Viruses at Solid–Water Interfaces: A Systematic Assessment of

Article pubs.acs.org/est

Viruses at Solid−Water Interfaces: A Systematic Assessment of Interactions Driving Adsorption Antonius Armanious,†,§ Meret Aeppli,† Ronald Jacak,∥ Dominik Refardt,‡ Thérèse Sigstam,§ Tamar Kohn,§ and Michael Sander*,† †

Institute of Biogeochemistry and Pollutant Dynamics (IBP) and ‡Institute of Integrative Biology (IBZ), Department of Environmental Systems Science, ETH Zurich, Zurich 8092, Switzerland § Laboratory of Environmental Chemistry, School of Architecture, Civil and Environmental Engineering (ENAC), École Polytechnique Fédérale de Lausanne (EPFL), Lausanne, CH-1015, Switzerland ∥ Applied Physics Laboratory, Johns Hopkins University, Laurel, Maryland 20723, United States S Supporting Information *

ABSTRACT: Adsorption to solid−water interfaces is a major process governing the fate of waterborne viruses in natural and engineered systems. The relative contributions of different interaction forces to adsorption and their dependence on the physicochemical properties of the viruses remain, however, only poorly understood. Herein, we systematically studied the adsorption of four bacteriophages (MS2, fr, GA, and Qβ) to five model surfaces with varying surface chemistries and to three dissolved organic matter adlayers, as a function of solution pH and ionic strength, using quartz crystal microbalance with dissipation monitoring. The viruses were selected to have similar sizes and shapes but different surface charges, polarities, and topographies, as identified by modeling the distributions of amino acids in the virus capsids. Virus-sorbent interactions were governed by long-ranged electrostatics and favorable contributions from the hydrophobic effect, and shorter-ranged van der Waals interactions were of secondary importance. Steric effects depended on the topographic irregularities on both the virus and sorbent surfaces. Differences in the adsorption characteristics of the tested viruses were successfully linked to differences in their capsid surface properties. Besides identifying the major interaction forces, this work highlights the potential of computable virus surface charge and polarity descriptors to predict virus adsorption to solid−water interfaces.

■

solution, respectively.32,43,44,46,48,49 In addition to electrostatics, van der Waals (vdW) interactions are typically considered, especially for virus adsorption to mineral surfaces.21,23,24,42 These interactions increase with the dielectric susceptibility of the sorbent and are thus higher for metals and crystalline minerals than for amorphous minerals and organic sorbents.50 Compared to electrostatic interactions, vdW interactions are shorter-ranged and hence operate at short virus-sorbent separation distances. Both electrostatic and vdW interactions are accounted for in DLVO theory. However, this theory failed to describe experimental virus adsorption data, suggesting the presence of additional interaction forces.26,27,48,51,52 Favorable contributions to virus adsorption may originate from the hydrophobic effect.25,32,52−56 These contributions arise from the gain in free energy for systems in which adsorption minimizes the interfacial area between water and apolar surfaces

INTRODUCTION Waterborne viruses are responsible for numerous human diseases.1−3 As a consequence, the processes governing the fate, stability, and the transmission patterns of waterborne viruses in natural and in engineered systems has received considerable research attention.4−8 Among these processes is the adsorption of viruses to solid surfaces in contact with water.1,4,9−17 Depending on the systems studied, these surfaces include minerals,9,18−25 dissolved and particulate natural organic matter,26−28 filter membranes,29−35 and vegetables and fruit skins36,37 as well as human skin.38−41 Electrostatic interactions between viruses and charged surfaces are generally considered to govern adsorption characteristics. Electrostatic attraction between viruses and oppositely charged sorbent surfaces typically results in fast and extensive adsorption.4,21,28,32,42−45 Conversely, electrostatic repulsion between viruses and like-charged sorbent surfaces hinders or completely inhibits adsorption.4,23,26,44−49 Electrostatic interactions are modulated by solution pH and ionic strength (I) by virtue of their effects on the virus and sorbent surfaces charges and on the decay lengths of the electrostatic potentials in © 2015 American Chemical Society

Received: Revised: Accepted: Published: 732

September 22, 2015 November 20, 2015 December 4, 2015 December 4, 2015 DOI: 10.1021/acs.est.5b04644 Environ. Sci. Technol. 2016, 50, 732−743

Article

Environmental Science & Technology

using the PROPKA model (Section S1).79 We performed two charge calculations: while the first included all ionizable AAs in the capsids, the second included only those ionizable AAs that were identified as lying on the outer capsid surfaces facing the bulk solution, as detailed in Section S1. Polarity Characteristics of the Virus Surfaces. Virus surface polarity characteristics were estimated using three different approaches: (i) by calculating the ratio of the solvent accessible surface area (SASA) of apolar AAs on the outer capsid surface to the total SASA of all AAs on the outer capsid surface, (ii) by calculating a hydropathy index (a measure of the relative hydrophobicity of different amino acid residues) for the virus coat proteins based on the Kyte and Doolittle80 scoring system, and (iii) by using the protein design software Rosetta81,82 and applying a hydrophobic scoring system that increases exponentially with the total area of each identified apolar patch on the outer capsid surface,83 as detailed in Section S1. Chemicals and DOM samples. All chemicals used were of analytical grade and were used as received. The DOM samples included Suwannee River humic acid and fulvic acid (SRHA and SRFA) and Elliot Soil Humic acid (ESHA) from the International Humic Substances Society (IHSS) and were chosen to cover a broad range in physicochemical properties. Detailed information on chemicals and DOMs is provided in Section S1. Virus Propagation and Purification. Phages MS2 and Qβ were from DSMZ (DSMZ 13767 and 13768; Braunschweig, Germany), fr (ATCC-15767-B1) was from LGC Standards (Molsheim, France), and GA was kindly provided by Dr. Joan Jofre (University of Barcelona). The phages were replicated in Escherichia coli strain W1485 (DSMZ 5695). Following inoculation with one of the phages, E. coli was incubated in LB for 4−5 h at 37 °C. The bacterial cells were subsequently lysed using chloroform. The bacterial debris was removed from solution by centrifugation (4000 g for 15 min; Hereus Megafuge 16R; Thermo Scientific). The phage-containing supernatant was purified by repeated centrifugal membrane filtrations through 100 kDa Amicon Ultra-15 centrifugal filters (Millipore). At least ten centrifugation steps were conducted (4000g for 5−15 min). After each of these centrifugation steps, new dilution buffer (5 mM HPO42−, 10 mM NaCl, pH 7.4) was added to the filter retentate that contained the virus. The final virus stock solutions were sterile filtered through 0.1 μm syringe filters and stored at 4 °C. Virus-like particles (VLP) of MS2 (MS2-VLP) were prepared using a modified approach from Hooker et al. (2004).84,85 Details are provided in Section S1. The infectivity, RNA and protein contents, hydrodynamic diameters, and electrophoretic mobilities of the viruses and MS2-VLP were experimentally determined as detailed in Section S1. Experimental Solutions. All solutions were prepared in Milli-Q water (resistivity ≈18.2 MΩ·cm; Barnstead NANOpure Diamond) as detailed in Section S1. Experimental virus solutions were prepared immediately before use by transferring 0.1−0.4 mL aliquots of the virus stock solutions to 100 kDa Amicon Ultra-15 centrifugal filters, followed by centrifugation until the retentate had a small volume of approximately 200 μL. A volume of 15 mL of buffered solution (3 mM Tris and 10 mM NaCl, pH 8.0) was added to the retentate, followed by centrifugation. The dilution and centrifugation steps were repeated a total of at least ten times, resulting in overall volumetric washing ratios between 500 and 2000. Adsorption Experiments. Virus adsorption was determined at pH 5 to 8 and at different I using a quartz crystal microbalance with dissipation monitoring (QCM-D) equipped

on the viruses and the sorbent. Contributions from the hydrophobic effect are therefore expected to increase with the increasing apolarity of the virus and sorbent surfaces. Additional favorable contributions may arise if hydrogen bonds form between donating and accepting moieties on the virus and sorbent surfaces.21 At the same time, hydrogen bonds (and hence, their favorable contributions) are likely suppressed in aqueous systems in which water molecules are expected to occupy a large portion of the H-bonding sites. Finally, steric effects arising from irregular topographies on the sorbent or virus surfaces (e.g., protein loops that protrude from the capsid into solution) may impair close contact between the virus and the sorbent and, hence, hinder adsorption.26,48,49 Although past studies on virus adsorption provided evidence for all of these interactions, a concise understanding of their relative importance is missing. Some interactions also remain poorly understood on a fundamental level. For example, it is controversially discussed whether virus surface charges (and, hence, the electrostatic interactions with charged sorbents) is governed solely by the ionizable amino-acids in the virus capsids48,57 or also by the negatively charged nucleic acid core inside the capsid.58−60 Similarly, although contributions from the hydrophobic effect to adsorption are considered important, there is no established procedure to parametrize virus surface polarity. Experimental approaches to determine surface polarities52,56,61 are susceptible to artifacts because apolar surfaces in water carry negative charges due to a preferential surface accumulation of hydroxylate ions.62−64 To advance, reliable descriptors of the physicochemical properties of the viruses are required that can be linked to virus adsorption characteristics. Such descriptors would be particularly beneficial for waterborne human viruses that have known crystallographic structures but cannot readily be studied in adsorption experiments. The objective of this work was to systematically assess the relative contributions of the different interaction forces to virus adsorption and to identify readily computable descriptors for the physicochemical properties of viruses that can be linked to their adsorption characteristics. We selected four extensively studied bacteriophages with known crystallographic structure from the Leviviridae family (i.e., MS2, fr, GA, and Qβ, all positive, singlestranded RNA viruses).26−28,43,47−49,52,55,59,60,65−77 These viruses were selected because they have similar sizes, shapes, and capsid geometries but vary in their surface charges, polarities, and topographies. We systematically studied the adsorption of these four viruses and of one genome-free virus-like particle to five model surfaces of varying surface charges and polarities as a function of solution pH and I. Finally, we systematically studied the adsorption to three dissolved organic matter (DOM) adlayers. These were included because they often form on surfaces in natural and engineered systems, they were shown to affect virus fate and infectivity,14−17,78 and because their use allowed us to discuss our findings in the context of previous work.

■

MATERIALS AND METHODS Virus Physicochemical Properties. The surface charges, polarities, and topographies of MS2, fr, GA, and Qβ were modeled and visualized based on their crystallographic structures (2MS2,74 1FRS,75 1GAV,76 and 1QBE,77 respectively), as detailed in Section S1. Virus Charge Characteristics. Charge characteristics for viruses were computed based on the compositions of ionizable amino acids (AAs) in the virus capsids according to Penrod et al. (1996).48 The acidity constants of these AAs were determined 733

DOI: 10.1021/acs.est.5b04644 Environ. Sci. Technol. 2016, 50, 732−743

Article

Environmental Science & Technology

Figure 1. Physicochemical properties of bacteriophages MS2, fr, GA, Qβ, and of MS2 virus-like particles (MS2-VLP; i.e., MS2 capsid devoid of singlestranded RNA core). (a) Visualizations of the four viruses based on the crystallographic structures 2MS2,74 1FRS,75 1GAV,76 and 1QBE77 for MS2, fr, GA, and Qβ, respectively. The whole capsids crystallographic structures were obtained from http://viperdb.scripps.edu89 (accessed August 2014). The crystallographic structures of capsid trimers were obtained from the Protein Data Bank (PDB; http://www.rcsb.org/,90 accessed May 2014). (b) The calculated surface charges of the four viruses considering all ionizable amino acids (AAs) and the C- and N-termini in the entire capsid protein over the pH range 2−9. (c) The calculated surface charges of the four viruses considering only the ionizable AAs and the C- and N-termini that are located on the outer surface of the virus capsids over the pH range 2−9. (d) Zeta potentials of the four viruses and MS2-VLP over the pH range 2−9 in 10 mM solutions (adjusted with NaCl) determined by electrophoretic mobility measurements. (e) Estimated hydrodynamic diameters of the four viruses and MS2-VLP over the pH range 2−9 in 10 mM solution (adjusted with NaCl) using dynamic light scattering. (f) Differences in the distribution of apolar patches of varying sizes on the outer virus surfaces. Shown are histograms of the absolute numbers of apolar patches of defined sizes (left ordinate) and the hydrophobic scoring for every area range (right ordinate). 734

DOI: 10.1021/acs.est.5b04644 Environ. Sci. Technol. 2016, 50, 732−743

Article

Environmental Science & Technology with four flow-through cells (E4 system; Q-Sense AB). Each flow cell is equipped with a surface-modified, oscillating piezoelectric quartz sensor. Changes in the adsorbed mass Δm (ng·cm−2) on the sensor surface were calculated from the changes in the resonance frequency, Δf n (Hz), of the fundamental tone (n = 1) and the nth overtone (i.e., n = 3, 5, 7, 9, 11, 13) of the sensor using the Sauerbrey equation:

Δm = −C·

above 5. The differences in calculated charges in parts b and c of Figure 1 reveal highly nonuniform distributions of ionizable AAs in the capsids of all four viruses with higher densities of acidic AAs on the outer surfaces and of basic AAs on the inner surfaces, as was previously reported for only MS2.48 The IEPs of the four viruses estimated by electrophoretic mobility measurements varied between 3.5 (fr) and 4.3 (Qβ) (Figure 1d). The IEP of MS2 of 3.9 agrees well with previously determined values.87 The experimental IEPs were also in reasonable agreement with the IEPs calculated when only ionizable AAs positioned on the outer capsid surface is considered (Figure 1c). The measured zeta potentials, however, showed smaller differences in the charge characteristics of the viruses than calculated based on their ionizable AAs on the outer capsid surfaces. The charge differences may have been attenuated due to loss of structural integrity of some viruses over the course of the long-term zeta potentials measurements or by contributions of the ssRNA cores to the electrophoretic mobilities of the viruses, as argued by others.58−60 Nevertheless, the good agreement between the experimental IEPs and the IEPs calculated based on the ionizable AAs on the outer capsid surfaces suggests that the surface-charge characteristics of the viruses were mainly governed by charged AAs on the outer capsid surfaces. Similar results were reported in previous studies, which also argued that the effective virus surface charges in solution are mainly determined by external ionizable AAs.48,57 Our finding is furthermore supported by the observation of virus aggregation at pH values close to the calculated IEPs shown in Figure 1c: aggregation is indicative of the viruses carrying no net charges and, hence, a loss of stabilization from capsid−capsid electrostatic repulsion. Consequently, when lowering the pH, we found that the aggregation of GA and Qβ started at pH 5 and of fr and MS2 at pH 4 (Figure 1e). To avoid confounding effects of virus aggregation, we conducted subsequent adsorption experiments at solution pH at which viruses were monodisperse (i.e., pH ≥ 5 for MS2 and fr and pH ≥ 6 for GA and Qβ). Figure 1d,e also shows data for MS2-VLP, which were obtained by removing the negatively charged ssRNA from MS2 while leaving the protein capsid intact. The zeta potentials, as well as the aggregation behavior of MS2-VLP, were very similar to those of MS2, suggesting that the ssRNA core had, at most, a small contribution to the effective surface charge of the viruses in solution. On the basis of these findings, we hypothesize that the electrostatic interactions of viruses with charged sorbents are primarily governed by the charges on the outer capsid surfaces of the viruses but do not involve charges carried by internal amino acids nor the nucleic acid. Virus Surface Polarities. Figure 1f shows the absolute numbers of apolar patches of defined sizes on the outer surfaces of the four viruses as well as the patch size-dependent hydrophobic scores used to calculate total hydrophobic scores of the capsids.83 The calculations reveal that the four viruses have very different surface polarities that decrease in the order of MS2 (smallest total hydrophobic score: 45.6; predominance of small apolar patches) ≫ GA (176.0) > fr (230.4) ≫ Qβ (592.8, higher number of large apolar patches). These polarity differences were not revealed by the other two calculation approaches; the ratios of SASA of apolar AAs on the surfaces to the total SASA of the capsid surfaces were approximately 0.6 for all four viruses. Similarly, the hydropathy index plots (proposed as a simple method for displaying the hydrophobic character of a protein)80 of all four viruses’ capsid proteins suggested that they had similar overall polarities (Figure S3). Previous work showed that

Δfn (1)

n

where C is a sensor-specific proportionality constant ( 17.7 (ng· cm−2·Hz−1)). Initial rates of virus adsorption to different surfaces were compared based on the adsorption efficiency value α:

α=

ka k max

(2) −1

where ka and kmax (both Hz·min ) are the initial virus adsorption rates to a given surface and the diffusion-limited (highest possible) adsorption rate to positively charged SAM-NH3+ surfaces, respectively (see below). The initial rates were obtained from the slopes of linear fits to the initial changes in Δf n versus time during virus adsorption. All QCM-D experiments were conducted at a constant volumetric flow rate of 20 μL·min−1 and a constant temperature of 20 °C. Virus adsorption was studied to model and DOM adlayer surfaces. The model surfaces included gold (Au), amorphous silica (SiO2), and self-assembled monolayers of alkyl-thiols on gold-coated QCM-D sensors formed from ethanolic solutions of cysteamine (SAM-NH3+), 11-mercaptoundecanoic acid (SAMCOO − ), and 1-dodecanethiol (SAM-CH 3 ), following a previously published protocol.86 DOM adlayers were formed by delivering DOM solutions (50 μgDOM·mL−1) over SAM-NH3+ surfaces, followed by rinsing with DOM-free solutions of the same pH and I prior to delivering virus-containing solutions.86 More details on the adsorption measurements, including SAM formation and sensor cleaning, are provided in Section S1.

■

RESULTS AND DISCUSSION Physicochemical Properties of Viruses. Virus Sizes and Surface Topographies. The full capsids of MS2, fr, GA, and Qβ are depicted in Figure 1a. All capsids are composed of 180 coat proteins in T = 3 icosahedral symmetry (and a single A protein; not shown), spherical with comparable outer diameters of approximately 29 nm, and enclose single-stranded RNA (+ssRNA) of comparable sizes. Loops of hydrophilic AAs (conserved sequence DNGGTGD) extend approximately 1 nm from the capsid surfaces of MS2, fr, and GA into the solution. Conversely, loops on the capsid surface of Qβ (sequence IGKDGKQ) are oriented into the capsid surface away from the solution. Previous work suggested that loops on the surface of MS2 result in unfavorable steric forces that hindered its adsorption.26,27,48 On the basis of the capsid topographies, we expect comparable steric effects for fr and GA but not for Qβ. Virus Surface Charges. Calculations considering all ionizable AAs and the charged C and N termini of the capsids led to isoelectric points (IEPs) above 8 for all four viruses and, thus, decreasing positive net charges with increasing pH from 5 to 8 (Figure 1b). Conversely, calculations considering only the ionizable AAs positioned on the outer surfaces of the capsids resulted in much lower IEPs between 3.9 (MS2) and 5.0 (GA) and, hence, increasing net-negative charges with increasing pH 735

DOI: 10.1021/acs.est.5b04644 Environ. Sci. Technol. 2016, 50, 732−743

Article

Environmental Science & Technology

Figure 2. Adsorption of bacteriophages MS2, fr, GA, and Qβ and of MS2 virus-like particles (MS2-VLP) to amine-terminated self-assembled monolayers (SAM-NH3+) and carboxyl-terminated SAMs (SAM-COO−). (a) Changes in resonance frequencies (Δf n/n) and in energy dissipation values (ΔDn) of the fundamental tone (n = 1) and five overtones (n = 3, 5, 7, 9, and 11) upon the adsorption of MS2 to SAM-NH3+ surfaces at pH 6 and ionic strength I = 10 mM (adjusted with NaCl). (b) Changes in the adsorbed mass during adsorption of MS2 to SAM-NH3+ surfaces at pH 6 and I = 10 mM, calculated from the frequency data in Figure 1a using the Sauerbrey equation (pink trace). The results of a replicate experiment conduced with a different MS2 solution and a different SAM-NH3+ surface under the same solution conditions are shown in red (raw data not shown). (c) Changes in the adsorbed masses during adsorption of MS2 (replotted from panel b), fr, GA, Qβ and MS2-VLP to SAM-NH3+ surfaces at pH 6 and I = 10 mM (adjusted with NaCl). (d) Final adsorbed masses of MS2, fr, GA, Qβ, and MS2-VLP on SAM-NH3+ surfaces at different pH values. All experiments were conducted at an ionic strength of I = 10 mM (adjusted with NaCl). (e) Adsorption profiles of MS2 to SAM-NH3+ and SAM-COO− surfaces at pH 7 and I = 10 mM (adjusted with NaCl). (f) Adsorption profiles of MS2-VLP to SAM-NH3+ and SAM-COO− surfaces at pH 7 and I = 10 mM (adjusted with NaCl).

amine in dissolved form) at pH 6 and I = 10 mM. Shown are the changes in resonance frequencies, Δf n/n, and in dissipation energies, ΔDn for the fundamental tone (n = 1) and five selected oscillation overtones. The experiment consisted of three consecutive phases. First, a virus-free solution was run over the sensor (t < 0 min) to obtain a stable baseline reading (i.e., Δf n/n and ΔDn= 0). Second, starting at t = 0 min, MS2-containing solutions were delivered over the sensor. Adsorption of MS2 to the SAM-NH3+ surface resulted in decreasing Δf n/n and increasing ΔDn values. The initial adsorption rate was high and approximately constant (i.e., linear decrease in Δf n/n over time). The adsorption rate started to decrease when values of Δf n/n ≈

differences in the adsorption of enzymes of varying polarities to lignin were better described by the hydrophobic scoring system rather than the SASA approach.88 On the basis of the total hydrophobic scores, we hypothesized decreasing energetic contributions from the hydrophobic effect to adsorption in the order Qβ (apolar) > fr > GA > MS2 (polar). Virus Adsorption to Model Surfaces. Positively Charged SAM-NH 3 + Surface. Figure 2a shows the results of a representative QCM-D experiment for MS2 adsorption to a positively charged SAM-NH3+ surface (with reported pKa values > 7.091−93 but likely a wider pH range of (de)protonation of the amino group in the SAM structure as compared to the same 736

DOI: 10.1021/acs.est.5b04644 Environ. Sci. Technol. 2016, 50, 732−743

Article

Environmental Science & Technology

Figure 3. (a) Adsorption efficiency values, α, of MS2, fr, GA, and Qβ to carboxyl-terminated self-assembled monolayers (SAM-COO−), amorphous silica (SiO2), methyl-terminated SAMs (SAM-CH3), and gold (Au) at pH 5−8 and I = 10 mM (adjusted with NaCl). All panels also include α values for the adsorption of MS2 and Qβ at pH 8 and I = 100 mM (adjusted with NaCl). (b) Adsorption profiles of Suwannee River humic acid (SRHA) onto positively charged amine-terminated self-assembled monolayers (SAM-NH3+) and, after rinsing the system with SRHA-free solutions, of MS2 and Qβ to the formed SRHA adlayers, all at pH 6 and an ionic strength of I = 10 mM (adjusted with NaCl). (c) Adsorption efficiency values, α, of MS2, fr, GA, and Qβ to Suwannee River humic acid (SRHA), Suwannee River fulvic acid (SRFA), and Elliot Soil humic acid (ESHA) adlayers at pH 5−8 and I = 10 mM (adjusted with NaCl) and of MS2 and Qβ to SRHA at pH 8 and I = 100 mM (adjusted with NaCl). DLA= diffusion limited initial adsorption rate (i.e., α = 1); LOQ= estimated limit of quantification of the QCM-D approach (i.e., α = 0.01).

−90 to −100 Hz were reached. Finally, adsorption leveled off at Δf n/n ≈ −120 Hz. During the adsorption, ΔDn values initially increased to maximum values and subsequently decreased to stable, nonzero values. Third, the system was rinsed with virusfree solutions (t > 85 min). Stable Δf n/n and ΔDn values demonstrated that MS2 did not desorb from the sensor surface. The data from Figure 2a is replotted in Figure 2b (pink trace) as changes in the adsorbed mass over time, calculated using the Sauerbrey equation. Eq 1 was applicable given that the formed MS2 adlayer was relatively rigid, which was apparent from overlaying Δf n/n traces and relatively small changes in the dissipation values (i.e., ΔDn/(Δfn/n)≪ 4·10−7 Hz−1).94 Figure 2b also shows the MS2 adsorption profile collected in a second, independent experiment under identical conditions, highlighting the reproducibility of the approach. The adsorption profiles of fr, GA, and Qβ as well as MS2-VLP to the SAM-NH3+ surfaces under the same solution conditions are shown in Figure 2c and Figure S5 and were very similar to those of MS2. The adsorption profiles of all viruses and MS2-VLP on the SAM-NH3+ surfaces are consistent with strong virus−surface electrostatic attraction. Initial adsorption rates were likely controlled by the rate of virus diffusion to the surfaces. This is supported by a linear increase in initial adsorption rates with increasing virus concentrations in the experimental solutions (Figure S6). Adsorption leveled off when the jamming limits of viruses on the sorbent surfaces were attained. We estimated that

the jamming limits corresponded to the surface coverages between 24 and 40% using a hydration model developed for the adsorption of spherical particles to QCM-D surfaces (Figure S7).95 Our data is in good agreement with final adsorbed masses (2440 ng·cm−2) and surface coverages (33%) reported for the adsorption of similarly sized cowpea mosaic virus (diameter of 28 nm).95 Adsorption into surface jamming limits is further supported by the pronounced maxima in ΔDn values at intermediate stages of adsorption (Figure S5). These maxima originate from hydrodynamic stabilization effects between adjacent spherical particles on a sorbent surface upon the attainment of extensive surface coverages (i.e., 10 to 20%).96 Adsorption to the SAM-NH3+ surfaces was carried out at pH 5 to 8 for MS2 and fr, pH 6 to 8 for GA and Qβ, and pH 6 and 7 for and MS2-VLP. No data is shown for GA and Qβ at pH 5 because it indicated multilayer adsorption, consistent with aggregation of GA and Qβ at pH 5 in solution (Figure 1e). Although the solution pH had little effect on initial adsorption rates, the final adsorbed masses decreased from pH 5 to 8 for all four viruses and MS2-VLP (Figure 2d). This decrease likely resulted from increases in the negative charges on the capsid surfaces with increasing pH (Figure 1c), thus increasing electrostatic repulsion between adjacent adsorbed viruses and resulting in jamming limits with fewer adsorbed viruses. Comparable final adsorbed masses of the different viruses at a given pH are consistent with their comparable sizes (Figure 1a). 737

DOI: 10.1021/acs.est.5b04644 Environ. Sci. Technol. 2016, 50, 732−743

Article

Environmental Science & Technology Attainment of virus jamming limits on the SAM-NH3+ surfaces provided compelling evidence that the used virus solutions were of high purity and, thus, that the measured rates and extents of adsorption indeed resulted from viruses attaching to the sensor surfaces. The use of solutions containing coadsorbing molecules smaller or larger than the viruses would have resulted in plateauing adsorption at much smaller or larger final masses, respectively.97 To achieve high-purity virus solutions, we used the purification protocol that entailed three critical features: (i) exclusion of a polyethylene glycol (PEG) treatment step, (ii) use of copious amounts of buffer in repeated washing cycles, and (iii) purification directly before the use of virus solutions in adsorption experiments. We omitted the PEG step because virus solutions “purified” with this step showed diminished adsorption (Figure S9). This finding is consistent with previous reports showing that the use of PEG during purification alters virus aggregation and electrokinetics.70 We propose that future studies on virus adsorption using QCM-D or comparable techniques routinely determine and report the final mass at which adsorption plateaus to directly confirm the purity of the used virus solutions. Such a confirmation through QCM-D measurements cannot be easily obtained if only initial virus adsorption rates are determined, as was the case in previous studies.26,27,45 In all subsequent adsorption experiments, one of the four flow cells of the QCM-D system always held a sensor with a SAMNH3+ surface, and the other three flow cells contained sensors with differing surface chemistries. Adsorption to the SAM-NH3+ surface served to confirm that the virus solutions used in each experiment were pure. The initial virus adsorption rate to the SAM-NH3+ surface was further used to determine kmax and allowed us to normalize the initial adsorption rates to the other surfaces measured in the remaining three flow cells, ka, according to eq 2. The obtained α values corrected for small variations in the virus concentration between experiments and thus allowed comparing of the initial adsorption rates of all viruses and MS2VLP to different surfaces and under different solution conditions across all experiments. Negatively Charged SAM-COO− and SiO2 Surfaces. We further investigated the role of virus−surface electrostatics by running pH-dependent adsorption experiments to SAM-COO− and amorphous silica (SiO2) surfaces. SAM-COO− surfaces are negatively charged above pH 5 (the carboxylate groups in the SAM-COO− surfaces have reported pKa values ≤5.564,91 but likely (de)protonate over a wider pH range in the SAM structure as compared to the same carboxyalkylthiol in freely dissolved form) and polar (contact angle of sessile water droplet≈ 15°).86 The SiO2 surface is negatively charged98 and highly polar (contact angle of sessile water droplet