 and heterocystous cyanobacteria 16,17 entrapped within thin Ca 2+ -alginate hydrogel cubes, beads and lms. The improvement of photosynthetic productivity was also conrmed for a few species of cyanobacteria entrapped in the hydrated latex coatings. 18 Although the data are not available, thin-layer immobilization is expected to improve production yields of other bio-based chemicals as well.
Unfortunately, thin-layer immobilization of cyanobacteria and green algae is currently limited to the use of alginate for cellular entrapment. Other materials, such as latex or sol-gel are either toxic to the cells or too expensive for industrial applications. For sustainable production, immobilization matrices are expected to be biodegradable. Although biodegradable, not toxic to the cells and available in industrial quantities, alginates have low mechanical stability and porosity. The low mechanical stability of alginate limits its utilization in long-term applications, especially in media with high contents of phosphates and other chelating agents. Low porosity leads to the accumulation of photosynthetically-evolved O 2 inside the matrix causing photoinhibition and oxidative damage to the entrapped cells, especially under high light conditions. 19 Importantly, the low porosity of alginate also precludes its application as a matrix for the entrapment of designed cyanobacterial or algal cells secreting large molecules (e.g. sesquiterpenes).
Cellulose nanobrils (CNF) offer a range of advantages for the immobilization of cyanobacteria and green algae. CNFs are plant-based building blocks manufactured via a so-called topdown approach, which involves disintegration of the plant cell wall structures by chemical/enzymatic and mechanical means. 20 These nanoscaled building blocks are virtually inexhaustible as a source for the construction of renewable, biocompatible and biodegradable materials which are already available in industrial quantities. In this context, the most attractive CNF grade is TEMPO CNF which is produced by TEMPO (2,2,6,6tetramethylpiperidine-1-oxyl radical)-mediated oxidation of cellulosic bres. 21 During TEMPO-oxidation, the primary hydroxyl groups of cellulose are selectively converted to anionic sodium C6-carboxylate groups. The release of individual and almost monodispersed nanobers is achieved with mild mechanical treatment due to electrostatic repulsion. The width of individual brils can be as low as 3-5 nm while the length is on the micrometer scale. Due to the small bril diameter and high aspect ratio, three-dimensional network structures of TEMPO CNF hydrogels are fully transparent, as are the twodimensional structures of lms aer the removal of water. 22,23 These qualities allow excellent light penetration. In addition, the nanoscaled brillar network brings high porosity coupled with high surface area, enabling the efficient diffusion of vital nutrients and gases, allowing higher cell concentrations in the matrix and offering numerous sites for biomolecules to interact. [24][25][26] Aforementioned features also improve the mechanical stability of the structures in the presence of water although the CNF based matrices are highly hydrophilic and hygroscopic. Especially when compared to the other biobased materials, CNF gels and lms have higher mechanical strength and higher durability in aqueous environments. Several routes to further improve the water stability, CNF gel strength and wet strength of CNF lms are available. Gel strength can be improved by e.g. gelation with di-or trivalent cations, such as Ca 2+ , 27 whereas TEMPO CNF lm wet strength can be enhanced by interbrillar bridging using e.g. polyvinyl alcohol, PVA 28 or by photocrosslinking. 29 Furthermore, tailoring the surface properties and pore size of TEMPO CNF matrices offers controllable cell attachment and facilitated secretion of various chemicals produced by the immobilized cells.
In this work, we investigated several novel approaches to design templates from plant-based TEMPO CNF for the attachment and entrapment of cyanobacterial and green algal cells. Beyond the most essential characteristics which are the photosynthetic activity and the capacity for H 2 production, these biocatalytic templates simultaneously offer a number of vital features such as transparency, exibility, porosity, and most importantly routes to enhance the water stability. It is important to note that these features are inherently derived from the nanoscaled brillar network structure of TEMPO CNF, and that this is in contrast to conventional polymeric biobased materials employed for similar purposes, but for which such a combination has yet to be reported. Our ndings reveal the high compatibility of CNF and photosynthetic microbes that were entrapped either in hydrogel layers, or in dry solid lms. Water stability of the hydrogels or solid lms was improved by ionic crosslinking with Ca 2+ or by interbrillar bridging with polyvinyl alcohol. In addition, we provide direct evidence on the mechanisms behind cellular interactions with the CNF matrix and nally suggest a simple strategy, based on electrostatic interactions, to tailor the immobilization mechanism towards even better performance. These ndings open new possibilities for developing a novel technology platform with tailored pore-size and controllable surface charge that targets sustainable chemical production by oxygenic photosynthetic microorganisms.

Materials
Strains and growth conditions. The DhupL mutant of lamentous N 2 -xing heterocystous cyanobacterium, Anabaena sp. strain PCC 7120 (hereaer, DhupL) was kindly provided by Prof. H. Sakurai (Waseda University, Japan). The mutant is decient in the large subunit (accession no. AAC79878.1) of the [Ni-Fe] uptake hydrogenase and, therefore, demonstrates increased hydrogen production yields compared to the wild-type strain. 30 In this study, we always used cells grown under diazotrophic (N 2 -xing) conditions for expression of the nitrogenase enzyme and following H 2 photoproduction under argon (Ar) atmosphere. The mutant was grown in 1 L Erlenmeyer asks containing 800 mL Z8x medium, 31 which lacks combined nitrogen. The asks were placed in a growth chamber at 26 C and illuminated from two sides with $60 mmol photons m À2 s À1 (Photosynthetically Active Radiation, PAR) cool-daylight uorescent light. The cultures were sparged continuously with sterile air ltered through 0.2 mm pore-size membrane lters (Acro 37 TF, Gelman Sciences, USA). Before immobilization, cyanobacterial laments were washed once in Z8x medium and pelleted by centrifugation at 3000g for 10 min.
The wild-type green alga, Chlamydomonas reinhardtii (here-aer, C. reinhardtii) strain CC-124 (mtÀ, nitÀ) was obtained from the Chlamydomonas Resource Center at the University of Minnesota, USA. The alga was grown under the same conditions as the DhupL mutant, but on the standard Tris-Acetate-Phosphate (TAP) medium. 32 Prior to immobilization, C. reinhardtii cells were sulfur-depleted to initiate H 2 photoproduction. 33 For this, algal cultures were grown to the late logarithmic phase ($25 mg total Chl per mL), harvested by centrifugation at 3000g for 3 min, washed once in TAP-minus-sulfur-minusphosphorus (TAP-S-P) medium to remove sulfates and phosphates, and pelleted by centrifugation prior to entrapment in TEMPO CNF and alginate hydrogels. Since green algal hydrogel layers were submerged into the medium, exclusion of phosphates from the medium helps to prevent dissolution of the Ca 2+ -stabilized hydrogel matrices.
TEMPO-oxidized cellulose nanobrils (TEMPO CNF). Neverdried bleached sowood pulp (spruce/pine mix) obtained from a Finnish pulp mill was used as a raw material for the production of cellulose nanobrils. TEMPO-mediated oxidation of the pulp bers was carried out by alkaline oxidation with hypochlorite catalysed by TEMPO, according to the procedure described by Saito et al. 21 The anionic charge of the oxidated pulp was 1.28-1.41 mmol g À1 , as determined with a standard conductometric titration method (SCAN 65:02). Aer the TEMPO-oxidation the pulp was washed and brillated into CNF (hereby called TEMPO CNF) in a microuidizer (Microuidics Int., USA) equipped with two Z-type chambers with respective diameters of 400 and 100 mm. The brillation was done in two passes at 1850 bar operating pressure, and the nal consistency of the TEMPO CNF was ca. 1 wt%. The appearance of TEMPO CNF is a viscous and transparent hydrogel (light transmittance of 74.3% at 800 nm by UV-Vis spectroscopy). 23 The carbohydrate composition of TEMPO CNF is 64.0% glucose, 6.2% xylose and 2.1% mannose. 34 Fig. 1 shows the transparent and exible appearance, topography and morphology of self-standing TEMPO CNF lm, in comparison with Ca 2+ -alginate lm, which was used as a reference material.
Polymers. Sterile 4 wt% alginate solution used as a reference matrix was prepared by mixing Na-alginate salt from brown algae (#71238, Sigma-Aldrich) in Milli-Q water and by autoclaving the mix at 121 C for 10 min. Polyvinyl alcohol (Mowiol 56-98, Mw 195 000 g mol À1 , DP 4300) used as a CNF lm wet strength additive was purchased from Sigma-Aldrich and dissolved to a 5 wt% solution in Milli-Q water at 95 C. Branched polyethylene imine (M w 70 000 g mol À1 , 30 wt% aqueous solution) used as anchoring polymer for CNF thin lms was purchased from Polysciences and used as received.
Other chemicals and materials. All chemicals were of analytic grade and used as received.
2,2,6,6-Tetramethylpiperidin-1-oxyl (TEMPO), sodium bromide (solid) and 10% sodium hypochlorite (aqueous) were purchased from Sigma-Aldrich. 0.1 M sodium hydroxide solution was obtained from Fluka Analytical. CaCl 2 (99%, #C7902) was purchased from Sigma-Aldrich. Ultrapure water (18.2 MU cm) was prepared with a Milli-Q purication unit (QPAK® 1, Millipore). Whatman® cellulose blotting paper (3MM Chr, #3030-931, Sigma-Aldrich) was used as a support. For submerged cultivation of Ca 2+hydrogels, instead of blotting paper, a plastic 120 mm-thick insect screen support was applied. Melamine foam sponge was cut into 3 Â 1 Â 1 cm pieces and boiled in Milli-Q water for 1 h before use. QCM-D experiments were performed on AT-cut quartz crystal sensors with gold electrodes purchased from Q-sense AB (Gothenburg, Sweden), with a fundamental resonance frequency of 5 MHz and a sensitivity constant of 0.177 mg m À2 Hz À1 as reported by the supplier.

Immobilization procedures
Approach A: TEMPO CNF hydrogel layers. The cyanobacterial laments were entrapped within thin hydrogel layers, which were formed on top of the blotting paper support. The DhupL pellets were mixed in 1 : 1 wet-mass ratio with different hydrogels: (i) 1 wt% TEMPO CNF; (ii) 1 wt% TEMPO CNF and 0.1 wt% polyvinyl alcohol (PVA); and (iii) 4 wt% alginate as a control. In the second formulation, PVA, which was used as a crosslinker in following studies, was added to investigate its biocompatibility with the cyanobacteria. The hydrogel layers were formed by drawing down the formulations with a stainless steel rod over the blotting paper to produce uniform layers ( Fig. 2A). Finally, the layers were cut into 3 Â 1 cm strips and placed on top of 3 Â 1 Â 1 cm melamine foam sponges that supply water to the strips. The sponges with strips were placed inside 33 mL gastight vials lled with 5 mL Z8x medium. The strips were incubated at 26 C under Ar atmosphere containing 6% CO 2 and light intensity of 140 mmol photons m À2 s À1 PAR. It is important to note that all previous studies on thin-layer immobilization of cyanobacteria and green algae were performed with Ca 2+ -alginate hydrogel lms. For the sake of comparison, we therefore applied Ca 2+ to alginate polymers in all immobilization approaches mentioned here and below.
Approach B: TEMPO CNF hydrogel layers crosslinked with Ca 2+ . Algal Ca 2+ -stabilized TEMPO CNF hydrogel layers were produced on top of the template consisting of a white-polymer insect screen placed over the sticky side of a Scotch-type tape (Fig. 2B) as described before. 14 The insect screen was embedded in the hydrogel layers for improving their mechanical stability. Sulfur-depleted C. reinhardtii pellets were mixed in 1 : 1 wetmass ratio with (i) 0.5 wt% TEMPO CNF, (ii) 0.5 wt% TEMPO CNF + 0.05 wt% PVA and (iii) 4 wt% Ca 2+ -alginate as a control. The algal hydrogel layers were formed as described in the approach A. In contrast to the approach A, both TEMPO CNF and alginate hydrogel matrices were stabilized by spraying the surface with 50 mM CaCl 2 solution. The Ca 2+ -stabilized hydrogel layers were washed in Milli-Q water, cut into 6 Â 1 cm strips and submerged into 10 mL TAP-S-P medium in 75 mL gas-tight vials. In the beginning of each experiment, the gas phase in the vials was replaced to pure Ar. The vials were incubated at 26 C under uorescence light of 50 mmol photons m À2 s À1 PAR. Approach C: dry PVA-crosslinked TEMPO CNF lms with immobilised cyanobacteria and recovery by re-wetting. The DhupL PVA-crosslinked TEMPO CNF lms were formed by drying the hydrogel layers (prepared as described in the approach A) on the top of the paper support for 22 h in complete darkness at 23 C and 70% relative humidity (Fig. 2C). The hydrogel formulations consist of cell wet biomass mixed in either 1 : 1 or 1 : 2 mass ratio with 0.5 wt% TEMPO CNF supplemented with 0.05 wt% PVA. The Ca 2+ -stabilized alginate hydrogel lms (made of 4 wt% alginate solution without PVA) were served as a positive control, while the matrix-free cell biomass placed on top of the paper support was used as a negative control.
The DhupL cells in dried lms were recovered by placing the lm on top of the melamine foam sponge, which supplied Z8x medium to the entrapped cells. The recovery was done in closed Petri dishes at 26 C under air atmosphere supplemented with 0.5% CO 2 . During the 4 days recovery, strips were illuminated with 6 mmol photons m À2 s À1 PAR. Aer the recovery stage, the lm-coated paper was cut into 3 Â 1 cm strips and strips were transferred into the 33 mL gas-tight vials containing 5 mL Z8x medium. To initiate efficient H 2 photoproduction by the DhupL cells, the headspace of the vials was replaced to Ar and supplemented with 6% CO 2 . The PVA-crosslinked TEMPO CNF and Ca 2+ -alginate lms were top-illuminated with 140 mmol photons m À2 s À1 PAR.
Photosynthetic activity monitoring. The photosynthetic activity of the immobilized cells was monitored with PAM-2000 (Walz, Germany). Cells were dark adapted for 3 min before the activity measurements. Red actinic light (655 nm) of 84 mmol m À2 s À1 was applied for 210 s and saturating white light pulses (0.8 s) of 2000 mmol m À2 s À1 PAR were red every 20 s to determine F 0 m , the maximum uorescence under actinic light. The effective yield of Photosystem II (YII) were calculated as (F 0 m À F s )/F 0 m , where F s is steady-state uorescence under actinic light.
Determination of H 2 and O 2 contents. The H 2 and O 2 concentrations in the headspace of the vials were monitored using Clarus 500 gas chromatograph (PerkinElmer, Inc.) equipped with a thermal conductivity detector and a molecular sieve 5 A column (60/80 mesh). 150 mL samples were taken with a gas-tight 250 mL gas chromatography syringe (Hamilton, USA) equipped with a sample lock.
Detection of TEMPO CNF and cyanobacteria interactions by QCM-D. Attachment of cyanobacteria to the surface of TEMPO CNF thin lms was investigated with Quartz Crystal Microbalance with Dissipation Monitoring (QCM-D) using the Q-Sense E4 instrument equipped with an Open Module 401 (Q-Sense AB, Gothenburg, Sweden), which allows direct application of desired sample solutions onto the sensor surface. The QCM-D instrument allows in situ detection of mass changes between solid-liquid or solid-gas interfaces. A detailed interpretation of QCM-D measurement data can be found elsewhere, 35,36 and is briey described in ESI. † The preparation of supported TEMPO CNF thin lms with a typical thickness of circa. 10 nm 34 on QCM-D sensors was performed according to the procedure described by Eronen et al. 37 with slight modications as described by Hakalahti et al. 34 and is described in ESI. † Cationization of TEMPO CNF lms was carried out by drop casting 200 mL of 1 mg mL À1 PEI solution in Milli-Q water onto the surfaces for 30 min, followed by rinsing with Milli-Q water and drying with N 2 gas.
Prior to the cyanobacteria attachment experiments the TEMPO CNF surfaces (anionic TEMPO CNF and cationized TEMPO CNF) were stabilized by immersing in Z8x medium for 16 hours. The surfaces were then quickly dried with N 2 gas and inserted into the QCM-D Open Module (QOM 401) cell. Schematics for different cyanobacteria attachment protocols involving anionic TEMPO CNF surface and cationized TEMPO CNF surface are presented in Fig. 3. First, 250 mL of Z8x medium was pipetted directly onto the surface and Df and DD values were monitored until a stable baseline was achieved. Stable baselines were attained within 1 hour. Aer stabilization, the medium was changed to 250 mL Anabaena DhupL suspension (OD 720 z 0.1, corresponding to dry weight percent of 0.05 wt%) in Z8x medium, and again Df and DD values were monitored for 2 h. Finally, the surfaces were rinsed with Z8x medium by sequential removal and addition of 250 mL of the medium until no changes in frequency and dissipation values were observed. Aer the measurements, the crystals were dried with N 2 gas and stored in desiccator in darkness.
The estimation of nal areal mass of the attached DhupL laments on TEMPO CNF surface was determined as described by Peresin et al. 38 with some modications, by following the difference in frequency response of the sensor. The measurement was carried out with the QCM-D Flow Module (QFM 401) in air at 23 C, and is described further in ESI. † Complementary analytical methods. Optical microscopy imaging of TEMPO CNF thin lms with attached cyanobacteria was performed using a Nikon eclipse Ci-L microscope with a Nikon DS-Fi2 digital camera (Nikon, Melville, NY, USA).
Scanning electron microscopy (SEM) imaging of TEMPO CNF-PVA and Ca 2+ -alginate lms (Approach C) was performed with Zeiss Merlin FE-SEM (Carl Zeiss AG, Oberkochen, Germany), with 2 keV accelerating voltage. The samples were coated with platinum with a sputter-coater (20 mA for 30 seconds) to improve the conductivity of the samples and thus the quality of the SEM images.
Atomic force microscopy (AFM) imaging of the CNF surfaces was performed with ANASYS AFM+® (ANASYS Instruments Inc., Santa Barbara, CA USA). The images were taken in tapping mode in air using aluminium coated n-type silicon cantilevers (HQ:NSC15/Al BS, Micromasch, Tallinn, Estonia) with typical probe radius of 8 nm, force constant of 40 N m À1 and nominal resonance frequencies between 265 and 410 kHz. The images were not processed in any way.
For Chl a determination in cyanobacteria, the strips were incubated in 90% aqueous methanol at 65 C for 30 min. The Chl a content in methanol extracts was determined spectrophotometrically using extinction coefficient given by Lichtenthaler. 39 The total Chl (a + b) in green algal samples was determined in 95% ethanol extracts as described before. 19

Results and discussion
We used three different approaches for the immobilization of the lamentous cyanobacterium, DhupL and the green alga, C. reinhardtii within thin TEMPO CNF matrix layers (see Fig. 2 and methods for more details). In the rst approach, DhupL was entrapped within TEMPO CNF hydrogel layers. For improving H 2 photoproduction yields of N 2 -xing laments, the hydrogels with entrapped DhupL were placed in gas-tight vials and the head-space was changed to Ar + 6% CO 2 . 16 In the second approach, green algal cells were entrapped within TEMPO CNF hydrogel layers, which were additionally crosslinked with Ca 2+ following the protocol developed for sulfur-deprived C. reinhardtii cultures. 14 Here, sulfur-deprivation was applied for initiation of H 2 production activity in algal cells. 33 These two approaches allowed for evaluation of the biocompatibility and suitability of TEMPO-oxidized CNF matrices for the immobilization of cyanobacteria and green algae. The third approach was applied for improving the mechanical stability of the TEMPO CNF matrices. PVA-crosslinked TEMPO CNF lms with entrapped DhupL laments were formed, dried and checked for H 2 photoproduction activity, following a 4 d re-wetting period. The third approach was also applied to C. reinhardtii without success and, therefore, is not described in this paper.

Approach A: TEMPO CNF hydrogels demonstrate cyanobacterial biocompatibility similar to alginate
In order to evaluate biocompatibility of TEMPO CNF for the immobilization of cyanobacteria, DhupL laments were entrapped within thin TEMPO CNF and TEMPO CNF-PVA hydrogel layers (Fig. 2A). Photosynthetic and H 2 photoproduction activities were compared to the laments entrapped in alginate hydrogels. The Ca 2+ -alginate polymer is known as a suitable, but not very mechanically stable carrier for thin-layer entrapments of H 2 -producing cyanobacteria 16 and green algae. 14 In addition, the hydrogel layers with entrapped laments were compared with the matrix-free cells applied directly on the top of the paper support.
As shown in Fig. 4A, CNF hydrogels with entrapped DhupL laments yielded similar amounts of H 2 as the alginateentrapped and the matrix-free laments exposed to the same conditions. The maximum specic rates were slightly higher in TEMPO CNF and TEMPO CNF-PVA hydrogels as compared to alginate or matrix-free samples, being 18.0 AE 0.7, 18.3 AE 0.9, 13.7 AE 1.3, 13.0 AE 0.3 mmol H 2 (mg Chl h) À1 , respectively. The difference between samples was statistically signicant (P < 0.05). In addition, all the samples showed similar effective Photosystem II yield and net O 2 evolution (Fig. S1 †). These data indicate the biocompatibility of TEMPO CNF and TEMPO CNF-PVA matrices for the immobilization of cyanobacteria. Indeed, the activity of CNF immobilized cells was slightly increased compared to alginate-entrapped and matrix-free cells.
Approach B: algal Ca 2+ -crosslinked TEMPO CNF hydrogel layers yield more H 2 than alginate-entrapped algae Similar to alginate, TEMPO CNF hydrogels can be stabilized by di-or trivalent metal ions such as Ca 2+ , Ba 2+ , Al 3+ and Fe 3+ . The cations draw adjacent brils together, which prevents water molecules from solvating their surface, increasing their mechanical stability in water. 27 Hence, in order to create hydrogel layers with increased performance in a submerged state, the layers with entrapped C. reinhardtii cells were stabilized by Ca 2+ and placed in sulfate/phosphate-free medium in the vials under an Ar atmosphere. Here, algae entrapped in CNF hydrogels yielded more H 2 than the cells entrapped in alginate (Fig. 4B), but all samples showed similar effective Photosystem II yield in the beginning and at the end of the experiment (Fig. S2 †). The effect was most likely due to the different structures of the materials in the nanoscale (Fig. 1). The TEMPO CNF hydrogel is formed by a highly porous brillar network, 26 whereas alginate hydrogels possess a more granular structure. This suggests lower porosity for alginate, which may restrict H 2 photoproduction by limiting diffusion of H 2 from the alginate layers. In contrast to nitrogenase-dependent irreversible H 2 evolution in heterocystous cyanobacteria, the [FeFe]hydrogenase enzyme in green algae catalyses a reversible reaction of H 2 production. Therefore, we hypothesise that accumulation of H 2 inside the alginate matrix would increase the back reaction (H 2 uptake) in algal cells. This is in line with studies showing that the H 2 photoproduction yield in C. reinhardtii declines with increased H 2 partial pressure. 40,41 Approach C: solid cyanobacterial TEMPO CNF lms recover biocatalytic activity aer drying To further increase the mechanical stability and wet strength of the hydrogel lms, the cyanobacterial TEMPO CNF-PVA hydrogels were exposed to a drying process for 22 h at 23 C and 70% relative humidity. Drought stress is known to reduce cyanobacterial cell tness and strongly decrease photosynthetic activity. 42 For evaluating the possible inhibition/recovery of photosynthetic apparatus throughout the lm formation process, we monitored the effective Photosystem II yield of the entrapped DhupL laments. As expected, photosynthetic activity dramatically declined upon drying in all studied lms (Fig. 5A). However, the re-wetting phase (Fig. 5B) allowed a gradual recovery of photosynthetic activity. On the third and fourth days of re-wetting, cyanobacteria entrapped in thin lms demonstrated a similar photosynthetic activity as observed initially. As opposed to the entrapped cyanobacteria, the matrix-free control laments could only partially recover to about 60% of the initial effective Photosystem II yield (Fig. 5B) and 55% of the initial Chl a level (Table S1 †).
The better recovery in TEMPO CNF lms may be explained by the hygroscopic structure of the material. Due to the inherent hygroscopic properties of the nanoscaled brillar network matrix, TEMPO CNF lms still contain structural water as multilayers and clusters at 70% RH, 34 and this moisture protects cyanobacterial laments during drying.
Aer the full four-day recovery period, DhupL solid lms were subjected to conditions favourable to H 2 photoproduction (Ar + 6% CO 2 ). As shown in Fig. 5C, the DhupL laments entrapped within 4 wt% alginate lms yielded the highest H 2 amount by the end of the $9 d (211 h) experiment. Among CNF formulations, the lms with higher matrix to biomass ratio showed better H 2 production yields, most likely due to a better cell recovery (Table S1 †).
Importantly, during the recovery stage, the kinetics and amplitude of the inhibition and recovery processes were similar in TEMPO CNF-PVA and alginate lms, suggesting applicability of these solid lms as a novel technology platform.
Revealing the interactions between cyanobacteria and TEMPO CNF using QCM-D Immobilization mechanisms and the interactions between TEMPO CNF and cyanobacteria were revealed using quartz crystal microbalance with dissipation monitoring (QCM-D). QCM-D measurements provide quantitative information on phenomena taking place at solid-liquid and solid-air interfaces, enabling the determination whether the cyanobacterial laments are attached to the TEMPO CNF brils, or merely passively entrapped in the matrix.
Here we investigated the interactions of DhupL laments to TEMPO CNF using a specically developed Open Module QCM-D measurement cell. With this set up, the lament-containing liquid sample is directly pipetted on the bril covered sensor surface and, thus, sensitive biomolecules with large size distribution can be introduced without e.g. pumping through the thin tubing. Fig. 6A shows the changes in frequency (Df) and dissipation (DD), which occurred due to changes in mass and viscoelastic properties monitored at solid-liquid interface of the TEMPO CNF ultra thin lm. First, the TEMPO CNF surface was stabilized in the growth medium (Fig. 6A, buffer stabilization step) for 1 h in order to attain a more or less stable baseline. This step was followed by the introduction of the cyanobacterial suspension. Sharp changes in both Df and DD signals were observed, due to the mechanical stress caused by the change of liquid, which was directly pipetted onto the QCM-D sensor surface. Attachment of cyanobacteria on the TEMPO CNF surface was monitored for 2 h and the system was then rinsed again with pure buffer medium in order to remove unattached laments. The minor negative change in frequency of À5.8 Hz indicates that cyanobacteria laments do not spontaneously attach to TEMPO CNF surface.
To further investigate the interactions between TEMPO CNF and cyanobacteria, mass changes were assessed at the solid-air interface under dry conditions (Fig. 6B). Here, the change in frequency was monitored before and aer the cyanobacteria attachment step, aer buffer rinsing. At the solid-air interface the frequency change (Df) was À32 Hz, corresponding to an increase in areal mass (Dm) of 560 ng cm À2 according to the Sauerbrey equation (eqn (1) in ESI †). The results indicate that minor lament attachment occurs when the laments are forced to approach at close proximity to TEMPO brils, due to removal of the buffer. Strong and spontaneous attachment is prevented by the overall anionic surface charges of TEMPO CNF and the DhupL laments at neutral pH. 43,44 As shown in Fig. 4 and 5, highly anionic TEMPO CNF provides a good immobilization matrix for cyanobacteria and green algae with good H 2 production yields, although the immobilization mechanism in this case appears to be a passive entrapment. However, the direct attachment of cells to the surface of the solid matrix support creates a natural and lessinvasive route to whole-cell immobilization, 45 which mimics the natural biolm formation process. 46,47 Due to the direct contact of cells and matrix from at least one direction, this allows the accurate control of nutrient distribution and gas exchange between the cells and environment is readily facilitated, especially for monolayer cell structures. Direct attachment is considered to be particularly benecial for cyanobacteria and green algae that oen experience oxidative stress due to excessive and unwanted photosynthetic O 2 production. 19,48 In order to further improve the immobilization matrix towards direct cell attachment, the CNF surface properties needed to be adapted, whilst retaining lm forming capacity and optical transparency. This could be achieved using a simple surface modication route involving the adsorption of oppositely charged polyelectrolyte on TEMPO CNF surface. Here, polyethylene imine (PEI) was utilised to cationize the TEMPO CNF surface in order to improve the attachment of cyanobacteria via electrostatic attraction. The surface interactions between PEI-cationized TEMPO CNF thin lms and DhupL laments were investigated with the QCM-D Open module, in exactly the same approach used for anionic TEMPO CNF. As hypothesized, the change in frequency aer cyanobacteria introduction was signicantly larger than with anionic TEMPO CNF (Fig. 6). Within 2 h, a frequency change of approximately À27 Hz was recorded, along with a rather large positive dissipation change of 6 Â 10 À6 . A moderately low but signicant frequency change coupled with such a high dissipation change indicate that a rather so and uneven layer is formed on the cationized TEMPO CNF surface. Furthermore, the rinsing with pure buffer solution did not generate any further signal changes, indicating that the layer is at least moderately attached on the surface i.e. it remains intact despite the mechanical stress caused by the rinsing step.
The same sensor was again analysed under dry conditions, before and aer the cyanobacteria attachment measurement. Interestingly, the frequency changes assessed at the solid-air interface resulted in a considerably larger areal mass change, being as high as 27 000 ng cm À2 . The changes in frequency and areal mass at the solid-air interface in fact exceeded the sensitivity limits of the QCM-D instrument. This substantial difference between the results observed at the solid-liquid and solid-air interfaces suggests that during the removal of liquid, other phenomena besides pure surface forces positively contribute to the attachment of cyanobacteria laments to the cationized TEMPO CNF surface. It seems that once the laments contact the cationic cellulose nanobrils, they strongly attach to the bril surface. The same remarkable difference is visually represented in the optical microscopy images (Fig. 6C). Only a few attached cyanobacterial laments are visible on the anionic TEMPO CNF surface, whereas signicantly larger amount of laments can be seen on the cationized TEMPO CNF surface. Moreover, AFM topography images taken from the regions between the laments reveal the ne structure of the surfaces. Small granular particles which are likely initially adsorbed via attractive surface forces are presumably extracellular polysaccharides and other macromolecules secreted by the cyanobacteria. The majority of the large laments are attached via oppositely charged sites but only once direct contact is achieved, for example during the removal of buffer. This attachment, taking place on several levels, can be considered as benecial to immobilization approaches where drying is involved.

Conclusions
Immobilization matrices for the photosynthetic microbial production of H 2 were fabricated using TEMPO-oxidised cellulose nanobrils with different strategies employed to enhance the matrix water stability. We showed that these versatile templates display high compatibility with oxygenic phototrophic organisms, and that they are suitable for thin-layer cell entrapments both in hydrogel and dry solid lm states. They possess several advantageous features, which make them more attractive than other natural matrix materials. Beyond their demonstrated suitability to maintaining cellular photosynthetic activity and H 2 production capacity, constructed TEMPO CNF matrices are also mechanically stable under wet conditions, transparent and have nanoscale porosity. As a material, CNF is abundant, renewable and biodegradable. To further tailor the system towards direct cell attachment, only a simple interfacial tailoring with cationic polyelectrolyte is needed. These ndings will facilitate the development of cost-effective approaches to large-scale bio-industrial applications.

Conflicts of interest
There are no conicts to declare.