Analytical Protocols for Separation and Electron Microscopy ofNanoparticles Interacting with Bacterial CellsClaudia Sousa,† Diana Sequeira,† Yury V. Kolen’ko,‡ Ines Mendes Pinto,‡ and Dmitri Y. Petrovykh*,‡

†Center of Biological Engineering, University of Minho, Braga 4710-057, Portugal‡International Iberian Nanotechnology Laboratory, Braga 4715-330, Portugal

*S Supporting Information

ABSTRACT: An important step toward understandinginteractions between nanoparticles (NPs) and bacteria is theability to directly observe NPs interacting with bacterial cells.NP−bacteria mixtures typical in nanomedicine, however, arenot yet amendable for direct imaging in solution. Instead,evidence of NP−cell interactions must be preserved inderivative (usually dried) samples to be subsequently revealedin high-resolution images, for example, via scanning electronmicroscopy (SEM). Here, this concept is realized for a mixedsuspension of model NPs and Staphylococcus aureus bacteria.First, protocols for analyzing the relative colloidal stabilities ofNPs and bacteria are developed and validated based onsystematic centrifugation and comparison of colony formingunit (CFU) counting and optical density (OD) measurements.Rate-dependence of centrifugation efficiency for each component suggests differential sedimentation at a specific predicted rateas an effective method for removing free NPs after co-incubation; the remaining fraction comprises bacteria with any associatedNPs and can be examined, for example, by SEM, for evidence of NP−bacteria interactions. These analytical protocols, validatedby systematic control experiments and high-resolution SEM imaging, should be generally applicable for investigating NP−bacteria interactions.

The unique physicochemical properties of nanoparticles(NPs) are the basis for their potential applications in

nanomedicine, whereby NPs interacting with cells provide ameans for detecting, monitoring, or controlling cell functionsvia nonbiological properties of NPs (magnetic, electronic,optical, mechanical).1−3 Interactions of NPs with bacterial cells,for example, have been studied for overcoming antibioticresistance,4,5 biosensing,6−9 and gene delivery.10

Despite the significant progress in characterizing interactionsbetween NPs and bacteria,11−14 an important remainingchallenge is the lack of a general method that can providedirect, systematic, and quantitative observations of NPsinteracting with bacteria. The current approaches tend toprimarily focus on biological or physical aspects of suchsamples. A commonly used biological assay, for example, teststhe antimicrobial effect of NPs by evaluating the bacterialviability after exposure to NPs.15,16 Physical aspects of NP−bacteria interactions tend to be analyzed by high-resolutionimaging techniques, such as electron,8,9,11,12,14 atomic force,17

or Raman13 microscopies, which can provide direct andquantitative information about NPs and cells (such asproximity, penetration, aggregation) but typically not directlyin solution environments where NPs and bacteria interact.Artifacts introduced during preparation for imaging, therefore,can obscure evidence of NP−bacteria interactions.

If a protocol can be developed to minimize sample-preparation artifacts, scanning electron microscopy (SEM)would offer significant benefits for imaging NPs interacting withbacterial cells. Field-of-view (FOV) dimensions accessible bySEM enable capturing statistically significant numbers ofresolved individual cells within a single image [Figures S1−S2in the Supporting Information (SI)].18,19 The high spatialresolution and depth-of-field of SEM also enable simultaneousobservation of μm-scale objects, such as bacteria, and NPsresolved individually or in aggregates (Figure 1).The concentrations of cells and NPs dramatically influence

the preparation and SEM imaging of NP−bacteria mixtures.Samples with low concentrations of both cells and NPs (Figure1e and f) are typically the easiest to image at high resolution;9,20

however, isolated individual areas of such samples (e.g., FigureS2b, SI) may not be representative of the whole and the lowconcentration of NPs may have limited their opportunity tointeract with particular cells in each FOV (Figure 1f).Conversely, having a high concentration of NPs (or cells) inmixed samples increases the possibility of observing artifactsfrom both preparation and imaging procedures (Figure 1a−

Received: October 13, 2014Accepted: March 12, 2015Published: March 12, 2015

Article

pubs.acs.org/ac

© 2015 American Chemical Society 4641 DOI: 10.1021/ac503835aAnal. Chem. 2015, 87, 4641−4648

d).14,15 While it obfuscates evidence of specific NP−cellinteractions, coating of cells with a dense layer of NPs issuccessfully used, for example, in membrane-protein enrich-ment protocols, which are commonly illustrated with SEMimages resembling our Figure 1a−c.21−24 A high concentrationof NPs during incubation, therefore, provides a simple generalmethod for ensuring that all the cells had the opportunity tointeract with NPs before having been prepared for imaging.The key to preparing optimal samples of bacteria and any

associated NPs for SEM imaging is, therefore, in controlling theNP concentration throughout the process. Specifically, startingwith an excess of NPs in the incubation solution produces amixture of free NPs and NPs associated with bacteria. To avoidartifacts during imaging, free NPs are then removed beforedrying the sample. Conversely, any NPs that became associatedwith bacteria during co-incubation may reveal evidence andproperties of the underlying NP−bacteria interactions underdirect, systematic, and quantitative ex situ analysis.The importance of ensuring that all the free NPs have been

removed after co-incubation depends on the context of anexperiment. For example, when verifying high loading of eachcell with NPs,21−24 including NPs in a form of dense layers(Figure 1a and b) or aggregates (Figure 1d), even a relativelyhigh fraction (perhaps, as high as 30%) of free NPs among allNPs sedimented with cells would not affect the interpretation

of results dramatically. The salient common characteristic ofsuch experiments is the relative unimportance of distinguishingbetween NP−NP and NP−cell interactions. Conversely, whenattempting to investigate specificity of NP−cell interactions,8,9any artifacts related to the presence of free NPs need to becarefully minimized, to avoid misinterpreting the apparentresults, especially in comparative studies of different NPs, forexample. This caveat becomes singularly important whenattempting to verify that NP−cell interactions are minimal(or even absent), as would be required in measurements ofnegative controls in studies of specific NP−cell interactions.For separating free NPs from a co-incubation mixture, we

propose centrifugation because it relies on the colloidal stabilityof bacteria and NPs; the latter can be controlled viaphysicochemical properties of NPs and the solvent.25

Furthermore, centrifugation is routinely used for separationand concentration of bacteria26,27 as well as NPs28,29 or NP-loaded fragments of cell membranes.21−24 For larger (∼5−15μm) cells, dense coating with NPs can be sufficient to sedimentthe NP-loaded cells from an NP-cell mixture based on theirhigher density.21,23,24 For mixtures of NPs and submicrometer-sized bacteria with relatively low NP loading, however, density-based sedimentation may not be very effective, so more generalcolloidal stability differences have to be used for differentialsedimentation, especially when a nearly complete removal offree NPs, as discussed above, is the goal. Here, we demonstratea basic example of a systematic methodology for performing adeterministic separation of free NPs from a co-incubationmixture, whereby the initial separate measurements of NPs andcells are used (1) to narrow down the mixed NP-bacteriasamples parameter space (which, a priori, is vast, owing to thecombinatorics of possible mixtures) and (2) to predict theparameters for separation of mixed NP-bacteria samples. Thisanalytical separation protocol produces samples of bacterialcells and any NPs associated with them that can be prepared forsubsequent direct observation by SEM (or by other microscopyor spectroscopy methods)8,13 to obtain information about NP−bacteria interactions that occurred during their co-incubation.

■ EXPERIMENTAL SECTIONNanoparticles with Organic Coatings. Magnetite NPs

were synthesized by coprecipitation (CP) in an automatedreactor or by a hydrothermal method (HT), as describedpreviously (SI).25 The synthesized NPs were stabilized byoleate (OL) or poly(acrylic acid) (PAA) capping ligands.Tetramethylammonium hydroxide (TMAOH) surfactant wasused to transfer hydrophobic as-synthesized OL-coated NPs toaqueous phase.30 Hereafter, the NPs are denoted by thecapping ligand (PAA or OL) and the synthesis protocol (CP orHT).

Colloidal Stability of NPs. The iron oxide NPs used in thiswork are superparamagnetic,25 so their magnetic properties donot induce colloidal instability. As described in the text and theSI, sedimentation threshold during centrifugation was used toevaluate the colloidal stability of NPs in four aqueous solutions:NaCl 0.9%, NaCl 0.45%, 0.01 M phosphate buffered saline, pH7.2 (PBS 0.01 M), and distilled water.

Staphylococcus aureus Suspensions. For each experi-ment, a stock suspension of Staphylococcus aureus (S. aureus)was prepared as described in the SI. The standard protocolinvolves 3 (washing) steps, during which bacteria are pelleted at8000 rpm and resuspended by vortexing. After the finalwashing, pellets were resuspended in 30 mL of the desired

Figure 1. Sample-preparation artifacts observed by SEM for mixedsamples of NPs and S. aureus. High concentration of both bacteria andNPs produces samples with several layers of bacteria covered with arough, dense overlayer of NPs (a, b). Low concentration of bacteriamixed with a high concentration of NPs produces individual bacteriacoated with overlayers (c) or aggregates (d) of NPs. Lowering theconcentration of NPs results in concentration-limiting their attach-ment to bacteria (e, f).

Analytical Chemistry Article

DOI: 10.1021/ac503835aAnal. Chem. 2015, 87, 4641−4648

4642

aqueous solution and sonicated under conditions that disruptcell clusters without impairing cell viability.31 The suspensionwas then adjusted to a specific optical density at 640 nm(OD640) for individual experiments.S. aureus Viability. The viability of S. aureus cells was

evaluated in aqueous solutions of the same composition asthose used to test the colloidal stability of NPs. Cellsuspensions were prepared and plated as described in the SI,followed by comparing the colony forming unit (CFU)counts.32

Sedimentation Efficiency for S. aureus. Sedimentationefficiency for S. aureus in water was determined starting with acellular suspension adjusted to OD640 ≈ 0.6. Additionalmeasurements at OD640 ≈ 0.1 and 0.2 are described in theSI. For each value of centrifugation rate (1000−8000 rpm), twoseparate 5 mL samples of S. aureus suspension were centrifugedfor 5 min. After centrifugation, the OD640 was measured for thesupernatant, which was aseptically decanted to a new tube andvortexed, and the pellet, which was resuspended in 5 mL ofwater, vortexed, and sonicated (as detailed in the SI).SEM Imaging of S. aureus and NPs. After centrifugation,

each sample (supernatant or resuspended pellet) was filteredthrough a sterile 25 mm-diameter polycarbonate filter with∼0.2 μm track-etched pores (Whatman, GE Healthcare LifeSciences) mounted in a funnel assembly connected to a vacuumpump. The material retained on the filter was fixed anddehydrated as described in the SI. Samples were imaged in aQuanta 650 (FEI) field-emission SEM (SI Table S1).Statistical Analysis. Confidence intervals (CI) of mean

values were calculated using the statistical package of Microsof tExcel 2010, assuming the Student’s t-distribution with nreplicates (n < 30). The CIs are reported at 95% confidencelevel (CL) as mean ± t0.95/2·s/√n (s = standard deviation).

■ RESULTS AND DISCUSSIONOur approach to developing and validating a protocol foroptimal separation and SEM imaging of nanoparticlesinteracting with bacteria is summarized in Scheme 1. Webegin by establishing the parameter window for differentialsedimentation, which we propose to use for removing excessfree NPs (Figure 1a−d) after co-incubation of NPs andbacteria. Multiple options for choosing NPs and/or solutionconditions (Table 1) make direct sedimentation measurementsfor all possible combinations of parameters impractical. Instead,we first evaluate the colloidal stability of NPs and bacteriaseparately in systematic centrifugation experiments that aredesigned to identify a range of rpm values where free NPsremain in the supernatant while bacteria sediment (Scheme1a).We take advantage of the relatively simple colloidal stability

measurements for NPs to isolate the combination of NP andsolution properties that is likely to produce efficient removal offree NPs from co-incubation mixtures (Table 1). Colloidalstability of S. aureus bacteria is then quantified in the solutionidentified from NP experiments and after validating themeasurement methods. Finally, we compare the observedsedimentation thresholds with the pattern in Scheme 1a, andperform the separation (Scheme 1b) followed by SEM imaging.Colloidal Stability of NPs. A common benefit of using

NPs in biomedical applications is the ability to optimize the NPcoating and solution conditions, for example, to maximize thedesired NP−cell interactions. The same two parameters (NPcoating and solution conditions) affect the colloidal stability

and thus their combination can be optimized for removing freeNPs from mixtures via differential sedimentation (Scheme 1).To illustrate our methodology, we consider a series of five

NPs in which similar superparamagnetic iron oxide cores arecoated with different organic shells (Table 1 and the SI).25 Thesolution options for NP-cell systems are commonly dictated bythe biological components, such as any antibodies or enzymesinvolved in biofunctionalization of NPs and biorecognition aswell as the cells themselves; the need to maintain viability of S.aureus, for example, determined our solution choices (Table 1).Having the solution conditions constrained by the biological

component limits applicability of the otherwise commonpractice whereby changes in solution conditions (e.g., pH,ionic strength, or temperature) are used to evaluate thecolloidal stability of NPs.25 Serendipitously, the intended use ofcentrifugation suggests the use of sedimentation thresholds forcomparing colloidal stabilities of NPs in one or more solutions.For each combination of NP type and aqueous solution,

sedimentation threshold, i.e., the lowest centrifugation rate atwhich NPs sedimented, was established (Table 1). For eachNP-solution pair, three separate aliquots were tested startingfrom 1000 rpm; if sedimentation was not observed, anotherthree aliquots were tested at a higher rate (in 500 rpmincrements). In all cases, we observed abrupt sedimentation,i.e., at a given rate, NP suspensions either remained stable orchanged color and formed a pellet (appearance similar to SI

Scheme 1. Differential Sedimentation: Colloidal StabilityAnalysis (a) Predicts the Rate for Efficient Separation (b)

Table 1. Colloidal Stability of NPs under Centrifugation

coprecipitation hydrothermal

poly(acrylic acid) oleic acid + TMAOH

PAA-CP1 PAA-CP2 PAA-HT OL-HT1 OL-HT2

solutionsthe lowest centrifugation rate (rpm) at which sedimentation

was observed

NaCl 0.9% 1000 2000 5500 1000 1000NaCl 0.45% 1000 3000 4000 1000 1000PBS 0.01M 1000 3500 3000 1000 1000water 3500 3500 3000 2000 6500

Analytical Chemistry Article

DOI: 10.1021/ac503835aAnal. Chem. 2015, 87, 4641−4648

4643

Figure S3c). Accordingly, thus determined sedimentationthreshold could be used as a quantitative index of relativecolloidal stability.The highest colloidal stability was observed for OL-HT2

(TMAOH-stabilized) in distilled water (Table 1), whichremained in suspension up to 6000 rpm, suggesting the useof water as a solvent and OL-HT2 as model NPs fordemonstrating the differential sedimentation approach(Scheme 1).S. aureus Viability. While water has been identified above

as a good solvent for OL-HT2 NPs, unlike saline solutions,distilled water is not commonly used to maintain Staphylococcusviability.33 Keeping in mind the ultimate objectives of studyinginteractions of NPs with live bacteria, we started by evaluatingthe viability of S. aureus in water.In a common viability assay for bacterial cells, diluted

aliquots of the bacterial suspension are placed onto a solidgrowth medium and incubated overnight. The number ofvisible colonies that form after the incubation is counted; theresulting number of colony forming units (CFUs) can be usedto evaluate bacterial viability, e.g., by comparing with CFUcounts from reference experiments.32,34,35 As detailed in the SI,we compared the CFU counts for S. aureus after incubation inwater to CFU counts reached after incubation in three othersolvents: PBS 0.01M and NaCl 0.9%, which are commonlyused with S. aureus, and NaCl 0.45%; the latter was chosen asan intermediate value between water and NaCl 0.9%, to checkfor any trends in viability with salt concentration.In three assays performed on different days, we started from

the same initial S. aureus concentration (adjusted to OD640 ≈0.1) and found countable plates (with 6−9 replicates) in thesame dilution, therefore, the CFU counts could be compareddirectly (Figure 2, SI Figure S4), rather than via conversion to

representation as nominal CFU/mL values commonly used inmicrobiology.35,36 CFU counts for plates prepared afterincubation for 1 h indicate some variability among resultsobtained for different solutions or on different days (Figure 2),but the mean values (dark shading in Figure 2) are notsignificantly different, based on the pooled 95% CIs. MeanCFU counts also were not significantly different at 95% CLamong plates prepared after incubation for 24 h: 134 ± 57, 97± 18, 111 ± 20, and 109 ± 16 for water, NaCl 0.45%, NaCl0.9%, and PBS 0.01M, respectively. Comparing these CFUcounts, we can reach two conclusions about S. aureus viability.First, there is no significant difference in S. aureus viability after

either 1 or 24 h incubation in the four solutions that we haveexamined. Second, there is no evidence of significant death orgrowth of S. aureus when incubated for up to 24 h in all foursolutions. Accordingly, even though a previous study indicateda difference in viability of S. aureus in water vs PBS over severalweeks,35 our measurements validate the use of distilled water inexperiments with S. aureus incubation times ≤24 h.Our statistical analysis, detailed in the SI, provides additional

important indications. The variability of our results is notunusual for CFU counting data, which typically are reported inlog10 plots and analyzed in terms of order-of-magnitudedifferences.36,38 Reproducibility (95% CI) of our same-dayreplicatesa common quality-assurance indicator37,38is in±10% to ±20% range, which is in line with best-practicemeasurements32,38 and thus would not provide, by itself, acompelling justification for additional measurements aftercompleting the first set of assays. Repeating the assays ondifferent days, however, revealed that the uncertainty of CFUmeasurements likely includes a significant systematic compo-nent, in addition to the expected random variation. Weemphasize that the level of variability in our data would beinconsequential in a typical microbiology context,36 e.g.,establishing a minimum bactericidal concentration, but itmust be considered when using CFU counting as a bacterialviability or concentration measurement.

Quantification of S. aureus Concentration. Havingverified S. aureus viability in water, we can proceed to measurethe colloidal stability of these suspensions. Unlike for the caseof NPs (Table 1), however, we do not have the advantage ofabrupt visible sedimentation for bacterial suspensions. Accord-ingly, to evaluate the sedimentation efficiency as a gradualfunction of the centrifugation rate, we need to identify amethod suitable for measuring bacterial concentrations.Turbidity or optical density (OD) is commonly measured for

suspensions of bacterial cells.39,40 At bacterial concentrationsrelevant for surface attachment, however, reliably measuringOD ≈ 0.1 can be a challenge, because blank buffers commonlyhave OD around 0.03−0.05. CFU counting can reach a lowerdetection limit than does OD but is more labor-intensive andtime-consuming than OD measurements. Another importantconsideration is obtaining practically useful uncertainty (e.g., of<50%) for sedimentation efficiency values, which involve ratiosof two measurements; therefore, the uncertainty of eachmeasurement of a bacterial concentration should be <20%.To test the resolution of OD and CFU methods, three

solutions were prepared to differ by 20% (based on dilutionfactors) in bacterial concentration. For a method withresolution sufficient to distinguish these 20% differences inconcentration, we would expect the measured means to trendand to be significantly different across the concentration series.From the comparison in Figure 3 (and data in SI Table S2),

the resolution of the CFU measurements clearly is notsufficient to distinguish the samples in this range of bacterialconcentrations. For OD measurements, the means trend asexpected and are significantly different at 95% CL for the twolarger values, indicating that the resolution of OD measure-ments is sufficient for quantifying concentration ratios forsamples with OD ≥ 0.1.

Sedimentation Efficiency for S. aureus. The appropriateinitial concentration for the samples is largely determined byour interest in following sedimentation efficiency as a functionof the centrifugation rate, whether that dependence is abrupt(as we observe for NPs) or gradual. For the latter case, the

Figure 2. Viability of S. aureus after 1 h incubation in water or in saltsolutions. For each solution, CFU counts (in the same dilution) areshown for 3 separate assays (lighter shading) and their pooled meanvalue (outline, dark shading). Error bars indicate 95% CIs, as detailedin the SI.

Analytical Chemistry Article

DOI: 10.1021/ac503835aAnal. Chem. 2015, 87, 4641−4648

4644

smallest detectable pellet dictates the choice of the initialconcentration. Measuring a pellet formed by as little as 20% ofthe sample is practically useful; taking into account the OD ≥0.1 limit established above, we arrive to OD640 in 0.5−0.6 rangeas an appropriate initial concentration.The data from systematic measurements presented in Figure

4 (and detailed in SI Table S3) validate our approach to

measuring the sedimentation efficiency for S. aureus. Inparticular, our choices of OD measurements and of the initialconcentration have enabled us to observe the gradual increasein sedimentation efficiency from 1000 to 4000 rpm. Thesedimentation efficiency clearly plateaus above 4000 rpm,indicating a parameter window between 4000 and 6000 rpm forattempting the differential sedimentation of bacteria and freeNPs (Scheme 1, Table 1). We repeated the measurements at1000, 4000, and 6000 rpm (in triplicate for each rpm value) toconfirm the presence of the efficiency plateau in the 4000−6000 rpm range (SI Figure S5) and the reproducibility of ourstrategy.We note that the sensitivity limit of the OD measurements

effectively restricts reliable measurements of the sedimentationefficiency to relatively high initial concentrations of bacteria.The OD value of ≈0.6 in Figure 4 is actually above the

concentration optimal for incubating with an excess of NPs andfor subsequently depositing the entire sample on thenanoporous support for SEM analysis. Our attempts tomeasure the sedimentation efficiency starting from OD valuesof 0.1−0.2 (SI Figure S6) were inconclusive because theyproduced much larger uncertainties, in agreement with resultsdiscussed in the previous two sections. We believe, however,that for our objective of producing a representative sample ofbacteria and any NPs that directly interacted with them duringco-incubation, assuring >90% sedimentation efficiency isdesirable but not indispensable. For example, if the practicalefficiency will enable us to recover about 50% of the bacteriafrom the mixed co-incubation sample, that fraction is likely tobe representative of the population that we intend to analyzebecause a 50% fraction will not be dominated exclusively bymutant strains or other similarly low probability artifacts.

Removal of Free NPs after Co-incubation with S.aureus. Having performed colloidal stability measurements onS. aureus and NPs separately (Scheme 1a), we proceeded to thefinal step of the approach outlined in Scheme 1b, whereby theresults of those separate measurements are combined to definethe parameter window for differential sedimentation of themixed co-incubation sample of S. aureus and NPs. We startedfrom a co-incubation sample prepared with an excess of OL-HT2 NPs (as described in the SI) to maximize the possibilityfor NP−cell interactions. The mixture was incubated for 1 h,homogenized by vortexing, and centrifuged for 5 min at 4000rpm, i.e., the value at which we expected to produce differentialsedimentation of bacteria (along with any NPs that had becomeassociated with them during co-incubation), while separatingfree NPs into the supernatant (Scheme 1). Both thesupernatant and resuspended pellet were then prepared forSEM analysis by filtering onto nanoporous membranesfollowed by fixation and dehydration. SEM images of thematerial retained on the nanoporous membranes are presentedin Figures 5 and 6, and SI Figure S7.An early indication of successfully retaining the free NPs in

the supernatant was its brown color. After preparation for SEM,the supernatant sample shown at different magnifications in thepanels of Figure 5 and SI Figure S7 is clearly dominated bystructures formed by the drying of free NPs, as evidenced, forexample, by observing similarly large/thick aggregates of NPsforming both on top of cells and on the supporting membranewithout any cells in direct proximity. Cracks in the NPoverlayer, e.g., visible around the “(a)” marker in Figure 5a, arealso likely drying artifacts, supporting the above interpretation.Figure 5 thus clearly illustrates how difficult (if not impossible)it would be to find indications of NP−cell interactions withoutremoving all free NPs after the co-incubation step.

Inferences from SEM Images for NP−Cell Interactionsin Solution. In contrast to artifact-dominated images in Figure5, features likely produced due to NP−cell interactions insolution are visible in SEM images that show the resuspendedpellet sample in Figure 6, at magnifications similar to those inFigure 5. The NPs bound to the bacteria in Figure 6 can bepresumed to have attached during co-incubation, as the bulk ofthe unbound NPs clearly have been removed from this sampleduring centrifugation. A close inspection of the supportingmembrane reveals that the removal of NPs not bound to thecells has been remarkably successful. Observing only a trivialnumber of free NPs, we conclude that any artifacts associatedwith them can be considered negligible in these images.

Figure 3. Resolution of cell concentration measurements. OD640values (a) and CFU counts (b) are compared for S. aureus suspensionsprepared as a dilution series with 20% increments. Error bars indicate95% CIs (n = 4); data summarized in SI Table S2.

Figure 4. Sedimentation efficiency for S. aureus. For each rpm value,OD640 was measured for a separate aliquot of the initial suspension.The measured OD640 values (SI Table S3) for both supernatant (graybars, left axis) and resuspended pellet (red bars, right axis) arerepresented relative to the initial OD640 ≈ 0.6; error bars indicate 95%CIs, n = 2 from two separate experiments.

Analytical Chemistry Article

DOI: 10.1021/ac503835aAnal. Chem. 2015, 87, 4641−4648

4645

Multiple features in panels of Figure 6 support the possibilityof making inferences about NP−cell interactions, based onsamples prepared using our protocols. Large differences in NPloading of adjacent cells (Figure 6e), apparent clustering of NPs

around some of the cell adherence regions (Figure 6b), and theprevalence of individual NPs or apparently loose NP aggregates(indicated by observing discrete NPs or linear rather than 3Daggregates on cell surfaces) are all examples of qualitativefeatures that could readily form during the co-incubation stagedue to NP−cell interactions, to which the discrete characterand localization of NP features may be attributed. Conversely,formation of such discrete and localized features during thestochastic processes, such as filtering and drying, would bemuch more difficult to rationalize.Considering the colloidal stability of NPs (Table 1) provides

additional insight for interpreting the NP aggregates observedin Figures 5 and 6. Abruptness of sedimentation observed forall our NPs strongly indicates that formation of even small NPaggregates rapidly destabilizes the suspension. Accordingly, anyNP aggregates formed during the co-incubation step would bepelleted during centrifugation. In contrast, any NPs retained inthe supernatant could not have been substantially aggregatedduring co-incubation. The absence of large NP aggregates inFigure 6 then supports the assumption that during co-incubation the individual NPs dominated the NP−cellinteractions; conversely, the presence of aggregates in Figure5 must be related to preparation artifacts.We note that even though Figure 5 suggests that the

sedimentation efficiency for S. aureus cells in this experimentdid not reach the 90% range achieved in Figure 4, it clearlyremained above 50%; therefore, as discussed before, ourpreparation protocol has successfully produced samplesamendable for elucidating the existence and properties ofsolution-phase NP−cell interactions. While our model NPs hadnot been functionalized to produce specific attachment to S.aureus bacteria, SEM images in Figure 6 illustrate the feasibilityof future investigations of NPs attached to bacteria via putativespecific interactions.8,9 Furthermore, any bacteria that internal-ize some of the NPs will be also naturally represented in NP-cell samples prepared using our protocol. Our model NPs arenot designed to promote specific internalization and, indeed,are not internalized in any significant numbers, as confirmed byEDX measurements (SI Figures S8−S9). For appropriatelyfunctionalized NPs, however, systematic analysis of theirinternalization by bacteria will be an important area of futureanalytical developments41,42 that can take advantage of ourprotocols.

■ CONCLUSIONSWe developed and validated a protocol for producing samplesthat are amendable for elucidating from SEM images theexistence and properties of solution-phase NP−cell inter-actions. The proposed elements of the workflow, including theinitial separate investigations of colloidal stabilities for NPs andcells, have been illustrated by successfully separating, viadifferential sedimentation, and preparing for SEM character-ization samples of NPs attached to S. aureus bacteria.We used sedimentation thresholds as quantitative indices to

compare relative colloidal stability of NPs under centrifugation.We found OD values to be suitable for quantifyingsedimentation of S. aureus cells under centrifugation aftersystematically comparing bacterial enumeration via OD andCFU methods. Our insights and methodology will beimportant for future analytical developments for NP-bacteriasamples.While illustrated using a specific model system, all the steps

of our protocol can be readily adapted to NPs or bacterial cells

Figure 5. SEM images of the supernatant after differentialsedimentation of an NP-cell mixture. Following Scheme 1b, freeOL-HT2 NPs remained in supernatant at 4000 rpm. NPs appear asresolved bright objects (∼20 nm) or as aggregates; S. aureus cellsappear as spherical objects (∼700 nm). Panels b and c correspond tothe outlined regions in panels a and b, respectively.

Analytical Chemistry Article

DOI: 10.1021/ac503835aAnal. Chem. 2015, 87, 4641−4648

4646

that have somewhat different properties, particularly becausemost biomedical applications involve cells and NPs that arecolloidally stable under a common set of conditions and offer

possibilities for controlling the relative stabilities of the differentcomponents. Our approach thus should be generally applicable

Figure 6. SEM images of the resuspended pellet after differential sedimentation of an NP-cell mixture. After centrifugation at 4000 rpm, no free NPswere expected in the pellet (Scheme 1). S. aureus cells appear as spherical objects (∼700 nm); numerous bright NPs (∼20 nm) and loose NPaggregates are observed on cells but not on porous membrane support; panel b corresponds to the outlined region in panel a.

Analytical Chemistry Article

DOI: 10.1021/ac503835aAnal. Chem. 2015, 87, 4641−4648

4647

for preparing samples that will enable direct and quantitativeinvestigations of interactions between NPs and bacteria.

■ ASSOCIATED CONTENT*S Supporting InformationMaterials and methods, SEM images, SEM settings, EDXspectra, detailed comparison of OD640 and CFU data, statisticalanalysis of viability data, and sedimentation efficiency. Thismaterial is available free of charge via the Internet at http://pubs.acs.org.

■ AUTHOR INFORMATIONCorresponding Author*E-mail: dmitri.petrovykh@inl.int.NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTSC.S. acknowledges financial support from the following sources:grant SFRH/BPD/47693/2008 from the Portuguese Founda-tion for Science and Technology (FCT); FCT Strategic ProjectPEst-OE/EQB/LA0023/2013; project “BioHealthBiotech-nology and Bioengineering approaches to improve healthquality”, Ref. NORTE-07-0124-FEDER-000027, cofunded bythe Programa Operacional Regional do Norte (ON.2−O NovoNorte), QREN, FEDER; project “Consolidating ResearchExpertise and Resources on Cellular and Molecular Bio-technology at CEB/IBB”, ref. FCOMP-01-0124-FEDER-027462. The authors thank Dr. Leonard Deepak Francis(INL) and Keriya Mam (FEI) for helpful discussions andassistance provided for SEM and EDX analysis.

■ REFERENCES(1) Huang, X.; El Sayed, M. A. J. Adv. Res. 2010, 1, 13−28.(2) Wang, W.; Li, S.; Mair, L.; Ahmed, S.; Huang, T. J.; Mallouk, T.E. Angew. Chem. 2014, 126, 3265−3268.(3) Chen, J.-S. Front. Mater. 2014, 1, 1−2.(4) Hajipour, M. J.; Fromm, K. M.; Ashkarran, A. A.; Jimenez deAberasturi, D.; de Larramendi, I. R.; Rojo, T.; Serpooshan, V.; Parak,W. J.; Mahmoudi, M. Trends Biotechnol. 2012, 30, 499−511.(5) Seil, J. T.; Webster, T. J. Int. J. Nanomedicine 2012, 7, 2767−2781.(6) Tallury, P.; Malhotra, A.; Byrne, L. M.; Santra, S. Adv. DrugDelivery Rev. 2010, 62, 424−437.(7) Syed, M. A. Biosens. Bioelectron. 2014, 51, 391−400.(8) Wang, J.; Gao, J.; Liu, D.; Han, D.; Wang, Z. Nanoscale 2012, 4,451−454.(9) Chang, Y.-C.; Yang, C.-Y.; Sun, R.-L.; Cheng, Y.-F.; Kao, W.-C.;Yang, P.-C. Sci. Rep. 2013, 3, 1863.(10) Lee, Y. H.; Wu, B.; Zhuang, W. Q.; Chen, D. R.; Tang, Y. J. J.Microbiol. Methods 2011, 84, 228−233.(11) Chwalibog, A.; Sawosz, E.; Hotowy, A.; Szeliga, J.; Mitura, S.;Mitura, K.; Grodzik, M.; Orlowski, P.; Sokolowska, A. Int. J. Nanomed.2010, 5, 1085−1094.(12) Hayden, S. C.; Zhao, G.; Saha, K.; Phillips, R. L.; Li, X.;Miranda, O. R.; Rotello, V. M.; El-Sayed, M. A.; Schmidt-Krey, I.;Bunz, U. H. J. Am. Chem. Soc. 2012, 134, 6920−6923.(13) Vishnupriya, S.; Chaudhari, K.; Jagannathan, R.; Pradeep, T.Part. Part. Syst. Charact. 2013, 30, 1056−1062.(14) Rodrigues, D.; Banobre-Lopez, M.; Espina, B.; Rivas, J.;Azeredo, J. Biofouling 2013, 29, 1225−1232.(15) Sondi, I.; Salopek-Sondi, B. J. Colloid Interface Sci. 2004, 275,177−182.(16) Chatterjee, S.; Bandyopadhyay, A.; Sarkar, K. J. Nano-biotechnology 2011, 9, 34.

(17) Gammoudi, I.; Faye, N. R.; Morote, F.; Moynet, D.; Grauby-Heywang, C.; Cohen-Bouhacina, T. Int. J. Chem. Mater. Sci. Eng. 2013,79, 607−613.(18) Afrikian, E. G.; Julian, G.; Bulla, L. Appl. Microbiol. 1973, 26,934−937.(19) Bergmans, L.; Moisiadis, P.; Van Meerbeek, B.; Quirynen, M.;Lambrechts, P. Int. Endod. J. 2005, 38, 775−788.(20) Pagnout, C.; Jomini, S.; Dadhwal, M.; Caillet, C.; Thomas, F.;Bauda, P. Colloids Surf., B 2012, 92, 315−321.(21) Chaney, L. K.; Jacobson, B. S. J. Biol. Chem. 1983, 258, 10062−10072.(22) Stolz, D. B.; Jacobson, B. S. J. Cell Sci. 1992, 103, 39−51.(23) Choksawangkarn, W.; Kim, S. K.; Cannon, J. R.; Edwards, N. J.;Lee, S. B.; Fenselau, C. J. Proteome Res. 2013, 12, 1134−1141.(24) Kim, S. K.; Rose, R.; Choksawangkarn, W.; Graham, L.; Hu, J.;Fenselau, C.; Lee, S. B. J. Nanopart. Res. 2013, 15, 2133.(25) Kolen’ko, Y. V.; Banobre-Lopez, M.; Rodríguez-Abreu, C.;Carbo-Argibay, E.; Sailsman, A.; Pineiro-Redondo, Y.; Cerqueira, M.F.; Petrovykh, D. Y.; Kovnir, K.; Lebedev, O. I.; Rivas, J. J. Phys. Chem.C 2014, 118, 8691−8701.(26) Maggi, C.; Lepore, A.; Solari, J.; Rizzo, A.; Di Leonardo, R. SoftMatter 2013, 9, 10885−10890.(27) Fukushima, H.; Katsube, K.; Hata, Y.; Kishi, R.; Fujiwara, S.Appl. Environ. Microbiol. 2007, 73, 92−100.(28) Kowalczyk, B.; Lagzi, I.; Grzybowski, B. A. Curr. Opin. ColloidInterface Sci. 2011, 16, 135−148.(29) Akbulut, O.; Mace, C. R.; Martinez, R. V.; Kumar, A. A.; Nie, Z.;Patton, M. R.; Whitesides, G. Nano Lett. 2012, 12, 4060−4064.(30) Salgueirino-Maceira, V.; Correa-Duarte, M. A.; Farle, M. Small2005, 1, 1073−1076.(31) Freitas, A. I.; Vasconcelos, C.; Vilanova, M.; Cerca, N. J. BasicMicrobiol. 2014, 54, 750−757.(32) Jett, B. D.; Hatter, K. L.; Huycke, M. M.; Gilmore, M. S.Biotechniques 1997, 23, 648−650.(33) Chapman, J. H. J. Bacteriol. 1945, 50, 201−203.(34) Greenberg, A., Clesceri, L., Eaton, A., Eds. In Standard Methodsfor the Examination of Water and Wastewater, 18th ed.; AmericanPublic Health Association: Washington, D.C., 1992.(35) Liao, C.-H.; Shollenberger, L. M. Lett. Appl. Microbiol. 2003, 37,45−50.(36) Pankey, G. A.; Sabath, L. D. Clin. Infect. Dis. 2004, 38, 864−870.(37) Barbosa, H. R.; Rodrigues, M. F. A.; Campos, C. C.; Chaves, M.E.; Nunes, I.; Juliano, Y.; Novo, N. F. J. Microbiol. Methods 1995, 22,39−50.(38) Sutton, S. J. Validation Technol. 2011, 17 (3), 42−46.(39) Sutton, S. J. Validation Technol. 2011, 17 (1), 46−49.(40) Koch, A. L. Anal. Biochem. 1970, 38, 252−255.(41) Kumar, A.; Pandey, A. K.; Singh, S. S.; Shanker, R.; Dhawan, A.Cytometry 2011, 79A, 707−712.(42) Yao, Z.; Carballido-Lopez, R. Annu. Rev. Microbiol. 2014, 68,459−76.

Analytical Chemistry Article

DOI: 10.1021/ac503835aAnal. Chem. 2015, 87, 4641−4648

4648

S1

SU P P O R T IN G IN F O R M AT IO N F O R

Analytical Protocols for Separation and Electron M icrosco py of N anoparticles Interacting w ith Bacterial Cells Cláudia Sousa,† Diana Sequeira,† Yury V. Kolen’ko,‡ Inês Mendes Pinto,‡ Dmitri Y. Petrovykh*,‡ †Center of Biological Engineering, University of Minho, Braga 4710-057, Portugal ‡International Iberian Nanotechnology Laboratory, Braga 4715-330, Portugal * Corresponding author email: dmitri.petrovykh@inl.int

S 1 E XP E R IM E N TA L S E CT IO N S1.1 SEM observation of S. aureus and NPs. After cen-

trifugation, each sample (supernatant or resuspended pellet) was filtered through a wetted sterile 25-mm-diameter polycar-bonate filter with ca. 0.2 µm pores (nuclepore track-etched membranes, Whatman, GE Healthcare Life Sciences), mount-ed in a funnel assembly connected to a vacuum pump. The material retained on the filter was fixed in aqueous 2.5% glu-taraldehyde for 1 h. After fixing, the filters were washed with water three times for 10 min each time. The samples were

subsequently dehydrated through a graded series of ethanol solutions [10%, 25%, 50%, 75%, 90% and 100% (vol/vol)] for 10 min in each solution and kept in a sealed desiccator until imaging. A portion of each filter was imaged in a Quanta 650 (FEI Company) field-emission environmental SEM using settings detailed in Table S1. In addition, the elemental com-position of selected samples was analyzed by energy-dispersive x-ray (EDX) spectroscopy with energy of the inci-dent electron beam set to 5 kV or 20 kV.

Table S1. SEM settings for images presented in the figures

Figure HVa (kV) Detectorb Magnification factor HFWc (µm) WDd (mm) 1a 10 ET 30000 9.95 9.8 1b 10 ET 75000 3.98 9.8 1c 10 ET 138391 2.16 9.1 1d 15 ET 218102 1.37 10.3 1e 20 LF 30000 9.95 10.4 1f 10 ET 100000 2.98 10.0 5a 20 LF 30000 9.95 10.3 5b 20 LF 100000 2.98 10.3 5c 20 LF 200000 1.49 10.3 6a 10 ET 30000 9.95 8.7 6b 10 ET 100000 2.98 8.7 6c 10 ET 129684 2.30 8.7 6d 10 ET 154221 1.93 8.7 6e 10 ET 200000 1.49 8.7 6f 10 ET 200000 1.49 8.7 S1a 15 ET 1000 298 9.3 S1b 15 ET 3000 99.5 9.3 S1c 15 ET 5000 59.7 9.3 S2a 15 ET 5000 59.7 9.9 S2b 10 ET 5000 59.7 8.7 S7a 20 LF 30000 9.95 10.3 S7b 20 LF 75000 3.98 10.3 S7c 10 ET 200000 1.49 8.9

ahigh voltage bEverhart-Thornley (ET) high-vacuum detector or large field (LF) low-vacuum detector chorizontal field width dworking distance

S2

Figure S1. SEM images of S. aureus bacteria supported on track-etched polycarbonate membranes. Under these imaging conditions (Table S1) bacterial cells appear as bright spherical objects <1 µm in size, polycarbonate support is visible as a uniform gray background, and sharply-defined track-etched pores appear as dark circles of ca. 0.2 µm in diameter. Individual sub-µm bacterial cells are clearly resolved in images spanning a wide range of FOV sizes (a–c); images cropped for side-by-side comparison.

S3

Figure S2. Different concentrations of S. aureus cells imaged by SEM. The number of cells resolved in the same FOV size under similar imaging conditions (Table S1) can span roughly two orders of magnitude (a, b, Figure S1c); images cropped for side-by-side comparison.

S4

S1.2 Nanoparticles with Organic Coatings. Magnetite NPs were synthesized by controlled coprecipitation (CP) in an automated reactor or by a hydrothermal method (HT), as described previously. [Ref 25] Stoichiometric amounts of FeCl2·4H2O and FeCl3·6H2O were used as iron precursors, while concentrated ammonia hydroxide solution was used as a precipitation agent. Colloidal stability of the resultant NPs was achieved by functionalizing their surfaces during the synthesis with poly(acrylic acid) (PAA) or oleate (OL) capping ligands. The final products appeared as brownish-black colloidal dis-persions in the solvent. While PAA-coated NPs can be dis-persed in water, OL-coated NPs are hydrophobic as-synthesized and thus can be dispersed only in nonpolar organ-ic solvents. The phase transfer of OL-coated NPs from organic phase to aqueous solution was accomplished via a surfactant-based procedure using tetramethylammonium hydroxide (TMAOH). [Ref 30]

Throughout this work, the NPs are denoted by the primary capping ligand (PAA or OL) and the synthesis protocol (CP or HT). The differences between the numbered samples are as follows: synthesis of PAA-CP2 was performed with double concentration of the starting reagents compared to that used for PAA-CP1; OL-HT2 was treated with a higher TMAOH surfactant concentration than that used for OL-HT1.

S1.3 Colloidal Stability of NPs. Sedimentation threshold for each of the NPs was used to evaluate their colloidal stabil-ity in four aqueous solutions: NaCl 0.9%, NaCl 0.45%, 0.01 M phosphate buffered saline, pH 7.2 (PBS 0.01M), and dis-tilled water.

Prior to adding NPs, all solutions were sterilized by auto-claving for 20 min at 121 °C. Before centrifugation, disper-sions of each NP were sonicated for 10 min, at 50 kHz (SC-52 ultrasonic bath, Sonicor, USA), to break up loose aggregates of NPs. Dilutions were prepared by adding 5 µL of the disper-sion of NPs (ca. 10 g/L) to 4995 µL of the respective aqueous solution, in a 15 mL tube, to produce the final aqueous disper-sion of NPs.

Static precipitation was tested as a means of evaluating the colloidal stability of NPs (Figure S3), but provided only lim-ited information. Specifically, instability was readily apparent for NP suspensions that changed color and formed precipitates (Figures S3b and S3c), but the degree of stability was difficult to quantify or compare for the more stable combinations (Fig-ure S3a). Accordingly, centrifugation was chosen to quantify colloidal stability of NP suspensions.

Figure S3. Static precipitation of PAA-CP1 NPs in different aqueous solutions. No color change or precipitate was observed overnight for NP suspension in water (a), both color change and precipitates were observed in NaCl 0.45% (b) and NaCl 0.9% (c), and minimal precipitation was observed in PBS 0.01M (d).

For each value of the centrifugation rate from 1000 and up to 8000 rpm, three separate 5 mL samples (of every NP-solution pair) were centrifuged for 5 min. If sedimentation was not observed at a given rate, the next sample of the same NP-solution pair was centrifuged at a higher rate (in increments of

500 rpm). The process continued until reaching the rate at which sedimentation first occurred; this rate was recorded for each NP-solution pair. We emphasize that to avoid contribu-tions from sample history, a different aliquot was tested at each centrifugation rate rather than merely subjecting the same aliquot to progressively increasing centrifugation rates. In triplicate experiments for each NP-solution pair, sedimentation first occurred always at the same rate; the corresponding val-ues are presented in Table 1. In effect, the abrupt sedimenta-tion behavior for each NP sample allowed us to quantitatively parameterize the colloidal stability of each NP sample with a single number (Table 1) rather than with a more complex function (Scheme 1a).

The trends in colloidal stability as a function of salt type and concentration in Table 1 suggest that different mechanisms dominate the stabilization of the five NP samples analyzed in this work. PAA-CP1, OL-HT1, and OL-HT2 share the pat-tern of low stability in all the salt solutions and increased stability in water, suggesting predominantly electrostatic stabilization. More complex stabilization mechanism, includ-ing possible steric effects, is indicated by the patterns observed for PAA-CP2 and PAA-HT, whereby the stability differs among the salt solutions and stability in water mirrors that in PBS 0.01M.

S1.4 Staphylococcus aureus Suspensions. For each exper-iment, Staphylococcus aureus (S. aureus) ATCC 25923 (ATCC, American Type Collection Culture) was subcultured on Tryptic Soy Agar (TSA, Merck, Germany) for approxi-mately 36 h at 37 °C and then grown for 18 h in 30 mL of Tryptic Soy Broth (TSB, Merck, Germany) at 37 °C under a constant agitation at 120 rpm. Cells were harvested by centrif-ugation (Sigma 3-16K, Sigma Laborzentrigugen GmbH, Ger-many) at 8000 rpm at 4 °C for 5 min, the supernatant was discarded, and the pellet resuspended in the desired aqueous solution and centrifuged twice at 8000 rpm at 4 °C for 5 min. The resultant pellet was resuspended in 30 mL of the same aqueous solution and sonicated (Cole-Parmer® 750-W Ultra-sonic Homogenizer, employing a 13-mm microtip) using the following cycle: (20 s + 40 s) at 30% followed by (40 s + 40 s) at 40%, for disruption of cell clusters without impairing cell viability. [Ref 31] For the subsequent assays, the cellular suspension in the different aqueous solutions was adjusted to a specific optical density at 640 nm (OD640) as determined using a microplate reader (Synergy HT, Biotek®).

S1.5 S. aureus Viability. Viability of S. aureus cells was evaluated in aqueous solutions of the same composition as those used to test the colloidal stability of NPs: NaCl 0.9%, NaCl 0.45%, PBS 0.01M, and water. Bacterial suspensions, prepared as described in the previous section, were adjusted to an OD640≈0.1 in each aqueous solution, and then left to incu-bate at room temperature for 1 h or 24 h. After the incubation, 10-fold serial dilutions were prepared by sequential transfer of 100 µL into 900 µL of the aqueous solution. A 10 µL sample from each dilution was deposited on a TSA Petri dish, in trip-licate, and the plate was tilted until the drops migrated across the agar surface to the opposite side of the plate. [Ref 32] After drying, the plated samples were incubated 24 h at 37 °C and the number of colonies was counted. The plates at the dilution that produced a countable number of colonies were selected for comparing the colony forming unit (CFU) counts

S5

among the samples. [Ref 32] Most of the successful assays produced 9 countable replicates; the number of countable replicates was not less than 6 in all cases.

In all our viability assays countable plates were produced at the same dilution, therefore, the CFU counts could be com-pared directly (Figure 2, Figure S4). We note that converting CFU counts to a representation as nominal CFU/mL values is commonly done in microbiology [Refs 35, 36, 38] largely because that representation is convenient for comparing CFU counts obtained from plates of different dilutions. In particu-lar, despite the CFU/mL values having the units of concentra-tion, the CFU terminology underscores that this method does not guarantee measuring concentrations of individual bacteria. In our case, directly comparing CFU counts also emphasizes the counting nature of CFU-based assays and informs the choice of the appropriate statistical methodology for analyzing the results, as discussed in more detail below.

Figure S4. Viability of S. aureus after 1 h incubation in water or in salt solutions. For each solution, CFU counts (in the same dilution) are shown for assays carried out on 3 separate days (b–d) and as a pooled mean value (a). Error bars indicate 95% CIs; a horizontal dashed line in each panel indicates the mean value for the panel and serves as a guide to the eye for comparing the data points.

Two types of criteria are commonly used to define a “signif-icant difference” when comparing data from microbiological experiments. (1) Most commonly, in bacterial enumeration experiments the minimum bactericidal concentration (or an-other similarly-defined parameter) is established based on observing a ≥99.9% (or ≥99%, for some applications) decrease in the treated bacterial population compared to the untreated control (or the value before treatment). In other words, only differences ≥3-log10 (or ≥2-log10, respectively) in CFU/mL values are considered to be significant in such CFU-based assays. [Ref 36, 38] Accordingly, the CFU/mL values for such measurements are typically plotted and compared on a log10 scale spanning multiple decades. (2) Alternatively, statistical definitions, e.g., based on CIs, can be used to identify signifi-cant differences among CFU data.

The log10 representation had been used in a previous com-parison of S. aureus viability over several weeks in two differ-ent storage solutions (water vs PBS). [Ref 35] Attempting to present our data for individual assays (days) in the log10 for-mat provides an apparent indication of the quality of our data: error bars corresponding to 95% CIs can only be clearly seen with expanded vertical scales (Figure S4 b–d). The reproduci-bility of same-day replicates is commonly used as a quantita-tive quality-assurance indicator in CFU counting [Refs 37, 38]. For our measurements, variation among same-day repli-cates is in the ±10% to ±20% range (Figures 2 and S4 b–d), which is in line with best-practice measurements [Refs 32, 38]; the data after 24 h incubation (not shown in detail) are of similar quality. Given the 2 log10 range of the panels in Figure S4, the differences (intra- or inter-day) among the S. aureus viabilities in the four solutions clearly are not significant under either the ≥3-log10 or ≥2-log10 criteria described above.

For a more formal and quantitative statistical analysis of our viability data (Figures 2 and S4), the appropriate procedure needs to be identified for pooling the data from the 3 separate assays (days). Such a comparison is not commonly done in the context of viability studies, but it provides some interesting insights. First, we note that the same-day variation among replicates for the same solution or among the means for differ-ent solutions is comparable and relatively small (in the ±10% to ±20% range, Figures 2 and S4 b–d). In contrast, the varia-tion among replicates of the same solution or among the means for different solutions is much larger than what may be expected from same-day values.

Having identified two scales of variability, we clearly would lose important inferences about its sources if the data from 3 days were considered simply as part of the same large set. The 95% CIs calculated under such a simplistic pooling assump-tion become comparable to the mean values, essentially sug-gesting that any comparison among the data more detailed than that under the ≥3-log10 or ≥2-log10 criteria is not statisti-cally meaningful.

By considering the CFU counts from each day as separate attempts to estimate the common underlying quantity (in this case, the true CFU concentration of a sample adjusted to OD640≈0.1), we realize that the inter-day variability serves as an indication of a systematic measurement uncertainty, which biases the measurement in a specific assay (among different days or different solutions) from the true value of the underly-ing bacterial concentration. Establishing the nature of this

Log 10

(CFU

s)

H 2O

NaCl

0.4

5%

PBS

0.0

1M

NaCl

0.9

%

Pooled

10

100

1000

Day 1

10

100

1000

Day 2

Log 10

(CFU

s)

10

100

1000

10

100

1000

H 2O

NaCl

0.4

5%

PBS

0.0

1M

NaCl

0.9

%

Day 3

(a)

(b)

(c)

(d)

S6

systematic uncertainty is beyond the scope of this work, but it has been important to repeat the assays over 3 different days to get an indication that this contribution is more likely systemat-ic than random.

In contrast, the intra-day variability provides an estimate of the random contribution to the measurement uncertainty. It may be improved, to some extent, by adding more same-day replicates, but likely has a lower limit determined by the in-trinsic variability of bacterial growth experiments.

We do not have a clear indication regarding the sign of the systematic bias revealed by inter-day comparisons, so the pooled mean value is still the best estimate of the true CFU count that corresponds to the underlying CFU concentration. The CI calculation is different, however, because it is calculat-ed by pooling the standard deviations of the 3 separate meas-urements (as a root-mean-square) and using the degrees of freedom of the smallest of the 3 pooled sets to calculate the Student’s t statistic (following the statistical procedure for sets with unequal variances). In other words, the pooled mean and CI values in Figures 2 and S4a provide the best estimate of what an unbiased (i.e., properly controlled for all systematic factors) measurement would produce for each solution. We note that while the systematic bias appeared to be correlated among the 4 solutions on a specific day in our measurements (Figures 2 and S4 b–d), without a better understanding of its origins, the sign and magnitude of the systematic bias can not be assumed to be day-specific: statistical conclusions regard-ing the effect of the different solutions on S. aureus viability, for example, would be different based on considering the data in Figure 2 (or Figures S4 b–d) for just one of the days.

Considering the pooled values, we note an apparent trend in CI values across the solutions, with the two “low-salt” condi-tions (water and NaCl 0.45%) having the largest CIs and PBS 0.01M having the smallest CIs. While a rigorous confirmation of this effect would require a dedicated study, we can specu-late that a practical motivation for using PBS 0.01M for han-dling S. aureus could be a reduced variability in CFU counts rather than an overall enhancement of viability.

The final speculation regarding the sign of the systematic bias is suggested by observing that the CIs are the lowest (in both absolute and relative terms) for the lowest mean CFU counts (Figure 2). A possible inference from this trend is that the systematic bias may randomly increase the CFU count in a given assay, while the lowest observed values correspond to minimally biased measurements. As with the other specula-tions offered above, formally investigating this hypothesis would require a dedicated study that falls beyond the scope of this work.

S1.6 Sedimentation Efficiency for S. aureus. In the prima-ry systematic evaluation, sedimentation efficiency for S. aure-us in water was determined starting with a cellular suspension adjusted to OD640≈0.6 (Figures 4 and S5), OD640≈0.1 (Figure S6a), or OD640≈0.2 (Figure S6b). For each value of the cen-trifugation rate (1000–8000 rpm), two separate 5 mL samples of S. aureus suspension were centrifuged for 5 min. To con-firm the results from the main series for the initial OD640≈0.6 (Figure 4), measurements were repeated in triplicate at 1000, 4000, and 6000 rpm (Figure S5). After centrifugation, the OD640 was measured for the supernatant, which was aseptical-ly decanted to a new tube and vortexed for ca. 60 s, and the

pellet, which was resuspended in 5 mL of water, vortexed (ca. 60 s), and sonicated ((20 s + 40 s) at 30% followed by (40 s + 40 s) at 40%) to disrupt cell clusters without impairing viabil-ity. [Ref 31]

Figure S5. Sedimentation efficiency for S. aureus at selected centrifugation rates. The rates have been selected to confirm the trend observed in Figure 4. For each rpm value, OD640 was meas-ured for a separate aliquot of the initial suspension. The measured OD640 values for the resuspended pellet are represented as per-centage fractions of the initial OD640≈0.6; error bars indicate 95% CIs, n = 3 from three separate experiments.

Figure S6. Sedimentation efficiency for S. aureus from suspen-sions of low initial concentrations. For each rpm value, OD640 was measured for a separate aliquot of the initial suspension. The measured OD640 values for both supernatant (gray bars, left axis) and resuspended pellet (red bars, right axis) are represented as percentage fractions of the initial OD640≈0.1 (a) or OD640≈0.2 (b); error bars indicate 95% CIs, n = 2 from two separate experiments.

Initial OD6405 0.6

Centrifugation rate (rpm)1000 4000 6000

Aver

age

% P

elle

t

100

75

50

25

0

(a)

400

300

1000

2000

3000

4000

5000

6000

7000

8000

200

100

0

-100

-200

-300

Initial OD6405 0.1-300

-200

-100

0

100

200

300

400

Aver

age

% S

uper

nata

nt Average % Pellet

Supernatant Pellet

(b)

Initial OD6405 0.2

1000

2000

3000

4000

5000

6000

7000

8000

Centrifugation rate (rpm)

Supernatant Pellet

400

300

200

100

0

-100

-200

-300

Aver

age

% S

uper

nata

nt

-300

-200

-100

0

100

200

300

400

Average % Pellet

S7

Table S2. Correlation between OD and CFU measurements

centrifugation rate (rpm)a

initial S. aureus suspension supernatantb pelletc

OD640 CFU OD640 CFU OD640 CFU

6000 0.08 109 0.05 0.05 22 6000 0.09 48 0.05 2 0.05 8 6000 0.12 54 0.05 0.05 6 6500 0.06 28 0.02 0.05 8 6500 0.08 27 0.04 2 0.06 11 6500 0.12 48 0.05 0.05 9 7000 0.07 30 0.04 0.05 28 7000 0.08 29 0.04 11 0.05 28 7000 0.11 40 0.04 0.05 46 7500 0.06 61 0.04 0.04 44 7500 0.07 66 0.05 36 0.07 71 7500 0.13 101 0.04 0.10 67 aThree solutions prepared to differ by 20% in bacterial concentration were aliquoted into tubes; each tube was centrifuged at a specific rate. bSupernatant obtained after centrifugation; CFU measurement was performed for only one sample at each centrifugation rate. cAfter centrifugation each pellet was redispersed by vortexing followed by sonication.

Table S3. Sedimentation efficiency for S. aureus

assay centrifugation rate (rpm)a

OD 640 % pelletd

initial S. aureus suspension supernatantb pelletc

assay 1 1000 0.66 0.54 0.11 17 2000 0.65 0.30 0.30 46 3000 0.66 0.06 0.47 71 4000 0.65 0.04 0.60 92 5000 0.63 0.02 0.52 83 6000 0.64 0.01 0.59 92 7000 0.64 0.01 0.58 91 8000 0.64 0.00 0.55 92

assay 2 1000 0.71 0.59 0.16 23 2000 0.72 0.21 0.50 69 3000 0.71 0.15 0.54 76 4000 0.71 0.02 0.62 87 5000 0.71 0.04 0.62 87 6000 0.71 0.01 0.64 90 7000 0.72 0.02 0.69 96 8000 0.71 0.01 0.68 96

aEight separate S. aureus aliquots were prepared and each aliquot was centrifuged at a specific rate. bSupernatant obtained after centrifugation cAfter centrifugation each pellet was redispersed by vortexing followed by sonication. dPercentage of pellet was calculated relative to the OD640 of the initial S. aureus suspension.

S8

The primary purpose of the replicate measurements for the initial OD640≈0.6 summarized in Figure S5 was to confirm the characteristics of the sedimentation pattern in Figure 4 (Table S3) that are critical for performing the differential sedimenta-tion outlined in Scheme 1. The measurements at 4000 and 6000 rpm confirm the indication of Figure 4 that the sedimen-tation efficiency plateaus at 4000 rpm and above, thus support-ing the choice of 4000 rpm as the rate for the differential sed-imentation experiment. The measurement at 1000 rpm con-firms both the low sedimentation of S. aureus cells at that rate and the ability of our methodology to reliably measure pellets of that size (as a fraction of the initial OD640≈0.6). The latter observation is in agreement with the resolution tests of Figure 3 (Table S2), as the ca. 25% pellet in Figure S5 corresponded to OD640≈0.15, validating our choice of OD640≈0.6 as a starting concentration for reliable sedimentation efficiency measure-ments.

In contrast to the reliable sedimentation measurements start-ing from OD640≈0.6, lower initial concentrations result in largely inconclusive measured values (Figure S6), primarily because the absolute OD640 values measured for both the su-pernatant and resuspended pellet are too close to the limit of OD640≈0.1 identified in Figure 3 (Table S2). We note that the magnitude of the 95% CI error bars in Figure S6 is partly due to the small number of replicates (n = 2); unfortunately, the time-consuming nature of such measurements limits a practi-cal number of replicates that can be performed for the full series (hence, the measurements in Figure S5 only partially replicated the range of Figure 4).

Their large uncertainties notwithstanding, the data in Figure S6 do indicate that at least some fraction of S. aureus cells sediments at ≥4000 rpm from suspensions with initial concen-trations of OD640≈0.1 and OD640≈0.2. Unlike in Figure 4, however, the sum of the measured supernatant and pellet frac-tions in Figure S6 is not close to 100% for ≥4000 rpm range (as indicated by vertical separation between the top/bottom edges of the red/gray bars in each pair), suggesting that one (or both) of the measured values is (are) underestimated.

The results summarized in Figures 4, S5, and S6 highlight a general challenge in developing, validating, and performing analytical protocols that involve measurements and systems measured both in solution and on surfaces (or supports, micro-fluidic channels, or any other arrangements limiting the total sample size). S. aureus suspensions with in the OD640≥0.6 range where we can readily validate the sedimentation proto-col (Figures 4 and S5) produce samples with an excess of cells (e.g., note the multilayer agglomerates of cells in Figures 1a and 1b) for SEM imaging. Reducing the initial cell concentra-tion to the levels compatible with subsequent SEM imaging, however, makes the samples difficult to measure by solution methods (Figures 3 and S6). Addressing this challenge in a systematic manner requires a more detailed investigation that falls beyond the scope of the current work, so here we focus on the ramifications for the goals of this investigation.

As indicated in the main text, assuring >90% sedimentation efficiency is desirable but not indispensable for our objective of producing a representative sample of bacteria and any NPs that directly interacted with them during co-incubation. A minimal criterion for a successful differential sedimentation would be to observe a large enough fraction of the original cell

suspension in a pellet to ensure that the sedimented fraction is not dominated by mutants or other, similarly low probability, artifacts, but rather is representative of the cell population that we intend to analyze. We believe that recovering about 50% of the bacteria from the mixed co-incubation sample would be sufficient to satisfy this practical criterion. As discussed above, the data in Figure S6 do not rule out that about 50% of the bacteria sediment at ≥4000 rpm even from the suspensions with low initial concentrations, so it should be reasonable to proceed with attempts to produce the differential sedimenta-tion outlined in Scheme 1.

S1.7 Correlation between OD and CFU measurements. In typical microbiology experiments, a correlation is assumed between bacterial enumeration using OD and CFU methods. [Ref 39] Verifying this correlation is obviously important, particularly when the system of interest has special properties. In our case, both the low OD640 values and centrifuga-tion/resuspension steps can affect the measurements by one or both of the methods and thus their correlation.

The initial OD640 measurements (“initial S. aureus suspen-sion” columns in Table S2) for three solutions prepared to differ by 20% (based on dilution factors) in bacterial concen-tration were used to compare the resolution of OD and CFU measurements in this concentration range, as discussed in the main text (Figure 3). Having found that only the OD values exhibited the expected trend of the means (statistically signifi-cant for OD640≥0.1) across this model series, we had selected OD measurements as appropriate for quantifying the sedimen-tation efficiency for S. aureus in water (Table S3, Figures 4, S5, and S6).

Aliquots of the same three solutions were then centrifuged at four different rates, as indicated in Table S2. The rates ≥6000 rpm were chosen to maximize the sedimentation effi-ciency even at the low initial OD640≈0.1 of these test samples (Figures 4, S6). While the analysis of the data in Figure 3 had already suggested that OD640 and CFU measurements are not strongly correlated in this concentration range, the following comparison of the OD640 and CFU values measured for resus-pended pellets and for selected supernatant samples summa-rized in Table S2 provides additional insights, and supports the use of OD640 rather than CFU values for quantifying sedimen-tation efficiencies (e.g., in Figure 4).

The OD640 values for supernatant and resuspended pellet in each row of Table S2 add up to approximately the OD640 val-ues measured before centrifugation. While the agreement is not perfect, it is consistent with uncertainties analyzed in Fig-ure 3 and clearly indicates that the same intrinsic characteristic of S. aureus suspensions is measured by OD before and after centrifugation.

In contrast, the CFU counts are essentially uncorrelated be-tween the initial suspension and centrifuged samples: in most cases the values differ dramatically between the initial suspen-sion and the pellet and adding up the CFU counts for pellet and supernatant does not account for the difference (Table S2). Clearly, the CFU counts do not measure the same intrinsic characteristic of S. aureus suspensions as that measured by OD. Neither are CFU counts appropriate for sedimentation efficiency measurements in this concentration range, as they produce essentially random “% pellet” values.

S9

Figure S7. SEM images of the supernatant after differential sedi-mentation of an NP-cell mixture. Following Scheme 1b, free OL-HT2 NPs remained in supernatant at 4000 rpm. NPs appear as resolved bright objects (ca. 20 nm) or as aggregates; S. aureus cells appear as spherical objects (ca. 700 nm).

A speculative inference from the CFU differences before and after centrifugation is that the pellet resuspension process (both after the final centrifugation and after the preceding washing steps) does not produce a consistent population of CFUs in suspension (e.g., average number of bacteria per CFU varies among samples). The OD measurements, in contrast, are much less sensitive to cell aggregation and thus are not affected as strongly as CFU counts.

S1.8 Removal of Free NPs by Differential Sedimentation. The parameter window for removing the excess free NPs from mixed NP–bacteria samples was determined based on the separate sedimentation measurements for NPs and S. aureus.

Equal volumes (2.5 mL) of S. aureus suspension (in water) adjusted to OD640≈0.2 and OL-HT2 colloid (also in water) at ca. 2 mg/mL concentration were mixed to produce an excess of NPs and thus maximize the possibility for NP–cell interac-tions. The mixture was then incubated for 1 h, homogenized by vortexing for 60 s, and centrifuged at 4000 rpm for 5 min at 4 °C. The resultant pellet was aseptically resuspended by vortexing in distilled water. Both pellet and supernatant were withdrawn for SEM analysis.

Panels of Figure S7 show evidence (complementary to that in Figure 5) of the structures being formed during the sample preparation for SEM rather than during the NP-cell co-incubation. Drying cracks are observed in a dense layer of NPs in the top-right corner of panel (a). Large, thick, three-dimensional NP aggregates are observed on and around cells in panels (a) and (b), on the bare porous substrate away from cells in panel (a), and as apparently self-supported three-dimensional arrangement in the center of panel (a).

Panel (c) of Figure S7 may appear contradictory to the above interpretation, as the cluster of cells in the foreground has very few NPs attached, i.e., shares an appearance with cells in the resuspended pellet samples in Figure 6. We note, however, that the cells visible in the background (particularly along the bottom edge of the panel) are covered with a dense layer of NPs, similar to that seen in the other supernatant sam-ples. Therefore, rather than indicating a strong avoidance of those specific cells by the NP aggregates during filter-ing/drying steps, the foreground cell cluster then is most likely a result of secondary re-deposition of cells during filtering, i.e., it probably has arrived to this location after the majority of the NPs have been deposited.

S1.9 EDX Analysis. Representative samples of the superna-tant and resuspended pellet after differential sedimentation were analyzed by EDX to confirm the effective removal of the free NPs by our protocol.

Resuspended pellet sample (Figure S8) provides direct evi-dence of the free NPs being nearly completely removed from the original NP-cell mixture. An area with a single isolated cell surrounded by bare supporting nanoporous membrane was selected for the EDX analysis to ensure that any free NPs present within the analysis region would be clearly visible. Some residual charging artifacts could not be completely removed under conditions optimized for the subsequent EDX analysis (Figure S8a), but their presence does not prevent us from clearly observing only a minimal number of free NPs in the entire area, as expected (Figure 6) after their successful removal by differential sedimentation (Scheme 1).

S10

Figure S8. SEM image and EDX spectra of the resuspended pellet after differential sedimentation of an NP-cell mixture. (a) As expected, only a few free NPs are visible on the bare sup-port surrounding an isolated S. aureus cell, which appears as a spherical object (ca. 700 nm). Edges of the ca. 200 nm pores are highlighted because of the residual charging artifacts on this sample under the imaging conditions that optimize the lateral localization of EDX analysis. EDX spectra are shown for nominal regions centered over the isolated cell (R1) and bare support (R2). (b) EDX spectrum centered over the isolated cell and acquired at 20 kV primary beam energy.

The EDX spectra acquired over the nominal regions cen-tered over the isolated cell (R1) and bare support (R2) in Fig-ure S8a are nearly identical, apart from the minor (<5 at. %) N and P signatures of the cell detected in the R1 region. The difference between the R1 and R2 spectra confirms that the EDX analysis regions in this measurement have been strongly localized at the scale comparable to the sizes of the nominal R1 and R2 regions selected in the software.

Notably, the Fe signal that would be expected if a signifi-cant number of iron oxide NPs were present in the sample is not detected in either of the two regions. Selecting larger areas over either groups of cells or bare support produced essentially the same EDX spectra (data not shown).

Figure S9. SEM image and EDX spectra of the supernatant after differential sedimentation of an NP-cell mixture. (a) As expected, both individual and aggregated free NPs are observed on top of S. aureus cells (ca. 700 nm spherical objects) and the nanoporous support (pores appear as dark ca. 200 nm circular objects). EDX regions are shown for nominal regions centered over a cell cov-ered by aggregated NPs (R1), an NP aggregate without a cell present (R2), bare support (R3), and a cell with a small number of visibly attached NPs.

To verify that a large number of NPs is not present either under or inside the cell in Figure S8a, we acquired an addi-tional EDX spectrum using the primary electron beam of a much larger energy (20 kV, rather than 5 kV used in Figure S8a), which dramatically increased the depth of analysis (Fig-ure S8b). Observing the signal from the Cu mounting tape under these conditions confirms that the entire thickness of the cell and the membrane is sampled by EDX, yet no Fe signal is detected. This measurement was repeated in several additional areas (data not shown) with analogous results. While the sensi-tivity of EDX measurements would limit our ability to detect a small (ca. 10) number of NPs, a larger number (ca. 100) would be clearly visible, as confirmed by the measurements of the sample in Figure S9. The absence of the Fe signal in Figure S8b then indicates that only a minimal number of NPs (if any) had been internalized by the S. aureus cells during our exper-iments. This observation is not unexpected, as the model NPs had been selected based on their colloidal stability (Table 1) and had not been modified to promote either specific attach-ment to or internalization by the model S. aureus bacteria.

The supernatant sample in Figure S9 provided complemen-tary information regarding the use of EDX for analyzing our

S11

mixed NP-cell samples. In agreement with the observations summarized in Figures 5 and S7, individual and aggregated free NPs are clearly seen in the SEM image of the selected area (Figure S9). We have acquired EDX spectra from regions nominally centered over the areas indicated as R1–R4 in Fig-ure S9. These four areas had been selected based on their visibly different nominal content: a cell covered by aggregated NPs (R1), an NP aggregate without a cell present (R2), bare support (R3), and a cell with a small number of visibly at-tached NPs. Comparing the EDX spectra for regions R1–R4, we observe the following systematic characteristics.

In agreement with the results in Figure S8, small (<5 at. %) N and P signals are associated with the areas that include a cell, whether accompanied by a large (R1) or a small (R4) number of visible NPs. Conversely, the strongest combination of Fe and O signals is associated with the areas that include the large (> 100) number of visible NPs, whether accompanied by a cell (R1) or not (R2). These correlated observations com-bined clearly indicate that the EDX spectra do include a signif-icant (if not dominant) contribution from the nominal specified analysis regions. They also directly confirm our ability to detect a large number (>100) of NPs, supporting our preceding conclusion about the absence of such large numbers of NPs in the sample of Figure S8.

In contrast to the apparently high localization of the EDX analysis region in Figure S8, however, a broadening effect is apparent in the EDX data from Figure S9, as small but clear Fe signals are observed in nominal areas R3 and R4 that do not contain a large number of visible NPs. Some broadening, sometimes referred to as “skirting,” can be expected under the low vacuum (90 Pa) and primary beam energy (5 kV) that were used for EDX in Figure S9 to minimize residual charging and other artifacts. Observing a mixture of contributions corre-lated and uncorrelated with the visible content of the nominal analysis areas indicates that the effective analysis area is likely up to 5 µm in diameter under these conditions. Significantly, the fact that a comparable Fe signal is observed for the cell in R4 and the blank support in R3 also strongly indicates that this signal is associated with lateral broadening of the analysis area in all cases, rather than with any putative internalization of a large number of NPs by the cell in R4.