Hydroxypropylcellulose stearate (HPCE, M_w 134,700 g mol⁻¹, M_n 107,400 g mol⁻¹, polydispersity index 1.25, degree of substitution (DS) 3.0) was synthesized according to a literature protocol (Nau et al., 2018). Trimethylsilyl cellulose (TMSC, From Avicel, M_w 185,000 g mol⁻¹, M_n 30,400 g mol⁻¹, polydispersity index 6.1, DS 2.8) was purchased from TITK (Thuringian Institute of Textile and Plastics Research, Germany). The structures are shown in Figure 1. Chloroform (99.3%), disodium phosphate heptahydrate (Na₂HPO₄·7H₂O), sodium dihydrogen phosphate monohydrate (NaH₂PO₄·H₂O), hydrochloric acid (37%), and bovine serum albumin (BSA) were purchased from Sigma-Aldrich and were used as received. Silicon wafers used as film substrates were cut 1 cm × 2 cm. Surface plasmon resonance (SPR) gold sensor slides (CEN102 AU) were purchased from Cenibra (Germany). Milli-Q water (resistivity = 18.2 Ω⁻¹ cm⁻¹ at 25°C) from a Millipore water purification system (Millipore, U.S.A.) was used for contact angle, SPR, and QCM-D investigations.

The film substrates were cleaned by immersing them in an in-situ produced peroxymonosulfuric acid containing H₂O₂ (30 wt%)/ H₂SO₄ (1:3 v/v) for 10 min for SPR spectroscopy slides or 30 min for silicon wafers, respectively. After rinsing with deionized water, the wafers were dried with nitrogen gas, rinsed and stored in deionized water. TMSC and HPCE were dissolved in chloroform in a concentration of 0.75 wt%, using a water bath heated to 30°C, and 120 h on a magnetic stirrer. Right before use the solutions were filtered through 0.45 μm PVDF filters (Chromafil) and mixed in volumetric ratios labeled further on as TMSC:HPCE 1:0, 1:3, 1:1; 3:1, and 0:1. A volume of 100 μl was used for spin coating and operated for 60 s with an acceleration of 2,500 rpm s⁻¹ and a speed of 4,000 rpm.

The conversion of TMSC into cellulose was implemented in a polystyrene petri dish (5 cm in diameter) containing 3 ml of 10 wt% HCl. The substrates were exposed to HCl vapor for 12 min. The regeneration of cellulose from TMSC was verified by contact angle and ATR-IR measurements (Alpha FT-IR spectrometer, Bruker, U.S.A.) using an attenuated total reflection attachment and obtaining spectra between 4,000 and 400 cm⁻¹ with 48 scans and a resolution of 4 cm⁻¹. The data was analyzed with OPUS 4.0 software.

Thickness of the thin films was determined by scratching the films with a scalpel and measuring the profile of a scan length of 1,000 μm and a duration of 3 s using a DETAK 150 Stylus Profiler from Veeco (Bruker, USA) on a hydraulic balanced stone table with a diamond stylus with a radius of 12.5 μm and a force of 3 mg. Three films of each sample were measured at 6 locations before and after regeneration. Film thickness and film roughness was calculated from the resulting profile using Software Vision 64.

Static contact angle measurements were performed with a Drop Shape Analysis System DSA100 (Krüss GmbH, Germany) with a T1E CCD video camera (25 frames per second) and the DSA1 v 1.90 software. All measurements were performed at least three times on minimum two manufactured films with Milli-Q water and diiodomethane using a droplet size of 3 μL and a dispense rate of 400 μL min⁻¹. Static CAs were calculated with the Young-Laplace equation, and the SFE was determined with the Owen-Wendt-Rabel-Kaelble (OWRK) method. Surface tension of 50.80 and 72.80 mN m⁻¹ for diidomethane and water were used, respectively.

A phosphate buffer containing 8.1 mM disodium phosphate, 1.9 mM sodium phosphate and 100 mM sodium chloride at pH 7.4 was used to carry out the adsorption experiments of BSA on the films. Surface Plasmon Resonance Spectroscopy (SPR) was performed with a MP-SPR Navi 200 from Bionavis Ltd (Finland), using 785 nm laser in both measurement channels. The attached autosampler MP-SPR Navi 210A was set to 20 μl min⁻¹ flow rate. The equilibration of the thin films was observed by measuring the spectra with full angular scan (39–78°) and scan speed of 8° s⁻¹ at 24.5° and plot the SPR-angle over time. A concentration of 1.0 mg ml⁻¹ of BSA was dissolved in the buffer and exposed to the thin films for 10 min. Adsorbed mass (Γ) was calculated with the de Feijter equation,

using the refractive index increment (dn/dc) 0.182 cm3 g⁻¹. The ΔΘ is the angular response of the surface plasmon resonance. For thin layers (<100 nm), k × d_p can be considered constant and can be obtained by calibration of the instrument by determination of the decay wavelength l_d. Here it was 1.09 × 10⁻⁷ cm/° (at 670 nm) and 1.9 × 10⁻⁷ cm/° (at 785 nm) in aqueous systems.

Quartz Crystal Microbalance and Dissipation (QCM-D) instrument (model E4) from Q-Sense (Sweden) was used with gold sensors purchased from QuartzPro (Sweden). The attached peristaltic pump was set to 0.1 ml min⁻¹. Adsorption was performed in the same conditions as the SPR analyses. The data was analyzed using Johannsmann modeling (Johannsmann et al., 1992; Naderi and Claesson, 2006).

Most AFM and all FFM measurements were acquired using an Asylum Research MFP-3D AFM (USA). The instrument is equipped with a closed-loop planar x-y-scanner with a scanning range of 85 μm × 85 μm and a z-range of 15 μm. The tapping mode AFM images were recorded with standard silicon probes (Olympus AC160TS, cantilever spring constant ~30 N m⁻¹, tip radius ~10 nm). The measurements were obtained in ambient conditions at 50 ± 8% relative humidity and a temperature of 22 ± 1°C. Topography images of three independent positions were recorded for each sample. All the data was processed in the open-source software Gwyddion (Necas and Klapetek, 2012). For the 5 μm × 5 μm images, a roughness analysis (Teichert, 2002) was performed by calculating the 1D height-height correlation function:

of each scan line and then averaging over all lines. The resulting values were fitted with the function:

The parameters σ, ξ, and α are used to characterize the surface roughness (Teichert, 2002). The σ denotes the root mean square (RMS) roughness, i.e., the standard deviation of the height values, which is a common measure for the vertical roughness. The lateral correlation length ξ describes the lateral fluctuation of the height values and α is the so-called Hurst parameter or roughness exponent. It determines the shape of C(x) and quantifies the jaggedness of the surface.

For FFM, which is recorded in contact mode, NT-MDT CSG10/Au probes with a tip radius of about 30 nm and a low cantilever spring constant of 0.1 N m⁻¹ were employed. Images with frame size of 5 μm × 5 μm were obtained with a constant scan speed of 2.5 μm s⁻¹. A vertical force of about 10 nN was applied during the measurements. For acquisition of an FFM image, standard contact mode AFM scan including the lateral trace and retrace channel were recorded. The raw lateral signals were converted to friction images by subtracting the lateral retrace from the lateral trace signal and dividing it by two for each image (Kalihari et al., 2010; Shen et al., 2014). This eliminates topography artifacts and a possible offset.

The quantitative interpretation in terms of friction coefficient is not straightforward, and literature includes several calibration methods (Klapetek, 2018). Lateral force sensitivity calibration was done here according to the wedge calibration method of Varenberg et al. (2003). Quantitative friction images were obtained by multiplying the resulting friction image data with the lateral force sensitivity using the Gwyddion software.

AFM force spectroscopy measurements to investigate the adhesion properties of the film surfaces were performed with a scan rate of 2 Hz and a force distance of 0.5 μm. For these measurements, HSC60 probes from Team Nanotec (Germany) were used which have a cantilever spring constant of about 50 N/m and a tip radius of 60 nm. Here, 32 × 32 px² maps were obtained on 5 μm × 5 μm topography scans.

A Veeco Multimode Quadrax MM AFM (Bruker, USA) in tapping mode using standard silicon probes (NCH-VS1-W, NanoWorld AG) was used for recording film topography after they were rinsed with chloroform.

To quantify the observed surface features for the individual blend films, ten individual cross-sections of the features were obtained from the topography images for each film with the Gwyddion software to determine the height and width of the features. The values are given as mean ± standard deviation.

The AFM topography images were used for calculating effective surface fractions of the blend films. These are referred to as “effective surface fraction” or “surface fraction” later in the manuscript. The film component fraction derived from spinning solution is referred to as “volume fraction” or “volume ratio.” Masked surfaces were evaluated by the surface area estimate method in Gwyddion, computed by simple triangulation that considers heights and spatial relations in the surface. For this purpose, additional points were added in-between four neighboring points using the mean values of these pixels. Four triangles are formed, and the surface area is approximated by summing their areas. Masking was done by threshold in z-value and adjusting with a pen tool. Single noise pixel was removed by grain filtering function. For films that showed a reliable phase contrast in the tapping mode topography analysis, the masks were determined from the phase information.

The contact mode FFM images (Figure 5A) enclose the distinct differences between the neat and the cellulose/HPCE blend films' friction (indicated by the contrast differences). The average friction coefficients (Klunsner et al., 2010) were lower for the blend films than for the pure films (Figure 5B). The same applied for the adhesion forces (Figure 5C) where the blends featured lower values (20–50 nN) than the cellulose and HPCE films (65 ± 2 nN, and 104 ± 5 nN, respectively). It should be noted here that friction force data is in good agreement with those reported for cellulose spheres interacting with modified silica surfaces in a similar applied force range (~5 nN) (Bogdanovic et al., 2001).

Cellulose in general is not very prone to non-specific protein adsorption. This originates from the highly hydrated, hydrophilic cellulosic material, having hydrogel characteristics. Upon protein deposition, the water that is close to the surface of the cellulose and on the surface of the protein needs to be replaced—a process which is, if there are not any specific contributions—entropically unfavorable. Several approaches have shown that either anionic or cationic coatings on cellulose thin films may alter the adsorption behavior of proteins (Orelma et al., 2011; Mohan et al., 2013, 2014) depending on the employed pH during adsorption and the isoelectric point of the protein. On hydrophobic surfaces, proteins adhere spontaneously in a non-specific fashion and often they (partially) lose their ternary or quaternary structure and denaturate during deposition (Norde and Lyklema, 1991; Sagvolden et al., 1998). One would expect that the incorporation of a hydrophobic component such as HPCE into the cellulose film would trigger non-specific adsorption different from affinity to the cellulose regions. We chose BSA (pH 7.4) as a demonstrator probe for non-specific protein interactions of the blend films and studied the adsorption process using a combination of SPR spectroscopy and QCM-D. The combination of these techniques gives complementing insight into the amount of adsorbed protein. While QCM-D is a gravimetric technique capable of sensing any type of mass (i.e., water, and BSA, Γ^QCM) on the surface, SPR spectroscopy allows for determination of dry mass (Γ^SPR) by employing the de Feijter equation (1). The difference between Γ^QCM and Γ^SPR is the amount of water in the film. This amount of water decreased with lowering the cellulose content in the films and reaches its minimum for the 1:3 cellulose:HPCE film (Figure 6A). The HPCE film should swell the least, and hence, the major fraction of water can be assumed to be associated with the BSA on the surface. The amount of water in the BSA layer is 65%. The HPCE film experiences the highest BSA adsorbed amount (based on SPR) of 1 mg m⁻². The other extreme is the neat cellulose film that features low BSA adsorption (Γ^SPR 0.1 mg m⁻²) which is in good agreement with other reports in these conditions (Niegelhell et al., 2017; Weißl et al., 2018). The highly swollen hydrogel character is revealed in the QCM-D data for the pure cellulose film, showing 1.7 mg m⁻² water in the layer corresponding to 96% of hydration.

The increase in volume fraction of the HPCE led to a linear increase in the amount of adsorbed BSA on the surfaces. Contrasting the SPR adsorbed mass to the effective surface fraction revealed that the 1:1 blend film (volumetric) was an outlier in the adsorption trend and a lower amount of protein was deposited than what would have been expected (Figure 6B). Comparing this data to the other few reports on protein adsorption on cellulose based blend films revealed that also for the other systems protein adsorption minima were determined close to the 1:1 volume ratio (Niegelhell et al., 2016, 2017; Strasser et al., 2016). The cellulose:HPCE 3:1 film contains 57% of cellulose based on surface fraction, but still, it shows a higher BSA adsorption (0.39 mg m⁻²) than the cellulose:HPCE 1:1 film (0.30 mg m⁻²) with lower cellulose content (37% surface fraction).

The influence of surface free energy, roughness, and friction coefficient on the BSA adsorption is plotted against Γ,QCM Γ,SPR and bound water (Figures 6C,D). The SFE has a decisive influence on the BSA adsorption (Figure 6C). For the neat cellulose and HPCE films, surface roughness as well as friction coefficient were similar, but the SFE significantly differed. For the blend films however, the correlation between SFE and adsorbed mass was straightforward at first glance: lower surface energy translated to more hydrophobic surface resulting in higher non-specific protein deposition (Figure 6C). However, the SFE development did not correlate linearly with the effective surface fraction (Figure S3). There is a plateau in the SFE development with surface fraction of 0.37 (1:1) after which the cellulose dominated the SFE. One would expect much lower SFEs due to the actual surface fraction of just 37% of cellulose which would lead to higher non-specific protein adsorption. It is therefore rather the wetting properties of the blend films that are tuned by the morphology and configuration and these further affect the adsorption. The roughness did not have a direct correlation to the adsorbed mass through the series, and neither did the friction coefficient (Figure 6D). Since the adsorbed mass and the SFE exhibited a linear correlation, it is clear that the roughness and friction coefficient correlation with SFE is similar to the adsorbed mass (Figure S5 vs. Figure 6D).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.