Ultrasensitive and unambiguous bacterial pathogen detection through super selective interactions between multivalent supramolecular immuno-nanoparticles (SINs)

Abstract

Unambiguous, rapid, cost-effective and simple methodologies for the detection of microbial pathogens in environmental and medical samples are vital for effectively controlling infectious diseases. Recently, the daunting task of early identification of pathogens has been addressed by using gold nanoparticles coated with antibodies that have shown to increase the sensitivity of detection several fold. However this usually affects the specificity of detection since background responses due to non-specific interactions also get amplified. In this study we have exploited multivalent supramolecular interactions between antibody-coated gold nanoparticles to achieve super selective detection of pathogenic bacteria. We term these as supramolecular immuno-nanoparticles (SINs) and used them to develop a simple, rapid and cost-effective immuno dot-blot detection method to selectively detect pathogenic bacteria even at extremely low densities (LOD: 50 cells per spot). Furthermore, we performed analytical dot-blot and ELISA assays using the antibodies, both in their free form and as SINs, to gain fundamental insights into the reasons behind the remarkable detection characteristics. We found that the combined effects of steric hindrance, signal amplification and surface density of the SINs resulted in the best detection characteristics in an immuno dot-blot format reported so far. These fundamental insights can be widely applied to improve various multivalent nanoparticle based detection methods.

---

1 Introduction

Unambiguous, rapid detection of microbial pathogens in clinical settings and their surveillance in environment, water and food are essential for the effective control of infectious diseases.¹ However, such methods are still elusive since the numerous molecular techniques that are available to specifically and sensitively identify pathogens are not readily field-deployable due to practical difficulties like cost and requirement of skilled workers.^2,3 Furthermore, most of these techniques involve indirect detection of pathogens by sensing biochemical analytes.^4–6 However, to prevent diseases, rapid detection of microbial contamination is required even before their toxins reach a measurable level, for which direct sensing of the microbial cells is required.⁷ Unlike simple biochemical analytes, complexly structured microbes pose unique problems, which are difficult to solve. Immunodiagnostic techniques have been relatively successful in addressing these issues and recently, nanoparticles have also been used to enhance the sensitivity of various cost-effective techniques such as dot blot,⁸ strip,⁹ and barcode¹⁰ tests. Furthermore, immuno-nanoparticles have abundantly been used as staining agents for detection purposes.^11,12 Recently we had reported the advantage of using secondary antibody-coated gold nanoparticles (GNPs) in the detection of White Spot Syndrome Virus (WSSV).⁸ There was an eighty fold increase in sensitivity compared to the use of normal secondary antibodies. Even though nanoparticle based techniques have greatly enhanced detection limits, non-specific interactions still plague such systems at high pathogen densities, leading to false positive readouts. For bacterial pathogens, O-antigen based detection is the one of the most common immuno-detection methods, however the specificity of such systems is relatively poor because of structural resemblances between O-antigens, resulting in cross reactivity not only among the serovars of the same species but also of other species. This is particularly true for clinical isolates, which have a constantly changing and heterogeneous mix of O-antigens. It is therefore important to devise methods that would improve this specificity problem along with enhancing the sensitivity.

Supramolecular chemistry, which deals with the assembly of molecules through non-covalent interactions, is one such field currently being employed to tackle such problems. Several novel and dynamic systems for biological purposes have been developed based on the principles of self-assembly, molecular recognition, multivalency, etc.¹³ In particular, nanoparticle based assemblies, formed by multivalent interactions, are being studied in great detail to understand and exploit the interesting phenomenon of super selectivity.¹⁴ Using theoretical¹⁴ and experimental means,^15,16 it has been shown that constructs containing multiple binding epitopes can selectively bind to surfaces that have ligand densities only above a certain threshold value. Such super selectivity allows us to design sensing systems that can distinguish target cells not only based on which ligands they display but also based on their spatial distribution.

In this study, we employ antibody-coated gold nanoparticles that interact with each other through such multivalent supramolecular interactions leading to super selective pathogenic detection. We term these as supramolecular immuno-nanoparticles (SINs) and have made 2 types – one with primary antibodies (primary SIN) and the other with corresponding secondary antibodies (secondary SIN). We have performed a comparative study with dot-blot assays using these SINs and free primary and secondary antibodies in four different types of combinations as shown in Fig. 1. In Schemes A–C, we observed different levels of sensitivity and selectivity however false-positive responses were always seen in cross-reactive and non-specific strains. Surprisingly, in Scheme D sensitivity improved by several orders of magnitude and non-specific signals were completely eliminated. All four schemes were further analyzed using various dot-blot and ELISA assays to gain insights into the reasons behind this highly selective response. The results show that we were able to develop a highly cost-effective (less than 0.30 US dollars for 10 samples) pathogen detection technique using dot-blot assays and antibody-coated nanoparticles, which was extremely sensitive and selective. Furthermore, the concepts learned from this study can lay the foundation for immense improvement in nano-diagnostic techniques.

Fig. 1 (a) Graphical representation of a pathogenic E. coli with magnified view of its outer membrane. The free antibodies and GNP-conjugated antibodies (SINs) used in this study have also been graphically depicted. (b) The four different combinations for immuno-detection of O-antigens have been shown graphically. The antibodies and SINs have been drawn to scale relative to the size of the O-antigens (O-antigen – 2.8 nm length, 1.2 nm width,¹⁷ IgG antibody – 14 nm length, 2 nm width,¹⁸ nanoparticle – ∼50–60 nm radius). This provides a better understanding of their assembled structures on the bacterial outer membrane.

2 Materials and methods

2.1 Chemicals and reagents

Chloroauric acid, trisodium citrate, nitro blue tetrazolium (NBT), 5-bromo-4-chloro-3-indolyl phosphate (BCIP), p-nitrophenyl phosphate (pNPP) and bovine serum albumin (BSA) were purchased from Sigma Aldrich. All solvents and reagents used in the experiments were analytical grade. Autoclaved and double distilled water were used throughout this work. Secondary antibody (goat anti rabbit IgG ALP conjugate) was purchased from Genei, USA. Primary rabbit antibody against O-antigens of different bacterial cells was purchased from Denka Seiken, Japan and Abd Serotech, India. The polyclonal antibodies against Uropathogenic E. coli (UPEC) were raised in rabbit against outer membrane polysaccharides (O-antigens) and isolated by ammonium sulfate precipitation method and dialyzed overnight.

2.2 Apparatus and instruments

UV-Vis absorption spectra of all the samples were recorded in Enspire, Perkin Elmer multi-plate reader in the range of 400–700 nm. The pH measurement was carried out in STL AP-1 plus, pH meter. Particle size distribution was determined by dynamic light scattering (DLS) apparatus, Malvern Instrument, UK. Bacterial cells OD were measured in Eppendorf Biophotometer. The NC membrane strips were imaged using Bio-Rad (Gel Documentation Unit) Molecular Imager® GelDoc™ XR. The SEM analysis was made in Tescan, Vega3 scanning electron microscope (SEM).

2.3 Bacterial cultures

Standard bacterial strains (ATCC & MTCC) and clinical isolates of different bacteria (Escherichia coli (E. coli), Shigella flexneri, Uropathogenic E. coli, Shigella dysenteriae, Pseudomonas aeruginosa, Klebsiella spp.) were characterized by standard microbiological and biochemical tests preceding the experiments.^19,20 The bacterial strains were grown in LB broth in an incubator shaker at 180–200 rpm at 37 °C overnight. 1 OD of bacterial cells were centrifuged and washed with PBS. These cells were re-suspended in PBS (10⁶ cells/1 μl/spot) for pathogen detection assay and the dilutions were verified by pour plate method on LB Agar.

2.4 Synthesis of GNP

GNPs were synthesized using standard citrate reduction method.^21,22 In brief, 100 ml of 1 mM aqueous HAuCl₄ solution heated to 100 °C under refluxing conditions were stirred vigorously and 10 ml of 30 mM sodium citrate was added rapidly. The yellow solution became transparent to dark blue and finally became wine red color which indicates end point of reaction. The nanoparticle solution was kept at 100 °C for 15 min with stirring and cooled at room temperature.^22–25 The prepared nanoparticles were characterized by using UV-Vis spectra analysis (Perkin Elmer, Enspire-Multimode Plate Reader), dynamic light scattering (DLS) Instrument and Scanning Electron Microscopy (SEM, Tescan-VEGA3).

2.5 Processing and formation of supramolecular construct

The GNPs were diluted in 1 : 1 ratio with autoclaved dd H₂O and the pH was adjusted to 7.5 by adding K₂CO₃. The flocculation assay was performed to determine the amount of antibody required to stabilize the GNP aggregation.²⁶ The antibody to be conjugated should be taken in higher amount than what is required to stabilize the GNPs calculated by the flocculation assay (ESI, M1†). The amount of antibody required to stabilize the GNPs was found to be 4 μg for 100 μl of diluted GNPs (0.75A₅₂₆ units per ml). Hence for bioconjugation, 5 μg of antibodies were added to 100 μl of GNP and stirred effectively for 30 min and kept at 4 °C overnight. Finally, 10 μl of BSA (10 μg μl⁻¹) were added and the mixture was mixed well and kept at 4 °C for 30 min and centrifuged at 10 000 rpm for 20 min followed by a wash with wash buffer. The final loose pellet was dispersed in 100 μl of preservative buffer for experimental use.

2.6 SIN based immunoassay for pathogen detection

Bacterial cells numbering 10⁶ CFU or less was bound to the membrane (nitrocellulose membrane, SCN, 0.2 μm, Nupore) for pathogen detection, because the bacterial endogenous ALP interfered with the results when more than 10⁷ CFU of cells was bound to the membrane (ESI Fig. S5†). Different bacterial strains were spotted (10⁶ cells per μl) on the NC membrane as an array. The spots were dried and the membrane was blocked with blocking buffer and washed with PBS twice. Subsequently, the membrane was incubated with 1 : 600 dilutions of primary SINs for 30 min at RT in a rocker. It was then washed with PBST once and followed by PBS twice at 3 min intervals and incubated with 1 : 600 dilution of secondary SINs (ALP conjugated secondary antibody) for 20 min at RT in a rocker. This was further washed with PBST once and then with PBS twice. The membrane was rinsed with substrate buffer and developed with ALP substrate (NBT, BCIP) chromogenic substrate. The color change can be observed visually by naked eye and the color intensity was measured by Molecular Imager® GelDoc™ XR.

2.7 ELISA based affinity comparison of different schemes

The binding characteristics of the different schemes were analyzed by ELISA experiments without pathogens. 96 well ELISA plates were loaded with 100 ng of the primary antibody with coating buffer and incubated for 2 hours at 37 °C. The plates were blocked with BSA for 1 hour at 37 °C and washed with PBS well. To coat the ELISA plate with the primary SINs 4 μl of the primary SINs from the stock (approximately 25 ng μl⁻¹ of antibody) were added to 96 μl of coating buffer. Then the plates were incubated with 100 μl of different dilutions of secondary antibody (1 : 250 to 1 : 64 000) and incubated for 1 hour at room temperature followed by wash with PBST and PBS. Then 100 μl substrate buffer was added and finally 1 mg ml⁻¹ pNPP substrate solution was added to each well and incubated in the dark for 20 min and the plates were observed at 405 nm. For all the schemes, the same procedure was followed and the results were observed and the calculations were made. The final average values were taken into consideration for analysis.

2.8 ELISA based pathogen detection in 96-well plate for different schemes

96 well ELISA plate wells were coated with different strains (ESI Table S2†) 10⁹ cells suspended in coating buffer (1 M NaHCO₃, 1 M Na₂CO₃) and left overnight. Wells were then blocked with 300 μl of 0.1% BSA and incubated for 30 min at 37 °C. The wells were washed once with PBS and 100 μl of 1 : 1000 dilutions (1 : 600 in case of SINs) of primary antibody (E. coli polyclonal) was added to the wells and kept for incubation at 37 °C for 1 hour. The wells were then washed with PBS, PBST and finally again with PBS. Similarly, 100 μl of 1 : 5000 dilution (1 : 600 in case of SINs) secondary antibodies was added to the wells and kept for incubation at 37 °C for 20 min. The wells were washed with PBS, PBST and again with PBS. Then 100 μl substrate buffers was added and incubated for 10 min and finally pNPP substrate solution (1 mg ml⁻¹) was added to each well and incubated in the dark for 20 min and absorbance values were recorded at 405 nm.

3 Results

3.1 Characterization of GNP and SINs

We determined the size of our GNPs using UV-Vis spectrophotometry, DLS and SEM. The absorption peak of the wine-red GNP solution was at 526 nm (Fig. 2c(i)) suggesting a size range of 30–50 nm.²⁷ The dynamic light scattering (DLS) analysis indicated a size range of 10–200 nm (Fig. 2a) with a predominant population around 50 nm and a polydispersity index of 0.24. This was further confirmed by SEM, where individual particles of approximately 50 nm in diameter were seen predominantly (Fig. 2b).

Fig. 2 (a) DLS data of GNPs showing the size distribution with the majority in the range of 50–200 nm, (b) SEM image of GNPs revealing predominance of approximately 50 nm particles, (c) UV-Vis spectrum of bare GNP with absorbance peak at 526 nm (i), slightly red-shifted spectrum of antibody-conjugated GNP (ii), and salt-induced aggregation of bare GNP (iii). Inset shows the change of colour of bare GNP (1) upon salt-induced aggregation (from wine-red to blackish blue) (3), but no color change in the case of antibody-coated GNPs even after the addition of salt (2).

The spectrophotometric analysis of GNP before and after bioconjugation with either primary antibody or secondary antibody revealed a slight red shift from 526 nm to 530 nm (Fig. 2c(ii)). This small red shift is in agreement with previous reports where proteins have been conjugated to GNPs.²⁸ Flocculation assays were performed to determine the amount of protein necessary to prevent GNP aggregation (Fig. 2c(iv)) in presence of high salt concentration (0.1% NaCl). The result showed that, 4 μg of antibody was required to stabilize 100 μl of GNP at a concentration (0.75A₅₂₆ units per ml). As can be seen in Fig. 2c(iii), such bioconjugated GNP retained the original wine-red color even after addition of salt.

3.2 Sensitivity of different schemes

Using these SINs, we first tested the detection limit of each scheme. This comparative analysis showed that Scheme A, in which free antibodies were employed, has the least sensitivity (10⁴ cells per spot). In Scheme B, wherein the primary SINs were used with free secondary antibody, the sensitivity was similar to that of Scheme A. However, when the secondary SINs were used along with free primary antibodies (Scheme C) the sensitivity improved by 2 orders of magnitude with faint spots visible even at 10² cells per spot. The most dramatic increase in sensitivity was observed when both primary SINs and secondary SINs were used together (Scheme D), which showed clearly visible spots even at 10² cells per spot and below that, the signal reduced drastically (Fig. 3a(i)D).

Fig. 3 (a) Comparison of sensitivities of different schemes, (i) dot-blot on nitrocellulose membranes having spots containing 10⁶ to 10² cells per spot showing intensity differences for each Scheme (A–D); for Scheme D, the sensitivity was tested on spots with as low as 10 CFU per spot. (ii) Column plot of quantified intensity values for spots versus the number of bacterial cells per spot. (b) Specificity comparisons of different schemes of the immunoassays, (i) two specific (P1, P2: UPEC475, UPEC149 respectively) and three different non-specific (N1, N2, N3: Proteus (5163), Klebsiella (340053), Staphylococcus (25923) respectively) strains were spotted on the membrane (10⁶ cells per spot) and detected using primary antibody against UPEC by the four schemes, (ii) column plot representing quantified intensity values of the spots for the different bacterial strains (the presented values are an average of triplicates ±SD, n = 3), (iii) comparison of sensitivity of the schemes with mixed cultures having different proportions of the positive and negative strains (total 10⁶ cells per spot). The numbers on top of the strips represent the proportion of the positive strain in each spot (iv) the graph generated from the intensity of the spots versus bacterial cells per spot for the four different schemes.

3.3 Reduction of non-specificity

In such immuno-detection assays, poor specificity of polyclonal antibodies, especially those directed against highly variable surface carbohydrates like O-antigens, is a common cause for the generation of false-positive readouts. Since such cross-reactivity is a major problem in immuno-detection techniques, we tested the specificity of detection of the four schemes using bacterial strains that were not specific for our polyclonal UPEC primary antibody. As expected, cross reactive binding of the primary antibody towards the non-specific bacterial strains resulting in faint spots appearing in Schemes A–C (Fig. 3b(i)-N1, N2, N3). No such spots were observed when only the secondary antibody was incubated with the bacterial cells even at 10⁶ cells per spot, confirming that it did not bind to the cell surfaces (ESI Fig. S1†). Surprisingly, in the case of Scheme D, these faint false-positive spots did not appear, indicating that the response due to cross-reactivity has somehow been eliminated. We also tested specificity of the different schemes under more stringent field-like conditions, by mixing negative strain (ESI, M2†) with positive strain in different proportions, keeping the total number of cells per spot constant (10⁶ cell per spot). Once again, we observed that ghost spots were produced in Schemes A–C (Fig. 3b(iii)), even in the absence of the negative strain. This was not the case with Scheme D – unambiguous dark spots were seen in the presence of only 10² cells per spot of the positive strain and no spot was seen in the absence of the positive cells. The spots were quantitatively measured and plotted as shown in Fig. 3b(ii & iv).

3.4 Dot blot and ELISA based pathogen detection assay for four distinct levels of cross-reactivity

To gain a better understanding of the mechanism behind Scheme D's impeccable detection specificity, we tested all four schemes using 12 different bacterial strains (ESI, Table S2†) that exhibited different levels of cross-reactivity. Using Scheme A, we observed that the E. coli O-antigen based polyclonal antibodies produced four distinct types of responses when tested with dot-blot (Fig. 4a) and ELISA assays (Fig. 4b). The strains, against which the antibodies were raised, produced clear positive responses and were termed as true positive (TP) strains; cross-reactive (CR) strains were pathogenic strains that produced responses similar to TP strains due to structural similarity of their O-antigens; non-specific (NS) strains were negative strains that produced ambiguous ghost spots in Schemes A–C due to cross-reactivity caused by weaker surface interactions or low-level resemblance of O-antigen structures; true negative (TN) strains produced no spots in all four schemes due to the absence of non-specific interactions. In all cases, Schemes A–C exhibited similar responses whereas Scheme D was able to eliminate the ghost spots due to low-level O-antigen resemblance, giving superior distinction between TP and NS strains. The TN strains do not interact with the primary antibody and hence there were no spots observed in all schemes in the dot-blot assay indicating that the antibodies do not non-specifically interact with bacterial surfaces as well as the dot-blot membrane. In TP and CR strains, Scheme D seemed to generate stronger responses compared to the other schemes, indicating that Scheme D effectively amplified the response when strong interactions occurred between the primary antibody and the bacterial surface. This led us to believe that Scheme D's improved sensitivity and selectivity might either be due to the affinity between the primary and secondary antibodies or the surface density of the primary antibody. We systematically tested these two effects.

Fig. 4 Four distinct types of responses against different bacterial strains observed using the polyclonal primary antibody (TP – true positive, CR – cross-reactive, NS – non-specific, TN – true negative). (a) Dot blot assay of 12 different strains showing the four distinct types of responses for all the Schemes A–D. (b) Absorbance values obtained from the ELISA experiment using all four schemes against the 12 different bacterial strains.

3.5 Affinity based binding characteristics

In order to explain the remarkable sensitivity and specificity seen in Scheme D, we further analyzed the data obtained from the average of dot-blot experiments in Fig. 3a. We plotted the intensity values against cells per spot for all four schemes and as expected they followed sigmoidal trends. We analyzed the data using the four parameter logistic model²⁹ given by the equation:

y = {(a − d)/[1 + (x/c)^b]} + d

where, a = lower asymptote, b = slope at inflection point, c = concentration at inflection point and d = upper asymptote. Scheme A showed the standard monovalent response normally obtained when using free antibodies. Here, the lower asymptote value (a = 0) indicated that no response was seen at low bacterial numbers per spot. The inflection point value (c = 1.4 × 10⁴), indicated that the half-maximum response was seen close to the spot with 10⁴ cells per spot and the slope at inflection point (b = 1.03) indicated that the response varies in direct proportion to the cell number. The upper asymptote value (d = 68) is the indication of the maximum possible response produced. In comparison, Scheme B has an inflection point value (c = 1.1 × 10⁵) that is almost one order higher and a slightly lower slope value (b = 0.92). This indicated that Scheme B, unexpectedly, performed worse than Scheme A for pathogen detection using this technique, in spite of presenting the primary antibodies in a multivalent manner. However, Scheme C showed better detection characteristics since the inflection point value (c = 6.5 × 10³) was about half an order lower than that of Scheme A and its upper asymptote value (d = 94) was higher than that of Scheme A. Even its lower asymptote value (a = 32) was non-zero, indicating an unusually high response at very low bacterial cells per spot. Finally, Scheme D showed highly sensitive detection characteristics. The inflection point value (c = 88) indicated that significant detection occurred even at low densities (100 cells per spot). The high slope value (b = 12.5) showed that the detection has an incredibly sharp off/on characteristic, where no response was seen below 50 cells per spot and maximum response (d = 78) occurred just above 100 cells per spot.

To further understand the scientific basis of Scheme D, the binding characteristics between the primary antibodies and secondary antibodies in all four schemes were studied in an ELISA experiment without cells, where the primary antibody (either free or as SINs) was coated on the wells and was titrated with different concentrations of the secondary antibody (either free or as SINs). As expected, we obtained sigmoidal response curves, which were fitted with the four parameter logistic equation (Fig. 5b). Here the inflection point concentration (c) was taken as the EC₅₀ value for the binding of secondary antibody to primary antibody. From Scheme A, we take the EC₅₀ value to be monovalent dissociation constant between the primary and secondary antibody (c = 5 × 10⁻⁹ M). Surprisingly, all four schemes displayed very similar EC₅₀ values. However, in Scheme B, a detectable response was observed only at higher secondary antibody concentrations (>2 × 10⁹ M) compared to the other schemes and a lower maximum response (d = 1.1) was seen compared to other schemes. This indicated that the number of secondary antibodies binding to the immobilized primary SIN was drastically reduced, possibly because of lack of access to the binding site as a consequence of compact organization of the primary antibodies around the GNP (Fig. 1b). Schemes C and D had detectable responses even at very low secondary antibody concentrations and higher maximum response values compared to Schemes A and B. This was probably so, since the response is caused by the secondary antibody and in the SIN form a large number of these enzymes are associated with every secondary antibody that binds to the primary antibody (Fig. 1b). This would also explain the higher sensitivity seen in Schemes C and D. However Scheme D did not exhibit a lower EC₅₀ value compared to Scheme C, indicating the secondary SIN adheres with similar binding strengths to well-packed surfaces of both free and GNP-conjugated primary antibodies.

Fig. 5 Four parameter logistic curve fit for different schemes (a = lower asymptote b = slope at inflexion point, c = concentration at inflection point, d = upper asymptote and adjusted R² values of the fits). (a) Curve fit for dot blot values (the presented values are an average of triplicates ±SD, n = 3), (b) curve fit for ELISA readings (the presented values are an average of duplicates ±SD, n = 2).

3.6 Surface density dependence of detection response

While using antibody-coated GNPs, apart from the naturally occurring antigen binding-site density and affinity, two other interplaying factors, steric hindrance and amplification, have to be taken into account to understand the differential behavior of the four schemes. Steric hindrance can be expected to occur in two forms; (1) binding of the primary SINs to the bacterial outer membrane causes a certain number of O-antigens to get buried because of the sheer size of the SINs and (2) the primary antibody forms a relatively well-packed oriented assembly on the GNP resulting in the reduction of the secondary antibody's ability to access its binding site that is presumably buried in the stem of the primary antibody. When one looks at the graphical representation of the different schemes with appropriate relative dimensions (Fig. 1, dimension based coverage calculations are provided in the ESI, Fig. S6†), in Schemes A and C (Fig. 1), the primary antibodies in their free form can easily access their O-antigens on the bacterial outer membrane, blocking only about ten antigens per antibody molecule. In Scheme A the secondary antibodies in their free form can access the primary in the ratio of 1 : 1 or 1 : 2 (secondary : primary), as in normal dot-blot or ELISA procedures. In Scheme C, even though the access of the primary is the same as in Scheme A, the secondary SINs will amplify the signal wherever the primary antibody binds irrespective of specific or non-specific binding of the primary antibody. In Schemes B and D, the binding of single primary SIN particle will block nearly a thousand O-antigens on the membrane thus restricting the surface area of interaction. Considering that the average 50 nm diameter nanoparticle can host a maximum of almost 350 antibodies and assuming that the free secondary antibody can bind all the available primary antibodies, amplification should be possible by a factor of 2 or 3. However the accessibility of secondary antibodies in free-form (Scheme B) or as SIN (Scheme D) to the binding site on the primary antibody will determine the actual signal. From Fig. 5b, it can be seen that Scheme B exhibits a poorer response compared to Scheme A, which indicate that the secondary antibody is severely hindered from binding to the GNP-conjugated primary antibodies. On the other hand, in Scheme D, signal amplification was observed at all secondary antibody concentrations in spite of the hindrance, very likely since the secondary antibody was in SIN form. In the negative control experiments (Fig. 3b(i)), Schemes A–C show low but significant non-specific responses probably because the primary antibody non-specifically binds to the bacterial surface. Normally, Scheme D would be expected to amplify these effects but there was no response seen in Scheme D in NS bacterial samples in Fig. 4.

The binding strength between multivalent entities is generally also dependent on the spatial distribution of binding epitopes. We postulated that in Scheme D, the binding of the secondary SIN to the primary SIN might be strongly dependent on the surface density of the primary antibodies.

To test whether this was the reason behind the behaviour of Scheme D, which completely eliminates the ghost spots, we performed an ELISA experiment by coating the well with a mix of two different primary SINs (ESI, M3 and S3†); one that specifically binds to the secondary antibody (specific primary SIN) and another one that does not bind (inert primary SIN). The ELISA plate wells were coated with; (1) 100% specific primary SIN, (2) 1% specific primary SIN, (3) 0.1% specific primary SIN. Since only Scheme B and D utilized primary SINs, these were the cases we tested by titrating various concentrations of the secondary antibody or SINs. The same titration was also performed on wells coated with only the inert primary SINs (ESI, Table S3†) and the values obtained were taken as the background response and subtracted from the other responses. The resulting data was fitted with the four parameter logistic fit as shown in Fig. 6. It was clearly evident that for Scheme B, even though the response decreases with lower densities of specific primary SINs, the binding affinity remained similar as indicated by the inflection point value (c ∼ 6.0 × 10⁻⁹) for all 3 cases. This is due to the fact that even though the amount of specific primary antibodies on the surface reduced, the free secondary antibody interacted only monovalently with it in a surface density independent manner.

Fig. 6 Four parameter logistic curve fits for ELISA experiments with different surface densities of the primary antibody, (a) Scheme B having surfaces of 100%, 1% and 0.1% specific primary SINs, (b) Scheme D having surfaces of 100%, 1% and 0.1% specific primary SINs. (c) Four parameter logistic fit values and adjusted R² values.

In Scheme D, we saw a different response. 100% and 1% relative densities exhibited similar affinity values (c ∼ 3 × 10⁻⁹) but in the case of the 0.1% a detectable response began only at the highest secondary SIN concentration employed. The slope in the fitting of the 1% surface response (b = 2.5) was also higher than that of 100% (b = 0.7) since the detectable response began at a correspondingly higher secondary SIN concentration and maximum response occurs at a correspondingly lower secondary SIN concentration. The maximum response of 1% surface (d = 1.48) was also lower than that of the 100% surface (d = 2.67). However the most striking result is seen on the 0.1% surface density of primary SINs where barely any response was observed for all the secondary antibody concentrations used. Above 5 nM of the secondary antibody, the secondary SINs seem to exhibit a complete shift from being unable to bind the primary SINs at 0.1% surface coverage to saturation level binding at 1% surface coverage. Such a distinct ‘on–off’ adsorption behavior seen within one order of magnitude of primary SIN surface densities indicates that the secondary SINs possibly bind in a super selective manner to these surfaces due to the multivalent supramolecular interactions between them. This seems to mimic the dot-blot results in Fig. 3, where Scheme B produced a ghost spot in NS strains whereas no such spot was seen in Scheme D. From this experiment, we propose that in case of TP strains, a relatively dense packing of the primary SINs occur on the bacterial surface because of which the secondary SINs can bind multivalently to them (Fig. 7). However in case of NS strains, the primary SINs bind at a lower density because of which the secondary SINs are not able to interact with them in a multivalent manner (Fig. 7). This results in the secondary SINs binding weakly and possibly monovalently to the primary SINs and causes their dissociation during the washing steps.

Fig. 7 Graphical depiction of the interactions between primary and secondary SINs on the outer membranes of TP and NS bacterial strains.

3.7 Reduced antibody coverage on SIN

To further validate our proposed mechanism, we tested whether the packing of the antibodies on the nanoparticles affected the sensitivity and specificity of the system. Normally we incubate antibodies with the GNP overnight to ensure relatively tight packing of the antibodies. This time we allowed the conjugation to occur for only 1 hour, hoping to achieve poor packing of the antibodies on the GNPs. This would provide the secondary antibody improved access to its binding site on the GNP-conjugated primary antibody. We then performed the sensitivity and specificity dot-blot assays similar to Fig. 3a(i) & b(i) and observed that the sensitivity of both Scheme B and Scheme D had drastically reduced. This is quite likely due to the presence of less number of primary antibodies for the secondary antibody to bind with. Below 10³ cells per spot no detection is seen even in Scheme D (Fig. 8a(i)). Also, as expected ghost spots were visible in two of the NS strains using Scheme D in the specificity test (Fig. 8b(i)). This indicated that the GNP-conjugated secondary antibodies were able to bind more strongly to the GNP-conjugated primary antibodies and prevented them from being completely removed during washing steps.

Fig. 8 (a) Sensitivity comparisons (10⁶ to10² cells) of Schemes B and D with reduced antibody coverage (a(i)). (b) Specificity comparisons of Schemes B and D with reduced antibody coverage (b(i)); two different Specific (P) and three different non specific (N) strains were spotted on the membrane (10⁶ cells per spot). P1 and P2 are TP strains (EP1 and T56). N1, N2 and N3 are NS strains (M8, T113 and M4 respectively). The graph generated from the color intensity of the dots versus the number of bacterial cells for sensitivity and specificity [a(ii) & b(ii)].

3.8 SIN based pathogen detection

Encouraged by the excellent specificity and sensitivity of Scheme D for O-antigen based pathogen detection, we applied this strategy using antibodies specific towards other bacterial strains (ESI, S2†). The negative strains selected for these experiments are pathogenic strains that are commonly present in field samples where such detection is required. The specificity of this immune detection method for identifying different bacterial pathogens is illustrated in Fig. 9 using Scheme D. The intensities of the spots were quantified and the corresponding values are shown as intensity peaks to judge the difference between the positive and negative spots. As can be seen, the method was able to clearly show positives vs. negatives without any ambiguity in all the three panels.

Fig. 9 SIN based pathogen detection (Scheme D) for UPEC identification (a), S. dysenteriae identification (b), S. flexneri identification (c) and their corresponding dot-blot strips (i), (ii) and (iii) respectively.

4 Discussion

Current worldwide biological concern about prevalence of pathogenic and antibiotic-resistant bacteria is resulting in technically sound methods to identify low abundance of these infectious pathogens using field deployable cost-effective techniques with sufficient sensitivity and specificity.^30–32 In this perspective, the classical dot blot method widely used for detection of virulence-associated proteins, virus, nucleic acids and pathogens has incorporated the advantages of nanotechnology like introduction of GNP and antibody–GNP conjugates to greatly improve its performance in terms of sensitivity. However, the direct capture of bacterial cells has been applied only in few reports and the assay sensitivity is rarely better than 10⁴ to 10⁵ CFU per sample.

Most of the methods still suffer from the serious problem of signal-to-noise ratio in many sensor types,^33,34 due to cross reactivity, especially if the antibody has been raised to a carbohydrate component like O-antigen. Researchers have achieved specific detection using techniques like PCR, DNA and RNA aptamer based detection,^35–37 but developing these techniques in a format that can be readily used in field conditions or without molecular engineering skills still remains challenging.

Hence, our interest was to solve these twin disadvantages using the emerging concepts of specificity enhancement by supramolecular organization of detection molecules. By systematically and comparatively studying the classical (Scheme A), improved (Scheme C) and the new (Scheme B and D) formats, we have found that Scheme D has remarkable sensitivity matching to that of sophisticated techniques like PCR and chemiluminescence. A striking feature is the clarity of results owing to elimination of non-specific signals. At the same time it does not eliminate significant cross reactivity of antibodies thus preserving their true behaviour (Fig. 3).

Recently supramolecular recognition leading to super selectivity has been theoretically dealt with by Martinez et al., 2011 based on multivalently interacting nanoparticles. The antibody molecules organize themselves because of their ability to bind GNPs forming constructs,³⁵ which can interact with each other in a multivalent supramolecular manner. This multivalent behaviour of supramolecular nanoparticles has been used for viologen dimer detection³⁸ and they achieved a two order of increase in sensitivity (11 μM to 0.1 μM). In our laboratory 80 fold increases could be achieved in the detection of WSSV⁸ using secondary SINs. Dually labelled GNP with antibody and ALP³⁹ resulted in 500 fold increase (5000 oocytes per ml to 10 oocytes per ml) in the sensitivity of detection of Cryptosporidium parvum. Though the enormous increases in sensitivity of detection in different systems are available, a comparative analysis of such supramolecular complexity employing the same system for bacterial detection has not been reported. The 2000 fold increase in sensitivity (50 cells per spot or 5 × 10⁴ cells per ml) of Scheme D employing both primary SINs and secondary SINs is a new high; the detection of as low as 50 bacterial cells is one of the best sensitivities achieved in pathogen detection using a simple procedure.⁴⁰ This is comparable and even better than what can be achieved by sophisticated techniques like SPR assay (LOD: 3 × 10³ to 10⁵ cells per ml), electrochemical impedance (LOD: 6 × 10⁴ to 6 × 10⁷ cells per ml), piezoelectric method (LOD: 1 × 10³ to 10⁸ cells per ml), chemiluminescence based (LOD: 10² to 10⁵ cells per ml) and optical absorbance assay (LOD: 5 × 10² to 5 × 10⁶ cells per ml).³⁶ Furthermore, this dot-blot based immuno-assay is relatively rapid and highly cost effective. Once the SINs and bacterial samples are available, the detection procedure requires only about 90 min with simple steps mostly involving dipping the strip in different solutions. A single analysis with up to 10 samples at once would cost less than 0.30 US dollars (calculations provided in the ESI†). Thus, this strategy is rapid, cost-effective and does not require skilled workers.

The schematic representation (Fig. 1) of different schemes with relative dimensions of the entities showed that Scheme D has the twin advantage of restricted surface interaction on the bacteria and amplification of signal by the secondary SIN. Though increase in sensitivity was anticipated, the vast improvement in specificity and the clarity of signal came as surprise. To account for it, we resorted to a more quantitative ELISA method and therefore affinity factors could be estimated. The four parameter logistic curve fits revealed the binding behaviour of the four schemes and superior performance of Scheme D, where the binding response starts at a much lower target concentration, as determined by both ELISA and dot-blot formats. In fact, in case of dot blot assay, it is highly sensitive with sharp on/off detection best suited for developing image based instrumentation for quick and high-throughput detection. Interestingly, the affinity between primary and secondary antibodies in free form or SIN form is fairly similar. However, when comparing Schemes A and B, the major difference is that, in Scheme B, the sensitivity is lower. This probably indicates that the secondary antibody is not able to access its binding site in the primary SIN as freely as it would access free primary. In case of Scheme C and D, however even with initial lower binding of secondary antibody to primary antibody, the amplification factor will be the major factor in enhancing the response. Further, specificity in Scheme D doesn't arise from affinity difference between primary and secondary antibodies in free form or SIN form, but probably arises from surface density packing of primary SINs on the bacterial surface (Fig. 7). This theory was proven to be likely by Fig. 6, where parameter c (affinity) does not change in Scheme B for different surface densities of primary SINs indicating similar binding affinities between primary and secondary antibody even at low surface densities. But in Scheme D at the lowest surface density of 0.1%, practically no signal is seen and other parameters change only a little between 100% and 1% surfaces indicating super selective binding between secondary SINs and primary SINs.

Hence the method could be used as a preventive or early detection tool for a variety of bacterial infections and even infections due to other types of pathogens and toxic molecules. Apart from exploiting in new ELISA and immuno-blot procedures, the strategy can be adapted for other solid and liquid based assays in which binding proteins other than immunoglobulins can also be employed.

5 Conclusion

In this work, we exploited the multivalent supramolecular interactions between antibody coated GNPs and studied their binding behaviour using both dot blot and ELISA formats to develop a highly sensitive and specific biosensor for bacterial pathogen detection. The comparative study using four different combinations of primary SINs and secondary SINs with respect to both sensitivity and specificity showed that the detection limit of as low as 50 cells per spot, currently the best possible in a simple dot-blot format, can be achieved when both the primary SINs and secondary SINs are used together (Scheme D). Moreover, the experimental evidence and analysis for drastic reduction in non-specific binding was studied and the effect of surface density packing was revealed out. These concepts can be applied in general for such other immune-detection techniques and also in the cases of detection using other multivalent constructs. It will also aid in developing superior microbial sensors for clinical and environmental applications.

Acknowledgements

We acknowledge the financial support from University Grants Commission (UGC) for Rajiv Gandhi National Fellowship to Hema; we acknowledge the financial support from DST-TDT funded National Hub for Healthcare Instrumentation Development for fellowship to Juhi and UGC-funded Centre with Potential for Excellence in Environmental Sciences and for supporting this research.

References

1. V. K. K. Upadhyayula, Anal. Chim. Acta, 2012, 715, 1–18.
2. N. Sanvicens, C. Pastells, N. Pascual and M. Pilar, TrAC, Trends Anal. Chem., 2009, 28(11), 1243–1252.
3. O. Lazcka, F. J. D. Campo and F. X. Munoz, Biosens. Bioelectron., 2007, 22, 1205–1217.
4. S. F. D'Souza, Biosens. Bioelectron., 2001, 16, 337–353.
5. T. Honda, T. Miwatani, Y. Yabushita, N. Koikem and K. Okada, Clin. Diagn. Lab. Immunol., 1995, 177–181.
6. K. Yagi, Appl. Microbiol. Biotechnol., 2007, 73, 1251–1258.
7. A. Ahmed, J. V. Rushworth, N. A. Hirst and P. A. Millner, Clin. Microbiol. Rev., 2014, 27(3), 631–646.
8. C. Thiruppathiraja, S. Kumar, V. Murugan, P. Adaikkappan, K. Sankaran and M. Alagar, Aquaculture, 2011, 318, 262–267.
9. P. Preechakasedkit, K. Pinwattana, W. Dungchai, W. Siangproh, W. Chaicumpa, P. Tongtawe and O. Chailapakul, Biosens. Bioelectron., 2012, 31, 562–566.
10. Y. Hui-qiong, J. Min-xian, S. Yang, W. Sheng-qi and Z. Jin-gang, Toxicon, 2012, 59, 12–16.
11. I. Nurulfiza, M. Hair-Bejo, A. R. Omar and I. J. Aini, J. Vet. Diagn. Invest., 2011, 23, 320–324.
12. M. Larguinho and P. V. J. Baptista, J. Proteomics, 2011, 75, 2811–2823.
13. J. Brinkmann, E. Cavatorta, S. Sankaran, B. Schmidt, J. V. Weerd and P. Jonkheijm, Chem. Soc. Rev., 2014, 43, 4449–4469.
14. F. J. Martinez-veracoechea and D. Frenkel, PNAS, 2011, 108, 10963–10968.
15. G. V. Dubacheva, T. Curk, R. Auzély-Velty, D. Frenkel and R. P. Richter, PNAS, 2015, 112(18), 5579–5584.
16. L. Albertazzi, F. J. Martinez-Veracoechea, C. M. A. Leenders, I. K. Voets, D. Frenkel and E. W. Meijer, PNAS, 2013, 110(30), 12203–12208.
17. S. Qiao, Q. Luo, Y. Zhao, X. C. Zhang and Y. Huang, Nature, 2014, 511, 108–111.
18. Y. H. Tan, M. Liu, B. Nolting, J. G. Go, J. Gervay-Hague and G. Liu, ACS Nano, 2008, 2, 2374–2384.
19. R. Aarthi, R. Saranya and K. Sankaran, Appl. Microbiol. Biotechnol., 2013, 98, 445–454.
20. T. Elavarasan, S. K. Chhina, M. Parameswaran (Ash) and K. Sankaran, Sens. Actuators, B, 2013, 176, 174–180.
21. J. Kimling, M. Maier, B. Okenve, V. Kotaidis, H. Ballot and A. J. Plech, J. Phys. Chem. B, 2006, 110, 15700–15707.
22. N. N. Long, L. V. Vu, C. D. Kiem, S. C. Doanh, C. Thi, N. Pham, T. Hang, N. D. Thien and L. M. J. Quynh, J. Phys.: Conf. Ser., 2009, 187, 012–026.
23. X. Ji, X. Song, J. Li, Y. Bai, W. Yang and X. J. Peng, J. Am. Chem. Soc., 2007, 129, 13939–13948.
24. K. C. Grabar, R. G. Freeman, M. B. Hommer and M. J. Natan, Anal. Chem., 1995, 67, 735–743.
25. A. Tabrizi, F. Ayhan, H. Ayhan and J. Hacettepe, Biol. Chem., 2009, 37(3), 217–226.
26. S. Thobhani, S. Attree, R. Boyd, N. Kumarswami, J. Noble, M. Szymanski and R. A. J. Porter, Immunol. Methods, 2010, 356, 60–69.
27. J. C. Martínez, N. A. Chequer, J. L. González and T. J. Cordova, Nanosci. Nanotechnol., 2012, 2, 184–189.
28. M. Iosin, F. Toderas, P. L. Baldeck and S. J. Astilean, J. Mol. Struct., 2009, 924–926, 196–200.
29. R. A. Herman, P. N. Scherer and G. J. Shan, Immunol. Methods, 2008, 339, 245–258.
30. D. Ivnitski, I. Abdel-Hamid, P. Atanasov and E. Wilkins, Biosens. Bioelectron., 1999, 14, 599–624.
31. K. Yagi, Appl. Microbiol. Biotechnol., 2007, 73, 1251–1258.
32. R. Edgar, M. McKinstry, J. Hwang, A. B. Oppenheim, R. A. Fekete, G. Giulian, C. Merril, K. Nagashima and S. Adhya, PNAS, 2006, 103, 4841–4845.
33. H. Wang, M. Sodagari, Y. Chen, X. He, B. Z. Newby and L. Ju, Colloids Surf., B, 2011, 87, 415–422.
34. C. Guegana, J. Garderesa, G. L. Penneca, F. Gaillardb, F. Faya, I. Linossiera, M. Herryc, M. N. Bellon Fontained and K. Vallee Rehela, Colloids Surf., B, 2014, 114, 193–200.
35. Y. J. Lee, S. R. Han, J. S. Maeng, Y. J. Cho and S. W. Lee, Biochem. Biophys. Res. Commun., 2012, 417, 414–420.
36. J. Sakata, K. Kawatsu, T. Iwasaki, K. Tanaka, S. Takenaka, Y. Kumeda and H. J. Kodama, J. Microbiol. Methods, 2012, 88, 77–82.
37. C. Chang, P. H. Hung, C. C. Wu, T. C. Cheng, J. Tsai, K. J. Lin and Y. Lin, Sensors, 2012, 12, 2710–2728.
38. H. Yamaguchi and A. Harada, Biomacromolecules, 2002, 3, 1163–1169.
39. C. Thiruppathiraja, S. Kamatchiammal, P. Adaikkappan and M. Alagar, Biosens. Bioelectron., 2011, 26, 4624–4627.
40. R. Joshi, H. Janagama, H. P. Dwivedi, T. M. A. Senthil Kumar, J. Lee-Ann, J. Schefers and S. Sreevatsan, Mol. Cell. Probes, 2009, 23, 20–28.

---

Footnote

† Electronic supplementary information (ESI) available: Supporting figures, tables, material and methods and the explanations. See DOI: 10.1039/c6ra03117j

---