Advanced Healthcare Materials

Here, we present a non‐woven, μm‐thick filter paper, consisting of pristine highly crystalline cellulose nanofibers and featuring a tailored pore size distribution particularly suitable for virus removal. To our knowledge, this is the first time a 100% natural, unmodified nanofibrous polymer‐based membrane is demonstrated capable of removing virus particles solely based on the size‐exclusion principle, with log₁₀ reduction value (LRV) ≥ 6.3, thereby matching the performance of industrial synthetic polymer virus removal filters.

The threat of the next great influenza virus pandemic is worrying as in the past millions of people died from its outbreaks. The problem is aggravated by the frequent mutations, which undermine the development of efficient vaccines against influenza virus. Therefore, in order to minimize the risks associated with pandemic, efficient, robust and affordable (air‐borne) virus removal filters are highly demanded for prevention of spreading viruses in hospitals, transportation hubs, schools, or other venues with high human turn‐over. Influenza virus is also well adapted to survive in aquatic environments,1 and can replicate and transmit via waterland birds.2 Furthermore, robust and affordable virus removal filters are also demanded by the biotechnology industry as there are hundreds of products possessing the potential risk of viral infection, including cell‐derived monoclonal antibodies, plasma‐derived coagulation factors (e.g., factors VIII and IX) and immunoglobulins, or proteins manufactured with processes that employ viruses as biological expression systems, e.g., human or animal vaccines.3 Absence of viral contamination further needs to be ensured for all types of therapeutic proteins, for example, human antithrombin III, derived from the milk of transgenic mammals.4 Since cell cultures and explants are often cultivated in serum‐derived media, both endogenous retroviruses and adventitious viral contaminants introduced during manufacturing also constitute a risk factor, for example, bovine viral diarrhoea virus, bovine parainfluenza 3 virus (PI‐3), parvoviruses.[3] The threat of viral contamination of biotechnology products is manifold, including active human infection, for instance, human immunodeficiency virus (HIV), hepatitis C (HVC), and enhanced oncogenic risk due to dormant infection, for example, avian leucosis virus.[3] In addition, the advances in viral vectors of gene delivery to cure cancer stipulate the development of purification techniques wherein the non‐immunogenic virus particles can be efficiently separated from proteins and other cell debris during production.[3]

Virus removal can be achieved by several means, for example, filtration (depth filtration or surface screening), partitioning and fractionation (centrifugation), and chromatography (ion‐exchange, affinity, gel‐permeation).[3],5 Filtration is attractive because it is both non‐destructive and non‐interfering, that is, does not compromise the integrity of biological samples and does not cause immune reactions. There are several parameters, which can influence the efficiency of the filtration process such as virus size, filter pore size distribution, filter thickness, pore tortuosity, number of filter layers, surface charge, surface chemistry as well as feed stream pH and ionic strength.5 With respect to robustness, size‐exclusion filtration is the method of choice since it is theoretically not restricted by the process parameters, as opposed to, for example, adsorptive entrapment (also known as interception).[3],5, 6

The materials used for virus filtration typically include various synthetic and semisynthetic polymers (e.g., polyvinylidene difluoride (PVDF), cuprammonium‐regenerated cellulose, cellulose acetate, and polycarbonate) as well as ceramic filters.[3] The ceramic filters are usually non‐disposable, heavy, brittle, and costly (about 10 times more expensive than synthetic polymer membranes) and therefore less common.[3] The majority of virus removal filters are synthetic or semisynthetic polymers, and these filters are typically produced via phase‐inversion processing requiring hazardous solvents and coagulants with rigorously controlled processing parameters and pore annealing to obtain desired narrow pore size distribution suitable for virus removal. To our knowledge, there are currently no virus removal membranes made from a single, 100% unmodified natural polymer and featuring virus removal efficiency matching that of industrial synthetic or semisynthetic polymers, viz. LRV ≥ 6–7 (for large viruses), while allowing for protein passage.[3],[3],5

Cellulose is a valuable industrial commodity thanks to its widespread availability in nature, renewability, mechanical strength, flexibility, inertness, and biodegradability. With respect to filtration applications, cellulose is attractive because it is inert, non‐toxic, hydrophilic, resistant to pH between 2 and 11, thermally stable (i.e., can be sterilized by autoclaving), cost efficient, and disposable. Normal filter paper has too large pores to retain viruses; however, it can be surface modified to impart virus interception properties.7 The advances in nanotechnology have stirred the interest in the development of nanocellulose‐based virus removal filters focusing mainly on the adsorptive type filters acting via electrostatic interactions rather than size exclusion.8 Furthermore, the reported membranes were composite materials of nanocellulose whiskers with, e.g., polyacrylonitrile/poly(ethylene terphtalate)8 since pure nanocellulose irreversibly agglomerates into a compact, essentially non‐porous mass upon dewatering, a process known as cellulose hornification.9 In order to avoid hornification, critical‐point drying, solvent exchange, or lyophilization are used to produce the so‐called porous aerogels of nanocellulose.9 However, the high costs associated with the processing (e.g., liquid carbon dioxide) are still an impediment for large‐scale industrial production of nanocellulose aerogels. Probably, the only known nanocellulose material, which retains its large surface area and porosity upon conventional drying, is nanocellulose from filamentous green algae, for instance, Cladophora algae, which is related to the structural peculiarities of this material such as superior degree of crystallinity, stiffness, and large thickness of its elementary fibrils as compared to, for example, land‐plant‐derived nanocelluloses.10 The Cladophora cellulose has been studied in several applications such as paper batteries,11 electrochemically controlled DNA extraction membranes,12 hemodialysis membranes,13 drug delivery vehicles,14 and rheology enhancers.15 Although it has been known that Cladophora cellulose membranes feature pore size distribution between 2 and 200 nm, as derived from N₂ gas adsorption and Hg intrusion studies,[13],16 dedicated studies on tailoring the pore size distribution in the region suitable for virus removal have not been performed.

In this paper, by using a conventional household heat‐press to dry the wet pulp, we show that one can obtain a pure nanocellulose filter paper featuring an average pore size of 19 nm and specific surface area of 88 m² g⁻¹ as derived from N₂ gas adsorption experiments. As it is seen in Figure 1, the predominant majority of pores is ≤30 nm. The heat pressing of nanocellulose produced membranes having an average thickness of 70 μm and a total porosity of 35% (See Figure S1, Supporting Information).

It is recognized that one of the most challenging tasks for designing virus removal membranes is tailoring the membrane upper pore size cut‐off so that the filter retains viruses having a particle size between 12 and 300 nm while allowing for unhindered passage of proteins, which typically range between 4 and 12 nm in size.17 We have previously shown that proteins pass unhindered through membranes of Cladophora nanocellulose.[13] Therefore, the pore size distribution presented in this work is promising for virus filtration applications especially for large viruses ≥50 nm and could be enhanced further to include small viruses too, for example, by adjusting membrane thickness.

In order to demonstrate the ability of nanocellulose filter paper to retain viruses, we first tested it using polystyrene latex beads tagged with fluorophore groups and having a bead size of 500, 100, and 30 nm, respectively. In order to avoid micro­cracks, the filtration was performed slowly by adjusting suction pressure to 10–15 kPa, with the corresponding hydraulic permeability of 50 ± 2 μL h⁻¹ cm⁻².

In Figure 2 the polystyrene latex beads of varying size are seen stacked on the surface of the porous filter paper. The underlying cellulose nanofibers are clearly visible in Figure 2a,b, whereas the surface of the membrane is almost fully covered with small latex beads in Figure 2c. As it is seen in Figure 2b, the irregularly shaped pores in the membrane arise due to the interstices between nanofibers and generally are much smaller than 100 nm. In order to estimate the filter retention efficiency quantitatively, fluorescence intensity in the filtrate was compared to that of the start dispersion.

Figure 3 shows the emission spectra related to the fluorophore groups tagged to each type of the latex bead. It was observed that the concentration of the fluorophores in the filtrate was below the detection limit and did not differ from that of the background for the pure solvent. Furthermore, upon filter examination, distinct coloration could be seen with an unarmed eye (see inserts in Figure 3a–c). It is thus concluded that the produced membranes are capable of surface screening the latex beads ≥30 nm, which correlates well with the upper pore size cut‐off obtained from N₂ gas adsorption analysis. Thus, the pore size distribution of the membrane remains more or less unchanged both in the dry and wet state, which is not surprising since it is known that the highly crystalline cellulose nanofibers do not swell.18

In order to further verify the size exclusion properties of the membrane, we tested the membrane retention properties with swine influenza virus (SIV) A as a model virus strain. SIV has a typical particle size of 80–120 nm in diameter and a spherical shape with characteristic hemagglutinin and neuraminidase protein structures extending outward from its surface.

Table 1 summarizes the results of the SIV retention test, and Figure 1d depicts the electron microscopy image of SIV particles retained on the surface of the nanocellulose filter paper. Following the virus retention test, no infectious SIV particles were found in each of the triplicate 96‐well plates in the filtrate. Because it is impossible to assert the absence of virus particles with certainty due to assay limitations, we assume that the log₁₀ virus titre in the filtrate is ≤ 0.8 mL⁻¹, viz. below the detection limit as calculated from the Kärber's formula. The corresponding virus removal probabilities are then LRV ≥ 5.2 or LRV ≥ 6.3 depending on the log₁₀ titre value of the feed solution. For large viruses, that is, viruses with particle size ≥50 nm, the state of the art industrial filters exhibit LRV ≥ 6–7. Thus, it can be concluded that the nanocellulose filter presented here matches the retention efficiency of the industrial filters. This finding is remarkable considering that the green filamentous Cladophora sp. algae are a commonly known seasonal water‐pollutant of coastal areas with a global negative environmental impact.10, 19 Further, the facile heat‐pressing manufacturing method is appealing from the industrial point of view and can be adapted to the existing capacities for roll‐to‐roll paper making processes.

Experimental Section

Materials: Crystalline nanocellulose from Cladophora algae was supplied by FMC Biopolymer (G‐3095–10 batch; USA). Fluorophore‐tagged polystyrene latex beads of various size, viz. 30 nm (L5155; 2.5% solids; carboxylate‐modified; yellow‐green; λ_ex ≈470 nm; λ_em ≈ 505 nm), 100 nm (L9902; 2.5% solids sulfonate‐modified; red; λ_ex ≈ 575 nm; λ_em ≈ 610 nm), and 500 nm (L1403; 2.5% solids; sulfonate‐modified; orange; λ_ex ≈ 520 nm; λ_em ≈ 540 nm), were used for retention tests as supplied by Sigma–Aldrich.

Microorganisms: SIV A/swine/Sweden/9706/2010(H1N2) was used. Propagation of SIV was done by methods described previously,20 using Madin Darby Canine Kidney (MDCK) (ATCC CCL‐34) cells.

Membrane Preparation: About 300 mg of Cladophora was dispersed in deionized water using high‐shear ultra sonic treatment (750 W; 20 kHz; 13 mm probe; Vibracell, Sonics, USA) for 10 min at 71% amplitude. The dispersed sample was then drained on a nylon filter having an average pore size distribution of 100 nm (R01SP09025; 90 mm; GE Water and Process Technologies). The collected cellulose mass was allowed to dry until slightly damp‐ just enough to allow the hydrogen bonds to form a coherent layer. Subsequently, the nylon support was easily delaminated using tweezers without affecting the integrity of the cellulose layer. The sample was then dried under load using a heat‐press (Rheinstern, Germany) at 105 °C to produce a flat paper sheet.

Membrane Porosity: The total porosity of the membrane was calculated from the ratio between the bulk and true density as follows:

where ε_% is the total porosity, ρ_bulk is the membrane bulk density calculated from membrane dimensions, and ρ_true = 1.64 g cm⁻³ is the true density of Cladophora cellulose.10 The thickness of the produced membrane was measured using a digital 10⁻³ mm precision calliper (Mitutoyo Absolute, Japan). The thickness of each membrane was measured in 10 different positions, and in total 16 membranes were evaluated.

Pore Size Distribution: The pore size distribution of the produced membrane was measured according to the Barret‐Joyner‐Halenda (BJH) method from nitrogen gas adsorption isotherms using ASAP 2020 (Micromeritics, USA) instrument.

Hydraulic Permeability: The rate of water flow through a 26 mm in diameter circular Cladophora cellulose membrane in a Büchner funnel was evaluated. The suction pressure was adjusted to 10 kPa.

Scanning Electron Microscopy: Following the filtration of latex beads and SIV, the membranes were studied with a scanning electron microscope (Leo 1550 FEG‐SEM, Zeiss). The membranes were sputtered with Au/Pd prior to analysis to avoid charging of the samples.

Particle Retention Test: A suspension (5 μL) of uniform polystyrene latex beads (2.5% solids) was diluted in 10 mL of water. The diluted dispersion was filtered through a Cladophora cellulose membrane (26 mm in diameter) in a Büchner funnel. The suction pressure was adjusted to 10–15 kPa. The filtrate was collected, and the fluorescence intensity was measured using a fluorospectrophotometer (Tecan Infinite M200, Austria) at the specified excitation and emission wavelengths.

Virus Retention Test: Prior to the virus retention test, the membranes were sterilized by autoclavation at 121 °C for 20 min. The tested samples were the stock, the feed, and the filtrate. The viruses were propagated as described earlier.20 The SIV feed solution was obtained by diluting the stock solution 10⁻¹ with PBS. Twenty (20) mL of the produced SIV feed solution was filtered through Cladophora cellulose membrane (26 mm in diameter) in a Büchner funnel. The suction pressure was adjusted to 10–15 kPa, and the filtrate was subsequently collected. Another 10 mL of the diluted SIV feed solution was frozen in −70 °C to be used as the hold‐control to measure the factual virus titre in the feed. The SIV titre was analyzed by the end‐point titration through cytopathogenic effect (cpe). Standard 96‐well plates containing MDCK cells were used in tenfold dilutions by assaying eight replicates of 50 μL per dilution.20 Negative controls were EMEM‐trypsin and PBS‐trypsin. The virus titres after 8 d were calculated according to Kärber21 and expressed as log₁₀ tissue culture infectious dose TCID₅₀ mL⁻¹.

The virus retention capacity of the membranes was expressed as the LRV as follows:

where C_feed and C_filtrate are the titres mL⁻¹ for the feed and filtrate, respectively.

The assay detection limit was 0.8 log₁₀ of TCID₅₀ mL⁻¹, calculated according to the Kärber's method. One additional experiment was performed using the undiluted stock solution as the feed.

Acknowledgements

G.M. and L.W. contributed equally to the first authorship. The authors thank Björn Syse for graphic design. The Bo Rydins Foundation, the Göran Gustafssons Foundation, and the FORMAS “Strong Research Environments” BioBridges 2011‐1692 program are gratefully acknowledged for financial support. A.M. is Wallenberg Academy Fellow and thanks the Knut and Alice Wallenberg Foundation for their support.

The license of this article was changed after online publication.