HESTIA aims to develop a new insulative material to meet climate needs, reduce waste and aims towards a ‘green revolution’ of building industry. Made of a cellulose aerogel coated with recombinant proteins to imbue modular properties such as fire and water resistance, this innovative material has excellent insulating properties and is biodegradable.

On this page, we present the results of our experiments, starting from the aerogel production, continuing to protein expression and purification and, edding with the binding of our modular protein coating and the aerogel.

Click the images below to read more about the respective sections.


A crucial part of the project was to manufacture a novel porous material containing ideally only cellulose fibres. The cellulose fibres were made from plant cellulose, since it was easier to find. Two main methods were explored in the following part to manufacture the aerogel and several tests were conducted to test the characteristics of the created samples. We started by assessing the material’s physical properties and we extended our analysis toward exploring the two main properties that this material should possess to be considered as an insulation material, mechanical and thermal resistance.

Physical Assessment

During the project, we implemented two methods well defined in the field1 using different approaches and equipment. Once the hydrogel was generated, there were two methods tested to generate an aerogel. The first method relied upon preparing an alcogel from the hydrogel samples by utlising ethanol solvent exchange and then exchanging the ethanol with CO2 in gaseous form to avoid damaging the cellulose structure in the next step Critical Point Drying (CPD) The second method was simpler: freezing the sample directly as a hydrogel and then Freeze Drying (FD) it in a lyophilizer at -50 deg with a vacuum pressure lower than 0.012 mbar.

In order to understand the aerogel manufactuing parts we exposed below a table giving the annotations we adopted:

Table (1)Celluse content in aerogels.
Cellulose aerogel concentration Cellulose (g) Dispersion solution* (g)
1X 5 100
1.5X 7.5 1000
1.75X 8.75 100
Critical Point Drying

During our initial experiments we had access to a Critical Point Dryer at the Centre to the MicroNanoTechnology (CMi), a clean room facility in EPFL. The use of the Critical Point Dryer proved unsustainable due to numerous technical constraints that come with a clean room, and was stopped after having successfully produced a viable sample

Figure 1Hydrogel and aerogels with initial 50 mL volume samples.(A) Hydrogel sample with cellulose concentration of 2 X, after first ethanol bath. (B) Hydrogel sample with cellulose concentration of 2 X, after the fifth ethanol bath. (C) Critical Point Dried (CPD) aerogel sample with cellulose concentration of 1X same day. (D) CPD aerogel sample with cellulose concentration of 1X after four days.

The aerogel samples displayed in (fig 1), were made of 50 g dispersing solution with 2.5 g Thiourea and 5 g NaOH. Closer analysis of (fig 1.C,D), showed a gradual appearance of red colour in the sample from an ethanol bath to another. The phenomena was explained by the fact that ethanol took the place of the initial salt and solvent. The obtained alcogel has a more solid structure then the hydrogel which was reinsuring in this process of solvent exchange.

Directly after Critical Point Drying (CPD), the porous sample (fig 1.C) turned white due to the loss of all the liquid content from the material. The texture of the material was generally flat with small bumps. After four days (fig1 D), The sample shrank considerably from the initial size immediately after the CPD step. Shrinkage could be slowed by protecting the aerogel from humidity through hydrophobization.

Freeze Drying using a Lyophilizer

Finding that the process of CPD was very slow and constraining due to the lower accessibility of the adequate machinery needed for the process. We acquired access to lyophilizing machines to produce the aerogels more easily. Several iterations and refinements of the procedure were needed to achieve a convincing material that remained intact. This was because in preparation for Freeze Drying ice crystals can form in the samples causing cracks as these crystals cause the pores of the aerogel to expand.

Figure 2Freeze Dried (FD) aerogels with different cellulose concentration 150 mL samples.(A) Cracked sample having 1X content of cellulose. (B) Homogenous sample having 1.5X content of cellulose. (C) Same sample as B after only one day of the FD.

The three samples above (fig 2) were made using a 150 g dispersed solution with 12.5 g NaOH, 6.25 g Thiourea and with the content of cellulose powder scaled according to the annotations in (Table. 1). The initial tests (fig 2.A) had several issues that we managed to troubleshoot. The major improvement was achieved by adding a higher concentration of cellulose (fig 2.B) for a firmer structure. The improvement was supplemented by optimising the process of gelation and transportation of the sample prior to freeze drying

The final result was a convincing intact aerogel, however we saw that after several hours the samples turned orange and there was an unpleasant slight smell in the samples. After several tests and discussions, we concluded that the remaining salts in the material were the reason for the discoloration of the material.

Those salts are or chemicals needed to be removed.

Solvent Exchange before Freeze Drying

In the final iteration of creating our insulative material (fig 3), we conducted several tests of solvent exchange using both bi-distilled water and ethanol 70 %. We rapidly discovered that combining both solvent exchange with freeze drying yielded the best results (fig 3. A,B). Influenced by the type of insulative material we desired and its application, we tested various combinations of constituents (Table 1, fig3). After comparison we settled upon the approach used to generate the second sample (fig 3. B). While the 3A material was lighter it was less robust. In contrast the 3B sample was a more robust and condensed material. The solvent exchange was a crucial step to purify the aerogel and permit avoid residual Thiourea and NaOH at the end product. This was confirmed by the lack of discoloration in any of the four solvent exchanged samples (fig 3).

Figure 31.25x cellulose concentration aerogels with different solvent exchange processes 14-20 mL.(A) Aerogel sample manufactured with bi-distilled water, then twice ethanol 70% baths and finally water solvent exchange. (B) Aerogel sample manufactured with twice ethanol 70% baths and finally water solvent exchange. (C) Same process as sample A, however longer bi-distilled water bath. (D) Aerogel sample manufactured with only water solvent exchange.
Density and Porosity Consideration

Having produced externally structurally convincing samples, we next conducted the initial tests of density and porosity. The method of measurement is elaborated upon in the Measurement page.

Table 2Density and porosity assessment of aerogel samples.
Aerogels Comments/Appearance Density (g/cm3) Porosity(%)
1.5X tube sample 50 ml without solvent exchange (batch 8) After FD the appearance of the aerogel was white, however the colour then rapidly changed. 0.198 64.59
1.5X sample, water solvent exchange 50 ml (batch 11) The sample lost some structure of the aerogel after FD. We identified visible volatile cellulose parts in the structure. 0.181 67.63
1.25 X ethanol solvent exchange then water (fig3. B) The sample maintained its aerogel structure. more firm than the other samples of aerogel. 0.184 66.84
1.25 X water , 70% ethanol then water (fig3. A) The sample appeared to keep its structure, probably the ethanol baths gave it a better structure 0.149 73.38

Considering the above values of (Table 2) and the values from the article of Z. Ahsan and all1, a porosity ranging between 64.59 and 73.38 % was reasonable, considering the non-specialized process in which we produced the project samples. Moreover, the density values also seemed to be reasonable, within the limits to our ‘non-industrial’ protocol. Deviations were predicted due to the fact that during lyophilisation the frozen water expands and this can expand pore size. But in sum, we managed to produce a ‘good enough’ if not perfect cellulose insulative aerogel sample that could transition to the proof-of-concept addition of protein components, fulfilling the sustainable aims we aspired to in our project.

Visualisation of our samples through Microscopy and Tomography Imaging

For the assessment of the aerogel samples, we had access to different characterisation methods, made available after training at the appropriate the EPFL facilities (fig 4). A very useful way to characterise the material was to explore its inner structure through X-Ray tomography imaging.

Figure 4Microscopy imaging of small aerogel cylinders.(A) 3D stitching image of sample with ethonal 70% solvent exchange. (B) 3D stitching image of sample without solvent exchange after some days from FD. (C) Scanning Electron Microscope (SEM) image of sample A. (D) SEM image of sample B .

The imaging of our samples helped confirm that the desired aerogel structure was preserved by our solvent exchange method. We identified clearly in (fig 4.A,C) a more homogenous porous structure in solvent exchange generated samples. This contrasted to samples when this method was not used (fig 4.B,D), which produced a ‘feathered structure’ with rigid macro cracks.

Interactive Tests

It was necessary to test the material to prove that it had the adequate properties as an insulative material. We explored both mechanical and thermal measurements, with the steps detailed in the measurements page.

To measure mechanical properties we used the Zwick-Roel instrument with a head capable of exerting 5kN, to conduct compression tests on three samples mentioned in notebook batch 10. We identified from (fig.5) that the measurements were coherent since the stress is higher for 1.75X than 1.5X and afterwards 1X, when the same strain was considered. When we compared the stress curve (fig.5) with the literature2, we found a similar curve which indicated that our samples behaved like foams, an expected result from a porous structure.

Figure 5Mechanical compression results with different cellulose concentration gradients.

For the thermal conductivity tests, we constructed an in house testbench to find out the thermal insulative properties of our samples. The apparatus was composed of two chambers isolated from each other with thick insulation panels. We placed in a cavity between both chambers the thinner porous material to assess its thermal conductivity. We heated one of the chambers and used two different sensors and a heat flux.

The thermal conductivity tests were performed with an in house testbench to find out the thermal insulative properties of the aerogel. Within the limits of our in-house apparatus thermal conductivity assay, we showed that the produced macro-porous aerogel had comparable insulative properties to a EPS.

Figure 6In-house Tesbench: Temperature values for a couple of hours.

The porous aerogel curve for the increase of heat in the non heated second chamber is comparable to the EPS curve(fig.6) suggesting our produced cellulose aerogel is an insulation material that has reduced heat conductivity.


Over the course of the manufacturing phase of our macro-porous cellulose aerogel material, we developed an efficient and easy way of benchtop production. If all the steps are undertaken correctly, especially the solvent exchange before the drying process to remove the NaOH and Thiourea salts, the resulting material should be nearly pure cellulose, and therefore completely biodegradable.

The final iteratively optimised process gave us a large amount of aerogel samples for downstream tests and augmentation. Better quality aerogel samples could conceivably be produced bacterial cellulose which has properties that enable finer quality cellulose aerogel samples. Another future improvement suggestion came from EMPA, who suggested to use the TEMPO Oxidation process3 coupled with Critical Point Drying to achieve samples of even higher quality.


Starting from the bacterial transformation and landing with the protein purification, we illustrate in this section the main results from the production of the engineered proteins used for the modular coating of the aerogel.

To address the water sensitivity of the aerogel, we decided to apply first a hydrophobic coating made of a recombinant silk biofilm inspired by green lacewing silk4,5. Afterwards, we decided to link a second layer of proteins with interesting properties to the silk biofilm. The linkage between the silk biofilm and secondary protein coating is mediated by biotin-streptavidin interaction6. For the scope of our project, we chose a fire-retardant, SRSF1 protein (SR in further text), already tested by the Mingdao 2015 iGEM team. However, we could've imagined any other protein of interest being used for secondary coating instead of the SR, making the whole material modulable to the objective. To obtain a visual control for the coating of the aerogel, we engineered a GFP fusion protein following the same construction as for the silk fusion protein.

Proteins Production

Ni-NTA Beads purification method leads to successful purification of the His-tagged recombinant proteins from BL21(DE3) bacterial strain.

To produce the fusion proteins in needed amounts, we did several trials of protein expression by changing some of the parameters of the usual set of experiment for protein production (please find the protocols here):

  • bacterial transformation

  • liquid culture growth,

  • IPTG induction (and in parallel DNA purification and Sanger sequencing)

  • protein purification, visualisation and concentration.

Firsts trials and sequencing results

The first purifications were done using MagneHis Protein purification kit from Promega. However, due to a 50 kDa contaminant protein, we were unable to detect our proteins of interest by SDS-PAGE. The contaminant 50 kDa protein was competing with our proteins of interest, making it infeasible to purify our proteins in sufficient amounts. Upon troubleshooting to understand the source of the contaminant protein, we concluded that it came from the BL21(DE3)pLysS bacterial strain. Therefore, we decided to change to the BL21(DE3) strain. Since at the same time we tried the Ni-NTA beads purification method which gave better results than the MagnHis kit, we performed further protein purifications using Ni-NTA beads at PTPSP. Nevertheless, for eluting all our proteins we had to use rather high concentration of imidazole. Upon examining our initial gene constructs for proteins of interest by performing Sanger sequencing, we noticed an additional sequence on the 5’ terminus of all our constructs. The additional sequence was composed of a thrombin site, a T7 tag and an additional 6x His-tag (fig 7.B, D & F), which added 102 nt to our initially designed constructs(fig 7.A, C & E). Due to the two His-tags, one on the N-terminus and one on the C-terminus of the recombinant proteins, all of the proteins showed high affinity to Ni-NTA beads and required high imidazole concentrations for the elutions.

Figure 7Plasmid Maps coding for the protein constructs made with SnapGene. Initially designed plasmid maps for the (A) silk fusion protein (01a),(C) SR fusion protein (01b), (E) GFP fusion protein (03a). Obtained plasmid maps after Sanger sequencing for (B) silk fusion protein (01a), (D) SR fusion protein (01b) and(F) GFP fusion protein (03a).

Silk Fusion Protein

Protein purification of the silk fusion protein (01a) showed a 100kDa protein present in the first elution lane (E1) in both the SDS-PAGE (fig 8.A) and the Western Blot (fig 8.B). Silk fusion protein was expected at 78kDa size. Using the nanodrop, we measured the concentration (0.72 mg/mL) and we obtained a final amount of 2.15 mg of the silk fusion protein (01a).

Figure 8Protein purification of silk fusion protein (01a)Both gels were run at 180V for 30 minutes. (A) SDS-PAGE gel stained with Coomassie Blue Protein Stain of all the fractions of silk fusion protein purification. (B) Western Blot of the elution 1 (E1 fraction on SDS-PAGE gel) visualised with anti-His antibody.W1 = Wash 1 with 20 mM imidazole; W2 = Wash 2 with 50 mM imidazole; E1 = Elution 1 with 250 mM imidazole; E2 = Elution 2 with 500 mM imidazole; E3 = Elution 3 with 1M imidazole; E4 = 2.5 M imidazole; E5 = 5 M imidazole.

SR Fusion Protein

For the SR fusion protein (01b) we observed a protein around 40 kDa in both quality controls (fig 9) which matches the expected size (36.4 kDa). Most of the protein was in elution 4 (E4 on SDS-PAGE gel) fraction (fig 9.A and B), which required very high imidazole concentration (2.5M). However, we still obtained a final concentration of 0.17 mg/mL for a total amount of 1.02 mg of the SR fusion protein (01b).

Figure 9Protein purification of SR fusion protein (01b).(A) SDS-PAGE gel stained with Coomassie Blue Protein Stain. (B) Western Blot ran with the sample of elution 1 and analysed by anti-His antibody.W1 = Wash 1 with 20 mM imidazole; W2 = Wash 2 with 50 mM imidazole; E1 = Elution 1 with 250 mM imidazole; E2 = Elution 2 with 500 mM imidazole; E3 = Elution 3 with 1M imidazole; E4 = 2.5 M imidazole; E5 = 5 M imidazole.

GFP Control Fusion Protein

On both the SDS-PAGE and the Western Blot for the protein purification of the GFP fusion protein (03a) (fig 10), we observed a band around 70kDa which matched the expected size of our protein (71.4 kDa). Purified GFP fusion protein (03a) was present in all eluted fractions. To concentrate the proteins, we used a 30 kDa filter. However, the filter was made of cellulose, so we observed the filter turning green. This showed the efficiency of the CBD domain of our construct to bind cellulose. We finally obtained a concentration of 0.35 mg/mL, which corresponds to 4.5 mg of GFP fusion proteins.

Figure 10Protein purification of the GFP fusion protein (03a). Both gels were run at 180V for 30 minutes. (A) SDS-PAGE gel stained with Coomassie Blue Protein Stain of all the fractions of GFP fusion protein purification. (B) Western Blot of the elution 1 fraction (E1 on SDS-PAGE) visualised with anti-His antibody.W1 = Wash 1 with 20 mM imidazole; W2 = Wash 2 with 50 mM imidazole; E1 = Elution 1 with 250 mM imidazole; E2 = Elution 2 with 500 mM imidazole; E3 = Elution 3 with 1M imidazole; E4 = 2.5 M imidazole; E5 = 5 M imidazole.

Since high imidazole concentration might denature the proteins, we had to either remove it or make imidazole inert. It was not possible to dialyse the proteins to remove imidazole, since the dialysis membrane was made of cellulose, and our fusion proteins would all bind to the membrane via their cellulose binding domain (CBD). We therefore decided to flash freeze the proteins and store them in the elution buffers so the imidazole will no longer affect them when frozen.


Cloning experiments lead to successful removal of the added site from our plasmids.

The elution of the proteins of interest in such high amounts of imidazole certainly comes from the double His-tag that composes our proteins. To remove it from the received plasmids, we performed some cloning experiments to obtain better yields in the future purifications.

By PCR And KLD For GFP And SR Fusion Proteins

In the case of the 03a plasmid (containing GFP) and the 01b plasmid (containing SR), we performed KLD cloning(fig 11). We amplified the plasmids by PCR without the undesired sequence, and we re-ligated them back by doing a KLD. The KLD reaction allows efficient phosphorylation, intramolecular ligation and template removal in a single 5-minute reaction step at room temperature. On the agarose gel we ran with the PCR products (fig 11.A), we observed bands around 7000 bp which are the expected sizes and means that we successfully removed the added site. After the KLD, we ran a second agarose gel with fragments obtained by cutting the new plasmids with different restriction enzymes (fig 11.B). We obtained similar patterns as the ones expected. By purifying and sequencing the plasmids(fig 11.C and D), we obtained the exact same sequence as designed. This confirmed that we removed the added site and re-ligated the DNA to obtain the good plasmids.

Figure 11Cloning experiment results for removal of the added site for 01b (SR) and 03a (GFP).(A) Agarose gel electrophoresis of PCR products of the 01b and the 03a constructs plasmid amplified without the GeneScript additional tags for KLD cloning. We used a 1% agarose gel stained with SYBR Safe DNA Gel Stain. (B) Agarose gel electrophoresis for restriction analysis of the 01b and the 03a constructs plasmid after the ligation by KLD. We used a 1% agarose gel stained with SYBR Safe DNA Gel Stain. (C) Plasmid map from the sequencing result of the obtained new SR plasmid. (D) Plasmid map from the sequencing result of the obtained new GFP plasmid.

With several minipreps we obtained big quantities of these new plasmids. We therefore transformed new BL21(DE3) E.Coli competent cells to start the protein production.

For the GFP fusion protein, the pellet after the first centrifugation step was not green, which means that there was no expression of our protein of interest. In a similar way, we couldn’t see a significant protein expression after the IPTG induction for the SR fusion protein. Even if the purification was performed, nothing would be visible in the elution lanes of the SDS-PAGE which suggests that we didn’t successfully purify our new SR fusion protein.

The cloning with PCR amplification and KLD for ligation was a success, but the bacteria were not able to produce our fusion proteins. Our hypothesis is that the number of bp between the RBS and the first Methionine was not optimal for the bacteria. However, by lack of time, we decided not to troubleshoot the reasons for this failure, and we concentrated on the protein production to use them in our further experiments.

By Digestion-Ligation for the Silk Fusion Protein

In the case of the silk fusion protein, because of its repetitive modules the previous experiment failed. To remove the added part, we therefore digested our plasmid (fig 12.A) with the NcoI enzyme since there are two NcoI restriction sites on both sides of the added site (fig 12.B) and re-ligated it back (fig 12.C).

Figure 12Cloning experiment results for the removal of the added site for 01a (silk).(A) Agarose gel electrophoresis of the digested product of 01a. We used a 1% agarose gel stained with SYBR Safe DNA Gel Stain. (B) Plasmid map of 01a (silk) from with the added site of GeneScript. (C) Plasmid map from the sequencing result of the obtained new 01a plasmid (silk).

Since we obtained the wanted plasmid, we transformed new BL21(DE3) competent cells to start a new protein production. By doing a gel comparing the protein expression before and after IPTG induction (fig 13.A), we could see a band at approximately 100 kDa. Protein purification of the silk fusion protein (01a) also showed a 100kDa protein present in the two elution lanes (E1 and E2)(fig 13.B). Even though we expected the protein to be around 74kDa, this 100kDa lane should correspond to our new silk fusion protein, since we obtained similar one in the previous purification. By measuring the concentration (0.37 mg/mL), we obtained a final amount of 11mg of the silk fusion protein (01a).

Figure 13Expression of the silk fusion protein from the plasmid without the added site.(A) SDS-PAGE gels run with the lysate before and after IPTG induction to see if the protein is well expressed. The gel was run at 180V for 30 minutes and stained with Coomassie Blue Protein stain. (B) SDS-PAGE gel ran with samples from protein purification of the new silk fusion protein induced in BL21(DE3). The gel was run at 180V for 30 minutes and stained with Coomassie Blue Protein stain. SN = Supernatant; FT = Flowthrough; W1 = Wash 1 with 20 mM imidazole; W2 = Wash 2 with 50 mM imidazole; E1 = Elution 1 with 250 mM imidazole; E2 = Elution 2 with 500 mM imidazole.

CBD Characterization

CBD-containing GFP fusion protein has a higher affinity for cellulose than a CBD-free GFP.

After the production of our proteins, we decided to test the binding of our Cellulose Binding Domain to cellulose. We therefore added drops of our mSA-GFP-CBD protein or of a GFP that doesn’t contain CBD on a nitrocellulose membrane (fig 14A). We then performed several washes using PBS to wash out the proteins that were not bound to the membrane. Our mSA-GFP-CBD protein was more fluorescent than the GFP control one, even if the concentrations were the same. After the third PBS wash (fig 14B), we observed a diminution in the intensity of the fluorescence of the control GFP, while no change in intensity for the mSA-GFP-CBD. Using ImageJ, we quantified the difference in fluorescence intensity between before (fig 14A) and after (fig 14B) the PBS wash. In the case of our mSA-GFP-CBD protein, the mean for the difference in fluorescence intensity is smaller than for the control GFP (fig 14C). With this quantification, we concluded that our protein mSA-GFP-CBD has a higher affinity with the nitrocellulose membrane than the GFP control protein. We also tried to wash the proteins by using solutions with decreasing pH values. However, since high pH-steps were used in the pH-washes, all the proteins were gone when they were exposed to an acidic solution of pH3.

Figure 14Comparisons of fluorescence intensity of control GFP and our GFP fusion protein (mSA-GFP-CBD) on nitrocellulose.(A) sfGFP control protein exposed to 2 vs 3 washes of PBS. (B) GFP fusion protein (mSA-GFP-CBD) exposed to 2 vs 3 washes of PBS. (C) BoxPlot of the difference of fluorescence intensity before and after the PBS wash for the two samples. Top of the boxes = mean of fluorescence difference per group; Black dots = individual measurements; Red vertical line = standard error of the mean. Data are mean ± s.d., n = 5 measurements per group.

Silk Biofilm

We successfully made a silk fusion protein biofilm, and we have shown that it provides hydrophobic properties.
Production of silk biofilm

After the production of our proteins, we created a silk fusion protein biofilm to coat an aerogel to make it water-resistant. Inspired by the work of Felix Bauer5, we produced two biofilms : one with the recombinant silk protein (fig 15.A) and one with the recombinant GFP protein, as a control (fig 15.B). In figure 15, we can see that we dried the biofilms too much due to the presence of cracks. Even if it was not a problem for hydrophobicity testing, it became one to remove the biofilm from the dish. Indeed, it was impossible for us to take the biofilm off, even by cutting the edges of the petri dish. This meant that we couldn’t coat aerogels with those biofilms. However, we saw that the GFP biofilm (fig 15.B) made some crystals and cubic structures so that the intended biofilm was less homogeneous than the silk one (fig 15.A). This result indicated that it is not possible to make a biofilm from any proteins, and in the case of the GFP fusion protein (03a), we didn’t obtain a proper one. Overall, we managed to generate a biofilm with our recombinant silk fusion protein (mSA-silk-CBD), but not with the GFP fusion protein (mSA-GFP-CBD).

Figure 15Silk biofilms inside polystyrene petri dishes.(A) Silk fusion protein biofilm, or (B) GFP fusion protein biofilm in a polystyrene petri dish after drying under the chemical fume hood for 24 hours.

Hydrophobicity Test

With these biofilms, we did a wettability test to compare their hydrophobicity. To do so, we added droplets of water on top of those biofilms and did several measurements. First, we measured the angle made by the droplet (fig 16.A and B) to assess the surface tension and the surface wettability. Then we measured the time it took for the water to be absorbed or evaporated, computed the mean values for each biofilm and compared them to conclude about hydrophobicity. For the first measures, we obtained an angle of 55.2° for the silk (fig 16.C) and 27.3° for the GFP biofilm (fig 16.D). This meant that the silk fusion protein biofilm (01a) had a higher contact angle than the GFP biofilm. Moreover, by measuring the time it takes for three water droplets to be absorbed, we obtained a mean of 466,33 s for the silk biofilm (fig 16.E), and 28,33 s for the GFP biofilm (fig 16.F). With these values, we concluded that it takes more time for a drop of water to be evaporated or absorbed from the silk biofilm. These two results suggested that our biofilm helps protect the aerogel from water droplets.

Figure 16Silk biofilms with water droplets for hydrophobicity tests.(A) Silk fusion protein biofilm (01a) with a drop of water (B) GFP fusion protein biofilm (01a) with a drop of water.(C) Measurement of the angle made by the drop on silk biofilm. (D) Measurement of the angle made by the drop on GFP biofilm. (E) Dried drops of water on the silk biofilm (F) Dried drops of water on the GFP biofilm.


After several trials of protein expression and purification, we optimised our protocols by using different bacterial strains and by purifying the protein with a Ni-NTA affinity column. By doing so, we reduced the number of contaminant proteins but had to elute our protein in high imidazole concentration. As we had a plasmid with two His-tag, we thought that removing one of them would improve the efficiency of the purification. We therefore deleted this site by doing several cloning experiments which were a success. However, our bacteria were not able to express all these new proteins, and we had to come back to our initial plasmid to produce them in bigger quantities. With them, we tested the affinity of the Cellulose Binding Domain with cellulose to prove that the silk fusion protein as well as the GFP fusion protein were able to bind the cellulose aerogel. Moreover, we produced a silk biofilm and concluded that it was water resistant.

Coating of The Aerogel

After the finding of the optimal concentration of cellulose in the aerogel, and the production of all of our recombinant proteins, we present in this section the main results for the coating of the aerogel. Since the aerogel is water sensitive, we decided to coat it with a layer of our silk fusion proteins that we have proven to bind cellulose and to provide water resistance. We used the GFP fusion protein as a visual control for this binding, because it follows the same construction as the silk fusion protein. Thanks to the the modeling part, we found the number of molecules needed to completely coat a 30 mm diameter and 3 mm thickness aerogel. With it we calculated that we had 7.64 x 10-10 mol of proteins for one such coating.

With GFP Fusion Protein

We successfully coated an hydrogel with our GFP fusion protein and freeze dried it to make an aerogel coated with this protein.
Fluorescent Aerogel

With the production of the GFP fusion protein, we coated a hydrogel of 30 mm diameter and 3 mm thickness with 7.64 x 10-10 mol of proteins. We used as a positive control a GFP that doesn’t contain the CBD domain. We then sent the coated hydrogels to the freeze drier to produce aerogels. As we can see on fig 17A, the aerogels were green under the fluorescence microscope which meant that the proteins bound the hydrogel and stayed on it during the freeze drying process. Even if we observed fluorescence for the three samples (water, control GFP and our recombinant GFP), the aerogel with mSA-GFP-CBD showed the most intense fluorescence. It showed that after the freeze-drying step, our mSA-GFP-CBD binds more to cellulose than usual GFP. The comparison of the fluorescence intensity between (fig 17.A) and (fig 17.B) showed that it has decreased with time. However, our mSA-GFP-CBD still showed the most intense fluorescence. This proves that our mSA-GFP-CBD bound tighter to cellulose thanks to its CBD domain compared to the usual GFP.

Figure 17Fluorescence gel imager images of the aerogels after the freeze-drying step produced from the soaked hydrogels.(A) Aerogels produced from hydrogels previously soaked with double-distilled water (left), GFP (middle and mSA-GFP-CBD (right). (B) Aerogels produced from hydrogels previously soaked with double-distilled water (left), GFP (middle) and mSA-GFP-CBD (right) five days after the freeze-drying step.

With Silk Fusion Protein

We have successfully coated the aerogels with our silk fusion protein and have shown that this coating adds hydrophobicity to the aerogel.

Silk protective coating

Now that we have proven that our construct binds to the hydrogel, and remains bound to the aerogel after the freeze drying step, we showed that coating the aerogel with the silk fusion protein makes it more water resistant than without the coating. We first coated hydrogels with proteins (either the silk fusion protein (01a), the GFP fusion protein (03a), the GFP control protein or water) (fig 18.A) and sent them to freeze dry (fig 18.B).

Figure 18Freeze drying of coated hydrogels.(A) Hydrogels before freeze drying. (B) Aerogel after 23 hours of freeze drying.

After this step, we saw that the coated aerogels had a more homogeneous shape, with less cracks and a sort of protective film on top of it compared to the uncoated ones (fig 19.A). Using the Keyence microscope, we filmed the application of 4 to 5 drops of water on the aerogels and quantified the water absorption time. For each aerogel we computed the mean delay time and used this data to perform a one-way ANOVA test to see if there is a significant difference between the means between the groups. We obtained a final p-value of 0.000576. As the p-value was less than the significance level 0.05, we concluded that there are significant differences between the groups. However, this result did not indicate which pairs of groups are different. As the ANOVA test was significant, we computed a Tukey HSD for performing multiple pairwise-comparisons between the means of groups. For each comparison with the silk fusion protein (01a), we obtained a p-value way smaller than 0.05, which means that there was a significant difference between the silk fusion protein and the other groups, while there was no significant difference among the other groups (dH20, GFP and GFP fusion protein 03a). Therefore we concluded that aerogels absorb water more slowly when they are coated with the silk fusion protein. Indeed, in the box plot we plotted to present the different populations (fig 19.B), it is clear that in the case of the silk fusion protein (01a), the delay times were higher. The water drop took longer to be fully absorbed by the aerogel coated with the recombinant silk, suggesting that the recombinant silk adds hydrophobicity to the aerogel. Overall, these results proved that the CBD domain we added to our silk construct allows binding to the aerogel and that this protein provides a protective coating for the aerogel.

Figure 19Protective coating given by the silk fusion protein layer on the aerogel.(A) Comparison between uncoated (top) and recombinant silk fusion protein (01a) coated (bottom) aerogels. (B) The delay time of the absorption of a drop of water to the aerogel coated with the different proteins. Top of the boxes = mean of delay time per group; Black dots = individual delay time measurements; Red vertical line = standard error of the mean. Data are mean ± s.d., n = 3 measurements per group; ***P < 0.001, **P < 0.01; one way ANOVA test was performed followed by a Tukey HSD for performing multiple pairwise-comparisons between the means of groups.

Experimental Continuation

Due to time restrictions, we were not able to troubleshoot each experiment. We therefore focused on the production of proteins and aerogels, in order to produce a recombinant silk biofilm, to proof the binding between our constructs containing the Cellulose Binding Domain (CBD) and aerogels, and to show the binding between the silk biofilm and the modular coating via biotin-streptavidin interaction.

To improve the protein production, we would have redone the cloning experiments to remove the added site in the plasmids, to better purify the proteins. Moreover, with more time, we would have improved the evaporation time in the biofilm production and troubleshoot to find a way to reduce the biofilm’s stickiness on the petri dish. In addition, to properly prove the hydrophobicity of the silk biofilm, the measurements would need to be redone in a statistical manner by increasing the number of samples to compute a statistical test. Finally, we tried the biotinylation of the SR fusion protein and then the binding between the SR and the silk. The results were not very conclusive, since we were not able to prove the binding between our two proteins of interest via iSCAT. With more time, we would have troubleshoot this experience. However, since the biotin-streptavidin is one of the strongest known non-covalent interactions7, if someone manages to purify and biotinylate the proteins, they should be able to prove their binding.


Table 3Achievements of our experiments
Success Misadventures
Training PTPSP First protein production for 01a
Third protein production for 01a Second protein production for 01a
Cloning of 01a: digestion-ligation First protein production for 01b
Upscaling for 01a Second protein production for 01b
First and second trial for silk biofilm Protein production with new 01b
Hydrophobicity Test for the biofilm Biotinylation and link between 01a and 01b
Third protein production for 01b First protein production for 03a
Upscaling for 01b Second protein production for 03a
Third protein production for 03a Protein production with new 03a
Upscaling for 03a
Troubleshooting for contaminant protein
Cloning of 01b and 03a
CBD characterization
Coating of hydrogels


  1. Zaman, Huang, Jiang, Wei & Zhou (2020)
    Preparation, Properties, and Applications of Natural Cellulosic Aerogels: A Review
    Energy and Built Environment, vol. 1, no. 1, pp. 60-76
  2. Buchtová, Pradille, Bouvard & Budtova (2019)
    Mechanical properties of cellulose aerogels and cryogels
    Soft Matter, vol. 15, no. 39, pp. 7901-7908
  3. Wang, Shao, Zhou, Yang & Xu (2017)
    Preparation of carbon aerogels from TEMPO-oxidized cellulose nanofibers for organic solvents absorption
    RSC Advances, vol. 7, no. 61, pp. 38220-38230
  4. Bauer & Scheibel (2012)
    Artificial Egg Stalks Made of a Recombinantly Produced Lacewing Silk Protein
    Angewandte Chemie International Edition, vol. 51, no. 26, pp. 6521-6524
  5. Felix Bauer (2013)
    Development of an artificial silk protein on the basis of a lacewing egg stalk protein
  6. Jansson, Thatikonda, Lindberg, Rising, Johansson, Nygren & Hedhammar (2014)
    Recombinant Spider Silk Genetically Functionalized with Affinity Domains
    Biomacromolecules, vol. 15, no. 5, pp. 1696-1706
  7. Weber, Ohlendorf, Wendoloski & Salemme (1989)
    Structural Origins of High-Affinity Biotin Binding to Streptavidin
    Science, vol. 243, no. 4887, pp. 85-88