Identification of Individual Bacterial Cells through the Intermolecular

Letter Cite This: Anal. Chem. XXXX, XXX, XXX−XXX

pubs.acs.org/ac

Identification of Individual Bacterial Cells through the Intermolecular Interactions with Peptide-Functionalized Solid-State Pores Makusu Tsutsui,†,§ Masayoshi Tanaka,‡,§ Takahiro Marui,‡ Kazumichi Yokota,† Takeshi Yoshida,† Akihide Arima,† Wataru Tonomura,† Masateru Taniguchi,† Takashi Washio,† Mina Okochi,*,‡ and Tomoji Kawai*,† †

The Institute of Scientific and Industrial Research, Osaka University, 8-1, Mihogaoka, Ibaraki, Osaka 567-0047, Japan Department of Chemical Science and Engineering, Tokyo Institute of Technology, 2-12-1, O-okayama, Meguro-ku, Tokyo 152-8552, Japan

‡

S Supporting Information *

ABSTRACT: Bioinspired pore sensing for selective detection of flagellated bacteria was investigated. The Au micropore wall surface was modified with a synthetic peptide designed from toll-like receptor 5 (TLR5) to mimic the pathogen-recognition capability. We found that intermolecular interactions between the TLR5-derived recognition peptides and flagella induce ligand-specific perturbations in the translocation dynamics of Escherichia coli, which facilitated the discrimination between the wild-type and flagellin-deletion mutant (Δf liC) by the resistive pulse patterns thereby demonstrating the sensing of bacteria at a single-cell level. These results provide a novel concept of utilizing weak intermolecular interactions as a recognition probes for single-cell microbial identification.

L

iving organisms protect themselves from pathogens through activation of an innate immune system that utilizes receptor molecules for pattern-recognition of extracellular entities through specific intermolecular interactions with conserved molecular structures.1−3 Human toll-like receptor 5 (TLR5) represents a germline-encoded pathogen sensor4 playing an important role in host defense against bacterial infections through recognizing their flagellin molecules.5−8 Inspired by the immunochemistry, we developed a solid-state sensor for label-free detections of single-bacteria that can identify the presence of flagella on individual microbes through analyzing an electrical pulse pattern associated with bacterial interactions. While pore sensors were generally investigated for various pathogen detection without any surface modification,9−13 the device consists of gold (Au) micropores functionalized with a screened peptide mimicking human tolllike receptor 5 (TLR5) (Figure 1a−c). Unlike the antibodydecorated nanopores used for the detection of analytes or ions bound to the channel surface through a change in the crossmembrane ionic current,14−16 we exploited artificial peptide sequence-derived weak interactions to add a ligand-specific perturbation to the translocation dynamics17−19 of individual bioparticles in the conduit for immunoselective single-bacteria identifications via resistive-pulse analyses.20 The peptide-based affinity sensor with a single-cell resolution should be useful for various applications in the environmental, medical, and nanobiological fields. A binding assay was conducted with the flagellated Escherichia coli (BW25113) cells for screening of active recognition peptides using a peptide array21,22 comprising © XXXX American Chemical Society

Figure 1. Peptide-embedded micropores. (a−c) Schematic depiction of the resistive pulse analysis using a peptide-functionalized micropore. (a) The ionic current Iion through an Au pore was measured under the applied voltage Vb. The channel wall surface was functionalized with flagellin-recognizing peptides (b) comprising an active sequence (FLLRVPHL), a spacer (GGG), and a cysteine anchor for the formation of Au-thiol bond (c). The molecular model was obtained by applying DFT calculations.

309 peptides of the sequential amino acid 8-mers derived from an extracellular region of TLR5. Its high bacteria-binding regions were found in a series of TLR5-derived peptides (Figure S1). Subsequently, a comparative peptide-binding assay was performed using the wild-type E. coli cells and the f liC deletion mutant strain (Δf liC) lacking the f liC gene that codes for the major flagellin component, wherein a conspicuous difference was observed in the affinity of the two bacteria to the FLLRVPHL peptide (position 443 from the N-terminus) Received: November 29, 2017 Accepted: January 15, 2018

A

DOI: 10.1021/acs.analchem.7b04950 Anal. Chem. XXXX, XXX, XXX−XXX

Letter

Analytical Chemistry

micropore sensors. Iion versus time (t) traces showed Iion spikes for both strains (Figure 3a) when positive Vb was applied, while the resistive pulses were shown to be absent after inverting the voltage polarity. Each Iion spike is therefore interpreted as stemming from the temporal blockage of the ion transport through the micropore by a bacterial cell expressing negatively charged phospholipids and lipopolysaccharides upon electrophoretic transit. Generally, as larger objects exclude more ions and bring more pronounced Iion drops, the height, Ip, of the ionic current signatures represents the size of the objects, whereas the width td reflects the translocation speed (Figure 3b; Figure S6).23 Pattern analyses of the resistive pulses can, therefore, provide a wealth of information concerning the presence or absence of the expected flagella−peptide interactions during the bacterial passing through the micropore. Roles of peptide molecules in the translocation dynamics of cells were first investigated by comparing the Ip−td scatter plots of the wild-type cells before and after the molecular functionalization (Figure 3c,d). We first note that the pulse height is becoming higher after the functionalization. Here, the wide variations in Ip is attributable to stochastic nature of the bacterial translocation dynamics that give rise to random orientations of their body with respect to the pore axis and associated difference in the amount of excluded ions by their volume (Figure S7). It is therefore anticipated that the rod-like cells point-contacted to the peptides tend to rotate via the electrophoretic forces to orient toward the radial direction of the channel. As for the pulse width, on the other hand, the plots revealed a significant retardation of the electrophoretic motions after the wall-surface modification using the flagella-binding peptide sequence (Figure 3c; Figures S8 and S9). An opposite trend was observed when the peptides lacking the active sites were used, with td becoming shorter, which indicates an increase in the speed of cells in the conduit (Figure 3b). These results demonstrate the key effects of the recognition probe on the mobility of bacteria transiting the conduit. A factor underlying the difference between the cell translocation dynamics of the linker and the peptide with the

(Figure 2a). Fluorescence observations with FITC-labeled peptides corroborated the specific binding of the active peptide

Figure 2. Screening of flagella-binding peptide from TLR5. (a) Screening of flagellum-binding peptides by comparing the fluorescence intensity after peptide binding to wild-type and Δf liC E. coli cells. Insets (left to right): representative fluorescence images of peptide spots containing FLLRVPHL and negative control peptide (GGG) after the binding assay. b Time-course analysis of E. coli cells bound to the FITC-labeled FLLRVPHL peptides. c Fluorescence images of the wildtype (left) and Δf liC (right) cells after binding of the FITC-labeled FLLRVPHL peptides.

to the flagellated cells (Figure 2b,c; Figure S2). These results indicate a specificity of this recognition molecule toward bacterial flagella. A peptide-functionalized Au micropore with a 3-μm diameter formed in a 50 nm thick Si3N4 membrane on a Si wafer was used for the detection of individual bacteria suspended in a diluted phosphate-buffered saline (×0.10 PBS) by a resistive pulse method. The cross-membrane ionic current Iion was recorded under the applied voltage Vb (Figure 1a,b; Figures S3 and S4). A peptide (12-mer) contained an active site, FLLRVPHL, a spacer (GGG), and an anchor (C) for the formation of gold−thiol bonds on the pore wall surface (Figure 1c; Figure S5). Wild-type E. coli and Δf liC strain cells were employed to verify the ligand-recognition capability of the

Figure 3. Detection of individual cells using the peptide-modified micropore. (a) Iion traces obtained for the wild-type (red) and Δf liC (blue) cells, showing current spikes corresponding to the translocation of single cells through the Au micropore before (two left curves) and after the addition of active peptides (two right curves). (b) Magnified views of the resistive pulses (red, wild-type; blue, Δf liC cells). Ip and td are the height and the width of the signals, respectively. (c,d) Resistive pulse height (Ip) versus width (td) scatter plots for the cells passing through the micropore containing active peptides (c) or a spacer (d), before (red) and after (green) the functionalization. B

DOI: 10.1021/acs.analchem.7b04950 Anal. Chem. XXXX, XXX, XXX−XXX

Letter

Analytical Chemistry

Figure 4. Peptide affinity to the bacterial flagella. (a,b) Translocation duration distributions obtained for Δf liC (a) and the wild-type cells (b) in an active peptide-functionalized micropore. Data obtained before the molecular functionalization are presented (gray). Gray and green curves represent a Gaussian fitting to the peak distributions. (c,d) Two-dimensional histograms of resistive pulses obtained for the wild-type cells with small (c) and large (d) widths using the peptide-modified Au surface during the translocation.

translocation of Δf liC after the functionalization is attributable to an increase in the nonspecific binding strength of the rodshaped body to the peptide-coated wall compared with that to the bare Au surface. Interestingly, on the other hand, the change in the translocation time was more pronounced for the wild-type cells, delivering a distinct distribution at td > 5 ms (Figure 4b). As expected, the active probes on the pore wall, designed to show the affinity toward bacterial flagella, bind strongly to one or more flagella protruding from the basal body.30 Further examination of the ionic traces showed that not all wild-type cells demonstrated a high affinity toward the peptide probes (Figure S10). For example, short spikes with td < 3 ms were shown to have single-pulse waveforms resembling to those recorded for the Δf liC strain and suggesting a nonspecific weak interactions between the wild-type cells and the probes during their transit (Figure 4c). However, the wider resistive pulses (td > 5 ms) rendered irregular line shapes characteristic for the flagellated cells (Figure 4d). These results demonstrate the stochastic nature of flagellar interactions with the functionalized pore surface: peptide recognition is probabilistic, involving the chance of each flagellum to approach the peptide in a right conformation that would provide a strong electrostatic coupling within the translocation duration. Further theoretical studies are needed to solve this interesting problem in the recognition micropore sensing. According to the model by Evans,30 the time of the presence of a bacterium under temporal trap via ligand−peptide interactions in the channel, tr, can be described as

active sequence (FLLRVPHLGGGC) would be the direct electrostatic interaction between bacteria and the pore wall,24−26 since the isoelectric points (pI) were 6.02 and 8.55 for the GGGC linker alone and the FLLRVPHLGGGC peptide, respectively. While the Au surface is hydrophobic in nature whereby interacting with the bacteria in the nonfunctionalized channels,27 the hydrophobicity is anticipated to decrease when functionalized with the linker peptides owing to the OH groups on the side chains that provide negative charges on the pore wall in the buffer. Therefore, together with the chemically inert alkyl chain backbone, GGGC sequence brings fewer interactions between bacteria and hence little contribution to the bacterial translocation speeds.26 In contrast, the positively charged active probe (containing basic amino acids, R and H) locally attracts the negatively charged bacteria toward the pore edge via Coulombic forces and retards their electrophoretic translocation. The flagella-binding peptide binds also via van der Waals and hydrophobic interactions through the aromatic (F, H) and the hydrophobic (L, V) residues, respectively. The anticipated affinity of the peptides toward bacteria described here qualitatively explains the sequence-dependent td variations, which in turn indicates the predominant effects of the transient single-cell adsorption to the pore-wall surface on their translocation dynamics. This can be viewed as an electrical analogue of the peptide-based affinity sensors with a single-particle resolution that exploits the specific interactions between the recognition molecules and certain analytes to facilitate their discrimination by analyzing the resulting change in the ionic current profiles. We further verified the peptide-decorated micropores can be used for the recognition of bacterial flagella by comparing the resistive pulses obtained for the wild-type and Δf liC strains. Without peptide functionalization, we observed a large overlap in the Ip−td plots (Figure S8), reflecting the similar size, rodlike shape, and surface charge density (zeta potential, −43 mV and −42 mV for the wild-type and Δf liC cells, respectively) of the two bacteria. This is not surprising as flagella are a relatively thin and charge-neutral filamentous nanostructures28,29 that contribute little to the ion exclusion upon bacterial translocation. In contrast, upon the modification of the pore-wall surface with the active peptides, we observed a significant increase in td when analyzing both strains (Figure 4a,b). The slower

tr = t0 exp((E B − F Δx)/kBT )

where t0 is the translocation time of single-bacteria under zero barrier, Eeff = (EB − FΔx) is the effective barrier energy with EB the energy barrier, the electrophoretic force F, and the displacement of the bacterial surface from the equilibrium toward the direction along the peptides Δx, kB is the Boltzmann factor, and T is the temperature.31,32 This model assumes expedited thermal dissociation of bonding between the peptide and a bacterium via weakening of the intermolecular interaction at the pore wall surface associated with the elongation of the bond length during the electrophoretic pulling of the bacterial cell. t0 can be roughly estimated from the mobility μ of the bacterial cells acquired in the zeta potential measurements C

DOI: 10.1021/acs.analchem.7b04950 Anal. Chem. XXXX, XXX, XXX−XXX

Letter

Analytical Chemistry

Figure 5. Discrimination between the bacterial strains. (a) Feature parameters representing the line shapes of current signals used for Prec estimations. (b) The precision Prec of the discrimination between the wild-type and ΔfliC cells. (c) Multiphysics simulation of the ionic current blockage during E. coli translocation through the peptide-modified Au micropore. The ionic current Iion levels were normalized to the pulse height Ip. E. coli is modeled as a rounded cylinder. Insets illustrate the bacterial position during the current drop, and E. coli cell enters the pore when Iion/Ip is 0.72 (blue dashed line). (d) Translocation mechanisms of a single cell involving specific ligand−peptide binding at the channel wall surface. Orange and red plots correspond to Iion trace during the out-of-pore and in-pore translocation of individual cells.

using a zeta-sizer as t0 = μELbac, where E ∼ 19 kV/m and Lbac = 2.6 to 0.8 μm are the electric field at the pore and the effective length of the bacteria considering the orientation variance, respectively. When considering the average bacterial tilt angle of 45°, t0 = 3 ms. The fact that td in the nonfunctionalized pore is not so different from t0 suggests significantly strong electrophoretic forces applied on the cells compared to the electroosmotic dragging (Figure S11); or in other words, negligibly small EB compared to FΔx. On the other hand, by considering tr ∼ td, we can qualitatively characterize the peptide affinities by comparing Eeff ∼ ln(spike width). Using the ratio of ln(td), the relative strength of the nonspecific cell−wall interactions is estimated to be 0.98 EAu and 1.19 EAu for the spacer and active peptide-modified micropores, respectively, where EAu is EB deduced from td obtained for the Au pores. Here, td was shown to be slightly shorter for the wild-type cells compared with that for the Δf liC cells. This unexpected finding can be attributed to motility of the flagellated microbes, which offers an additional driving force, i.e., larger F for propagation through the channel. In contrast, minor differences were found in the td distributions derived from the nonspecific interactions in the active peptide-modified pore channels between two bacterial strains, which reflects the minor impact of F compared with that of the relatively high EB stemming from the strong affinity of the E. coli surface toward the peptides. The specificity of the recognition probe to flagellin was shown to be the highest EB of 1.38EAu. As described, EB represents a useful parameter for the evaluation of the affinity of the recognition molecules in the geometrically confined space of pore channels, wherein the probe characteristics differ greatly from what a dissociation constant in the bulk media describes.33 To determine the usefulness of our approach, and since the resistive pulse patterns contain a wealth of information such as size, shape, and surface charges, we applied a machine learning

approach to identify any bacterial traits based on the pulse waveforms, by defining feature parameters of not only Ip and td, similar to the conventional resistive analyses, but also many other factors including the pulse bluntness β, onset angle θ, pulse peak position r, pulse area ratio Sr, and inertia Jσ (Figure 5a; Figure S12). We evaluated each parameter by examining the difference in their distributions between two strains in a framework of an expectation-maximization theory.34 The precision Prec of the estimated values, calculated through TP/ (TP + FN) with TP and FN being the number of true-positive and false-negative outputs, respectively, was shown to be approximately 50% in the nonfunctionalized pores, irrespective of the spike characteristics used for discrimination. The discrimination between the strains was especially difficult when using the pulse height Ip, manifesting the difficulty to distinguish the bioparticles with rodlike morphologies. In contrast, Prec improved to above 60% with the active probe, which affected the translocation dynamics of the wildtype cells in particular. Surprisingly, although using td, the parameter directly reflects the peptide-modulated bacterial mobility, the classification was shown to be better, with Prec approaching 82%, when β, the width at 30% from the pulse apex, was used (Figure 5b). This result was explained by performing a theoretical analysis of the ionic current blockage, which demonstrated that it is efficient in providing the information on the physical mechanisms underlying the resistive pulse waveforms. As the pore channels used in this study have relatively low thickness-to-diameter ratio, the Iion spikes reflect the motions of particles outside the conduit,35,36 due to the considerable contribution of the access resistance.37 Quantitatively, a multiphysics simulation revealed that the bacteria are still at the orifice when the ionic current drops by 72% with respect to the pulse top (Figure 5c, d). This explained the usefulness of β in the description of the translocation D

DOI: 10.1021/acs.analchem.7b04950 Anal. Chem. XXXX, XXX, XXX−XXX

Letter

Analytical Chemistry

(7) Kawai, T.; Akira, S. Nat. Immunol. 2010, 11 (5), 373. (8) Yoon, S. -i.; Kurnasov, O.; Natarajan, V.; Hong, M.; Gudkov, A. V.; Osterman, A. L.; Wilson, I. A. Science (Washington, DC, U. S.) 2012, 335 (6070), 859. (9) Darvish, A.; Goyal, G.; Kim, M. In Proceedings of SPIE; Southern, Š. O., Ed.; The International Society for Optical Engineering: Bellingham, WA, 2015; p 94900M. (10) Apetrei, A.; Ciuca, A.; Lee, J.; Seo, C. H.; Park, Y.; Luchian, T. Nanoscale Res. Lett. 2016, 11 (1), 501. (11) Tsutsui, M.; Yoshida, T.; Yokota, K.; Yasaki, H.; Yasui, T.; Arima, A.; Tonomura, W.; Nagashima, K.; Yanagida, T.; Kaji, N.; Taniguchi, M.; Washio, T.; Baba, Y.; Kawai, T. Sci. Rep. 2017, 7 (1), 17371. (12) Yang, L.; Yamamoto, T. Front. Microbiol. 2016, 7, 1500. (13) Goyal, G.; Mulero, R.; Ali, J.; Darvish, A.; Kim, M. J. Electrophoresis 2015, 36 (9−10), 1164. (14) Zhang, H.; Tian, Y.; Jiang, L. Nano Today 2016, 11, 61−81. (15) Yusko, E. C.; Johnson, J. M.; Majd, S.; Prangkio, P.; Rollings, R. C.; Li, J.; Yang, J.; Mayer, M. Nat. Nanotechnol. 2011, 6 (4), 253. (16) Chaturvedi, P.; Rodriguez, S. D.; Vlassiouk, I.; Hansen, I. A.; Smirnov, S. N. ACS Sensors 2016, 1 (5), 488. (17) Iqbal, S. M.; Akin, D.; Bashir, R. Nat. Nanotechnol. 2007, 2 (4), 243. (18) Krishnakumar, P.; Gyarfas, B.; Song, W.; Sen, S.; Zhang, P.; Krstić, P.; Lindsay, S. ACS Nano 2013, 7 (11), 10319. (19) Wang, D.; Harrer, S.; Luan, B.; Stolovitzky, G.; Peng, H.; AfzaliArdakani, A. Sci. Rep. 2015, 4 (1), 3985. (20) Harrer, S.; Kim, S. C.; Schieber, C.; Kannam, S.; Gunn, N.; Moore, S.; Scott, D.; Bathgate, R.; Skafidas, S.; Wagner, J. M. Nanotechnology 2015, 26 (18), 182502. (21) Tanaka, M.; Hikiba, S.; Yamashita, K.; Muto, M.; Okochi, M. Acta Biomater. 2017, 49, 495. (22) Okochi, M.; Muto, M.; Yanai, K.; Tanaka, M.; Onodera, T.; Wang, J.; Ueda, H.; Toko, K. ACS Comb. Sci. 2017, 19 (10), 625. (23) Luo, L.; German, S. R.; Lan, W.-J.; Holden, D. A.; Mega, T. L.; White, H. S. Annu. Rev. Anal. Chem. 2014, 7 (1), 513. (24) Fahie, M. A.; Chen, M. J. Phys. Chem. B 2015, 119 (32), 10198. (25) Tsutsui, M.; He, Y.; Furuhashi, M.; Rahong, S.; Taniguchi, M.; Kawai, T. Sci. Rep. 2012, 2 (1), 394. (26) Wei, R.; Pedone, D.; Zürner, A.; Döblinger, M.; Rant, U. Small 2010, 6 (13), 1406. (27) Chapman, R. G.; Ostuni, E.; Liang, M. N.; Meluleni, G.; Kim, E.; Yan, L.; Pier, G.; Warren, H. S.; Whitesides, G. M. Langmuir 2001, 17 (4), 1225. (28) Brinton, C. C., Jr.; Buzzell, A.; Lauffer, M. A. Biochim. Biophys. Acta 1954, 15 (29), 533. (29) Tagliazucchi, M.; Szleifer, I. J. Am. Chem. Soc. 2015, 137 (39), 12539. (30) Bacri, L.; Oukhaled, A. G.; Schiedt, B.; Patriarche, G.; Bourhis, E.; Gierak, J.; Pelta, J.; Auvray, L. J. Phys. Chem. B 2011, 115 (12), 2890. (31) Evans, E. Annu. Rev. Biophys. Biomol. Struct. 2001, 30, 105. (32) Tabard-Cossa, V.; Wiggin, M.; Trivedi, D.; Jetha, N. N.; Dwyer, J. R.; Marziali, A. ACS Nano 2009, 3 (10), 3009. (33) Petrosko, S. H.; Johnson, R.; White, H.; Mirkin, C. A. J. Am. Chem. Soc. 2016, 138 (24), 7443. (34) Duda, R. O.; Hart, P. E.; Stork, D. G. Pattern Classification; John Wiley: New York, 2001; p 680. (35) Tsutsui, M.; He, Y.; Yokota, K.; Arima, A.; Hongo, S.; Taniguchi, M.; Washio, T.; Kawai, T. ACS Nano 2016, 10 (1), 803. (36) Weatherall, E.; Hauer, P.; Vogel, R.; Willmott, G. R. Anal. Chem. 2016, 88 (17), 8648. (37) Hall, J. E. J. Gen. Physiol. 1975, 66 (4), 531.

motions of E. coli inside the channel. The resistive pulses demonstrated widening primarily at the 30% apex (Figure S13). Furthermore, ligand−peptide interactions do not occur until the bacteria are residing inside the pore. Therefore, while the td parameter depicts the entire translocation dynamics of bacteria, including those at several-micrometers distance from the channel entrance, β represents a measure of the bioparticle mobility inside the conduit, emphasizing the effects of the adsorption to the wall surface. Therefore, β yields a higher Prec than td, since the random electrophoretic motions of bacteria in the capture stage are expected to contribute little to the bacterial discrimination.

■

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.analchem.7b04950. Detailed experimental methods, screening of E. coli binding peptides using peptide array comprising peptide library designed from amino acid sequences of TLR5 extracellular region, fluorescent observations of wild-type and Δf liC strain, structure of Au micropore chip, fabrication process of Au micropores, peptide adsorption on Au surface, resistive pulse extraction procedure, tilt angle dependence of resistive pulse height, ionic spike height versus width scatter plots, capture rate of bacteria, ionic current spikes, electrokinetics of single-bacteria, feature parameters used in a resistive pulse pattern analysis based on nonparametric density functions for single-bacteria discriminations, and ionic trace characteristics of peptide−bacteria interactions (PDF)

■

AUTHOR INFORMATION

Corresponding Authors

*E-mail: [email protected]. *E-mail: [email protected]. ORCID

Mina Okochi: 0000-0002-1727-2948 Author Contributions §

M. Tsutsui and M. Tanaka contributed equally.

Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This work was supported by the ImPACT Program of the Council for Science, Technology and Innovation, Cabinet Office, the Government of Japan.

■

REFERENCES

(1) Akira, S.; Takeda, K. Nat. Rev. Immunol. 2004, 4 (7), 499. (2) Janeway, C. A.; Medzhitov, R. Annu. Rev. Immunol. 2002, 20, 197. (3) Delves, P. J.; Martin, S. J.; Burton, D. R.; Roitt, I. M. Roitt’s Essential Immunology, 12th ed.; Wiley-Blackwell: Chichester, U.K., 2011. (4) Amini, K.; Kraatz, H.-B. Int. J. Environ. Anal. Chem. 2016, 7319, 1. (5) Hayashi, F.; Smith, K. D.; Ozinsky, A.; Hawn, T. R.; Yi, E. C.; Goodlett, D. R.; Eng, J. K.; Akira, S.; Underhill, D. M.; Aderem, A. Nature 2001, 410 (6832), 1099. (6) Smith, K. D.; Andersen-Nissen, E.; Hayashi, F.; Strobe, K.; Bergman, M. A.; Barrett, S. L. R.; Cookson, B. T.; Aderem, A. Nat. Immunol. 2003, 4 (12), 1247. E

DOI: 10.1021/acs.analchem.7b04950 Anal. Chem. XXXX, XXX, XXX−XXX