Article pubs.acs.org/ac
Surface Modified Microprojection Arrays for the Selective Extraction of the Dengue Virus NS1 Protein As a Marker for Disease David A. Muller,†,‡ Simon R. Corrie,† Jacob Coffey,† Paul R. Young,‡,§ and Mark A. Kendall*,† †
Australian Institute for Bioengineering and Nanotechnology, University of Queensland, Australia Australian Infectious Disease Research Centre, School of Chemistry and Molecular Biosciences, University of Queensland, Australia § Institute for Molecular Bioscience, University of Queensland, Australia ‡
S Supporting Information *
ABSTRACT: While advances in assay chemistry and detection continue to improve molecular diagnostics technology, blood samples are still collected using the 150-year-old needle/syringe method. Surface modified microprojection arrays have been developed as a novel platform for in vivo, needle-free biomarker capture. These devices are gold coated silicon arrays with >20,000 projections per cm2, which can be applied to the skin for tunable penetration into the epidermis or dermis. The microprojection array conceptually offers several advantages over the current methods including: minimally invasive sample collection, no need for sample processing and concentration of specific markers at the device surface for sensitive detection. In this study, Microprojection arrays were coated with antibodies to capture an early marker of dengue virus infection, NS1, from the skin of live mice. We also developed a complementary “total IgG” assay which could be used as a positive control for adequate penetration of the projections. Surface modifications designed for selective extraction were tested against standard microtiter plate ELISA. We also investigated the use of Protein G-mediated antibody immobilization in order to orient capture antibodies. While we found that capture efficiency could be improved, the direct EDC-based antibody immobilization resulted in a significantly higher surface density leading to a higher degree of NS1 capture. Using mice intravenously injected with recombinant dengue virus type 2 NS1 as a pseudomodel for dengue infection, NS1 was successfully extracted using microprojection arrays sampling from skin fluid, with a detection limit of 8 μg/mL.
C
devices.3 Indeed, the MPAs used in the current study are 10− 100-fold smaller than related devices and hence are expected to be even less invasive. The extracellular fluid present in the skin epithelia is known to contain a mixture of plasma proteins and local tissue proteins.8 Focusing on the plasma component, this is a particularly complex mix of proteins, with individual concentrations spanning 10 orders of magnitude. Furthermore, over 90% of the total plasma protein content is made up of only a handful of “high abundance” proteins which mask the presence of the lower concentration target proteins.1 Confronted with this environment, suitable surface modification of MPAs for selective capture must focus on (a) reducing nonspecific adsorption of high abundance proteins, while (b) covalently attaching suitable “probes” which capture the desired target with high affinity. While protocols for direct adsorption of probes to metallic surfaces exist, it has previously been shown that this passive binding leads to low analytical sensitivity of biomarker detection.9 The reduction in sensitivity can be overcome by the addition of covalently immobilized oligomeric or polymeric spacers. The addition of heterobifunctional Poly ethylene glycol (PEG) to gold surfaces has been a common approach to solving this problem10 and serves two purposes.
urrently most diagnostic assays rely on blood or blood products for biomarker detection.1 This usually requires a venous blood draw or fingerprick sample, requiring trained staff and access to laboratory facilities. This is not ideal for resource limited settings2 and is uncomfortable at best for patients,3 especially those requiring frequent testing. However, while diagnostic technologies are becoming simpler, rapid, and more sensitive, the effects of sampling on downstream sensitivity and utility have been largely neglected. Here we show the potential for the use of microprojection arrays (MPA) as a new method of directly extracting circulating protein biomarkers from the skin. MPAs have been traditionally applied to skin for the purpose of delivering high molecular weight compounds which cannot be delivered efficiently by other means. When applied with a customized spring-loaded (low impact velocity; 2−3 m/s) applicator onto the skin surface, the projections breach the outer layers of skin, lodging in the epidermis and/or dermis.4 Our group, and others, have targeted the immunologically rich skin epithelia to deliver a range of vaccines (typically mDa range) showing strong immune responses and protection against challenge.5 While transdermal glucose sensing has been investigated briefly in the past,6 to our knowledge we published the first study demonstrating that surface-modified MPAs could be applied to the skin for capture and detection of antigenspecific IgG antibodies.7 This technology is considered to be minimally invasive in comparison to standard needle/syringe © 2012 American Chemical Society
Received: December 22, 2011 Accepted: February 26, 2012 Published: March 12, 2012 3262
dx.doi.org/10.1021/ac2034387 | Anal. Chem. 2012, 84, 3262−3268
Analytical Chemistry
Article
NS1 protein expression details) was incubated at 37 °C for 1 h and then washed 6 times in phosphate buffered saline with 0.05% (v/v) tween-20 (PBST). Fifty μL of the detection antibody anti-NS1-HRP (horseradish peroxidase; Panbio) was incubated for 1 h at 37 °C, and then plates were washed a further 6 times in PBST before being developed. Once the detection antibody was removed, plates were washed a further 6 times and developed using 50 μL of 3,3′,5,5′-tetramethylbenzidine (TMB) (ELISA systems) for 10 min in the dark before the reaction was stopped with 50 μL of 1 M phosphoric acid and absorbance was read at 450 nm and quantified using a standard curve of known NS1 concentrations.17 For the total IgG ELISA, Nunc maxisorp plates were coated overnight at 4 °C with 50 μL of a 5 μg/mL solution of goat antimouse Fc specific IgG (Sigma Aldrich) in carbonate buffer (0.05 M carbonate-bicarbonate buffer pH 9.6). Plates were then blocked as above. Serum samples (50 μL) were added, incubated for 1 h at 37 °C, and then washed 6 times in PBST. Captured IgG was then detected with 50 μL of goat antimouse Fab specific IgG HRP conjugate (Sigma Aldrich) (1:4000), incubated for 1 h at 37 °C, and then washed 6 times. Plates were developed as described above. Antibody Coupling to MPA. Five μg/mL of capture antibody (or 80 μg/mL Protein G) was covalently attached to MPAs using EDC/NHS chemistry. Polyethylene glycol (PEG)coated gold MPAs were washed with 100 mM N-morpholino ethanesulfonic acid hydrate (MES) pH 5 (MPI biomedicals) in preparation for EDC/NHS activation. MPAs were incubated for 1.5 h at room temperature on a shaker (300 rpm) in 200 μL of EDC/NHS activation buffer (100 mM MES, 5 mM EDC, 5 mM NHS), washed twice by emersion in 100 mM MES, and incubated in a solution of (a) capture antibodies at 5 μg/mL in 200 μL or (b) Protein G at 80 μg/mL in 200 μL, for 2 h at room temperature in a shaker (300 rpm). The reaction was stopped by flooding with 200 μL of 100 mM glycine for 1 h, and the MPAs were then stored at 4 °C in phosphate buffered saline (PBS). Confirmation of protein immobilization was performed by X-ray Photoelectron Spectroscopy (XPS) as previously reported.7 We also measured the surface density of IgG-HRP against a dilutions series of the same antibody, measuring absorbance at 450 nm following enzymatic conversion of TMB substrate (see Figure S5 in the Supporting Information). To immobilize capture antibody on the protein G surface, 200 μL of 20 μg/mL of antibody was incubated on MPAs for 2 h at 37 °C in PBS, followed by washing in PBST (6 times) to remove excess antibody. To determine the amount of antibody bound to protein G, 200 μL of 1:1000 dilution of goat IgG biotin conjugate was incubated with the MPAs for 1 h at 37 °C and washed 6 times in PBST. MPAs were then probed with 200 μL of 1:1000 dilution Cy5-streptavidin conjugate for 1 h at 37 °C and then washed a further 6 times. MPAs were imaged in a Tecan laser scanner. Relative fluorescence intensity was determined using Image J18 and analyzed (see the Supporting Information for details of data analysis) in PRISM 5.02 (GraphPad Software, Ja Jolla, USA). In vitro MPA-ELISA. Following EDC/NHS coupling of capture probes, MPAs were transferred into preblocked 96 well plates. NS1 and IgG capture ELISA’s were performed as described above; prior to development with TMB MPAs were transferred to a new well to avoid any nonspecific binding of the detection probe to the reaction well. In vivo MPA-ELISA. In vivo antigen (NS1) and antibody (total IgG) detection was performed on BALB/c mice. All
First, it allows for capture antibodies to be covalently immobilized to amine or carboxylic acid functional groups via N-(3-dimethylaminopropyl)-N′-ethylcarbodiimide hydrochloride (EDC)/N-hydroxysuccinimide (NHS) chemistry. Second, PEG creates a nonfouling layer that significantly reduces nonspecific binding of proteins to the surface. However, using this approach alone does not solve the problem of random probe orientation which leads to low capture efficiency. In an attempt to address the issue of suboptimal orientation of capture antibodies, protein G (a bacterial protein from streptococcal bacteria which binds the Fc region of antibodies with high affinity) has been utilized by several groups with varying success.9b,11 In a comprehensive recent study, Tajima et al. (2011) found that the positive effect of antibody orientation on assay sensitivity decreased with increasing molecular weight of the protein target, to the point at which there was no benefit for proteins >100 kDa.11 In this study, we targeted dengue fever detection as a test case, as all of the available commercial assays to our knowledge (lateral flow assays, ELISAs, etc.) accept only serum/plasma samples, thus requiring significant laboratory-based processing. Dengue virus is an important human pathogen causing up to 100−200 million infections each year across the tropical belt, and with half of the world’s population at risk of infection. Dengue virus is spread through the bite of infected Aedes aegypti (A. aegypti) mosquitoes. Disease resulting from this infection varies greatly from asymptomatic infection and selflimiting illness, dengue fever, through to the life threatening forms of disease associated with secondary infection, dengue hemorrhagic fever (DHF), and dengue shock syndrome (DSS).12 The viral nonstructural protein 1 (NS1; a 46−55 kDa glycoprotein, naturally present in patient sera)13 is secreted from infected cells from the onset of disease symptoms with as much as 50 μg/mL being reported out to 12 days post disease onset.14 Elevated levels of NS1 early in secondary infection have been observed to correlate with the progression to severe disease later in infection suggesting NS1 may act as a prognostic marker in some settings.14b,15 Diagnosis of dengue virus infection is routinely performed by serological (IgM and IgG capture) and antigen tests (NS1), currently performed in enzyme linked immunosorbent assay (ELISA) or rapid point of care formats. However, following needle/syringe blood draw, serum/plasma is separated from the cellular components of whole blood prior to analysis. Here we report NS1 protein capture from the skin of live mice, followed by detection on an MPA device without the need for sample processing. First, we show the development of MPA-based capture of NS1 and total IgG in comparison to a gold-standard ELISA. BALB/c mice were then injected intravenously with NS1 as a pseudomodel of disease, and we successfully captured NS1 from these mice using noninvasive sampling/extraction on the MPA.
■
MATERIAL AND METHODS Nunc Maxisorp Plate ELISA. The NS1 antigen capture ELISA was performed using protocols previously established by Young et al. (2000).16 Briefly, plates were coated with 50 μL of carbonate buffer (0.05 M carbonate-bicarbonate buffer pH 9.6) containing 5 μg/mL anti-NS1 monoclonal capture antibody overnight at 4 °C. Following antibody attachment, plates were blocked with 200 μL 1 × milk serum diluent (KPL, Inc., Gaithersburg, USA) 1% sucrose for 2 h at room temperature. Fifty μL of the diluted samples (see Supporting Information for 3263
dx.doi.org/10.1021/ac2034387 | Anal. Chem. 2012, 84, 3262−3268
Analytical Chemistry
Article
Figure 1. Assay investigation and development (A) IgG Capture ELISA with mouse IgG being captured with a goat antimouse IgG Fc specific antibody and detected by a goat antimouse IgG Fab specific HRP conjugate antibody. IgG capture ELISA titration curves for IgG captured on either the (B) Nunc maxisorp and (C) MPA surfaces. (D) NS1 Capture ELISA with NS1 being captured with an anti-NS1 monoclonal antibody and detected by an anti-NS1 HRP conjugate antibody. NS1 capture ELISA titration curves for NS1 captured on either the (E) Nunc maxisorp (0.5 ng/ mL sensitivity) and (F) MPA surfaces at 100, 50, 25, 12.5, 6.25, and 0 ng/mL Signal from the Maxisorp plates was normalized to the surface area of the MPA (6 ng/mL sensitivity in PBS).
Figure 2. Effect of Protein G-mediated antibody coupling in comparison to EDC coupling. (A) Relative antibody density using EDC or Protein G antibody coupling methods in comparison to negative control PEG coated surface. (B) Effect of antibody orientation on NS1 capture/detection in 10% serum. Anti-NS1 capture mAbs was either orientated with protein G or randomly coupled using EDC attachment chemistry.
animal experiments were performed in accordance with ethical guidelines set by The University of Queensland and the National Health and Medical Research Council of Australia. Animals were anaesthetized with 60 μL of intraperitoneal injection of xylazine, ketamine, and saline solution in the volume ratio 1:1:2. Excess hair on the flank was removed using Nair and an electric shaver. Flank skin was secured to vinyl backing using double-sided tape under a spring-loaded applicator. MPAs were attached to the applicator device with a carbon tab and applied to the mouse skin upon spring release, at a velocity of 2.1 m/s for the 110 μm microprojection array (110-MPAs) and 3.1 m/s for the 260 μm microprojection array (260-MPAs). For details of MPA penetration depth analysis please see the Supporting Information. Purified NS1 was injected into anaesthetized BALB/c mice by the tail vein as a pseudomodel for disease. Mice were prepared for MPA application as described above. The required amount of NS1 (2 mg−3 μg) was mixed with Evans Blue dye (0.001% w/v) and injected into the tail vein and allowed to circulate for 5 min. During the 5 min period, mice were prepared for MPA application by drawing flank skin away from the body taping it
to the stand. MPAs were left on the animal for 20 min to allow for in vivo sampling of NS1 or IgG, removed and placed into 96 well plates, and washed 6 times with PBST. The MPA ELISA was performed as described above. Following removal of the MPAs mice were bled and euthanized.
■
RESULTS AND DISCUSSION In developing the appropriate surface chemistry for high capture efficiencies of antigens from skin, we compared the MPA devices directly against Nunc maxisorp plate surfaces. First, a key requirement for a diagnostic assay is a “positive control” which confirms successful operation of the device. For the MPA technology, a control which confirms successful penetration into the relevant tissue compartment (i.e., the dermis) is needed. We chose total IgG as a useful positive control (Figure 1A), given the range of reagents available and the various reaction schemes possible. After investigating three capture ELISA strategies for IgG capture (Figure S1) we compared our preferred assay (Fc-capture, Fab detection) on Nunc microtiter plate surfaces against the MPA surfaces (Figure 1B and C, respectively). After taking into account the 3264
dx.doi.org/10.1021/ac2034387 | Anal. Chem. 2012, 84, 3262−3268
Analytical Chemistry
Article
surfaces we observed lower initial slopes for all three curves (Figure 1F). As the saturation signals are similar for PBST on both surfaces, binding of the detection antibody to NS1 does not appear to be affected. Therefore, it appears that the capture process is responsible for the difference in selectivity between the two surfaces − likely due to (a) serum/plasma components limiting the access of NS1 to the capture antibody, (b) competition for NS1 binding between capture antibodies and other potential NS1-binding partners (a number of which have been recently identified),20 or (c) a combination of both. Nevertheless, as NS1 concentration was increased, there was an increase in signal. During the development of both NS1 and Total IgG MPA-ELISAs (Figure 1) a blocking step was included in all plate/MPA assays for consistency. However, it was later found that the addition of the blocking agent had a negligible effect on the signal when compared to the PEG surface (Figure S2). As a result the blocking step was omitted from all subsequent in vitro and in vivo MPA-ELISAs. This was considered an advantage as it was undesirable to have adsorbed milk proteins on the MPA surface for in vivo biomarker extraction. We reasoned that Protein G-mediated antibody immobilization might result in improved capture efficiency of NS1 in comparison to the direct EDC-immobilization method. On Nunc substrates, capture antibodies passively adsorb to the surface mainly via the Fc region, thus allowing Fab fragments to be accessible for antigen binding in solution.21 In contrast, using EDC-mediated cross-linking to a COOH-terminal surface (as in the case of MPA-ELISA), orientation is expected to be more random due to the covalent attachments forming between activated NHS esters and lysine side-chains or Ntermini. However, while there has been much published in recent literature on the positive effects of protein G mediated antibody immobilization,9b,22 to our knowledge there have been no reports comparing the effect of antibody orientation on the capture efficiency of functionalized surfaces exposed to natural biological fluids. Against this backdrop, we coupled biotinylated antibodies to the MPA surfaces via direct EDC-mediated antibody coupling or Protein G-mediated antibody coupling (protein G coupled to surface by direct EDC) and detected their relative abundance by Cy5-streptavidin binding and fluorescence analysis (Figure 2A). The surface prepared using Protein G-mediated antibody immobilization was very stable when immersed in solutions containing high antibody concentrations, with negligible reduction in Cy5 signal following prolonged exposure (Figure S3). We observed, using the same concentration of biotinylated antibody, that the EDC-mediated immobilization yielded significantly higher Cy5 signals in comparison to that for the protein G surface, suggestive of a higher surface density. It is possible that using higher concentrations of protein G (80 μg/mL in this study) may yield higher antibody surface density; however, this is not an economical option. Repeating the same antibody immobilization processes to attach the anti-NS1 capture IgG to both surfaces, we again performed an NS1 titration experiment using a 10% serum matrix. Surprisingly, these curves were almost identical, suggestive of higher capture efficiency for the protein G immobilized surface. This suggests that the capture efficiency does improve with Protein G-mediated immobilization, presumably by improving the orientation of the anti-NS1 capture antibody. At first, our results appear contradictory to a recent study showing that larger proteins did not benefit from antibody orientation.11 However while the hexameric (soluble)
Figure 3. Microscopy of MPAs and their interaction with mouse flank skin. (A) 110-MPAs showing 110 μm projections of the “cone-oncylinder” design; (B) 260-MPAs showing 260 μm projections of the “bullet” shape design; (C) cryo-SEM image of the 110-MPAs and (D) 260-MPAs penetrating into skin (note: skin material is obscuring the view of penetration depth in this image); (E) fluorescence microscopy images of cryo-preserved skin sections of 110-MPAs and (F) 260MPAs showing FluoSpheres delivered to the skin by MPAs, enabling the resolution of individual projections (see the Supporting Information for details of MPA depth penetration measurement method).
difference in surface area, the titration curves appeared almost identical. This is not surprising, given the similar surface densities on the MPA (∼250−300 ng/cm2, based on an HRPlabeled IgG which might underestimate surface density; Figure S5) and the polystyrene plate (∼260−550 ng/cm2).19 Furthermore, Tajima and colleagues11 clearly showed that IgG, as a target biomarker, is large enough that small changes in capture antibody density or orientation are unlikely to affect capture. For detection of the dengue virus antigen, NS1 we performed a similar experiment to compare target capture in a capture ELISA format on the Nunc plate in comparison to the MPA (Figure 1D). After normalizing HRP signals for the lower surface area of the MPA we found striking differences between the two surfaces. On the Nunc plate (Figure 1E), dilution of NS1 into either a simple (e.g., Bovine serum albumin only) or complex protein source (e.g., plasma or serum) had no significant effect on the shape of the curve. However, on MPA 3265
dx.doi.org/10.1021/ac2034387 | Anal. Chem. 2012, 84, 3262−3268
Analytical Chemistry
Article
Figure 4. In vivo biomarker sampling (A) NS1 extracted from skin fluid by 260-MPAs and 110-MPAs in mouse groups 1 and 2. (B) Levels of NS1 in the serum of mouse groups 1 and 2 at 20 min post injection. (C) Penetration control, IgG extracted from skin fluid by 260-MPAs and 110-MPAs. (D) Location of flank MPA application. (E) Lowest detectable level of NS1 in skin via 110-MPAs and (F) the corresponding serum NS1 levels, including the difference in signals for 300 μg at 5 or 20 min.
surface chemistry and probe immobilization may improve detection limits on MPA devices. With our surface modifications established in vitro, we then moved to investigating the MPA designs applicable to extraction of circulating antigens from the flank skin of live mice. We selected both our standard mechanical design (“110MPAs”) from a previous study,7 along with a device of significantly larger total area and projection length (“260MPAs”). SEM analysis of the MPAs used in this study revealed regular arrays of projections at >20,000 cm−2 density (Figure
form of NS1 is 300 kDa, it is well-known to be detergent labile (e.g., PBST solutions) with exposure to nonionic detergent resulting in the breakdown of hexameric to 100 kDa dimers.13b,20,23 These dimers are in the correct molecular weight range to be affected by antibody orientation. However, it appeared that this benefit was offset by the larger degree of capture antibody bound directly via EDC (Figure 2B). We therefore continued to use direct EDC grafting of the NS1 capture antibody for the remainder of this study, but it is clear from the results in Figures 1 and 2 that further investigation of 3266
dx.doi.org/10.1021/ac2034387 | Anal. Chem. 2012, 84, 3262−3268
Analytical Chemistry
Article
3A and B). The “110-MPAs” were 110 μm in height, and, based on penetration depth analysis, the penetration depth for the 110-MPAs into mouse flank skin was ∼79 ± 14 μm, indicating complete dermal penetration (see the Supporting Information for details of MPA depth penetration measurement methods).24 Penetration depth analysis was performed by coating the MPAs with fluorescent nanospheres (FluoSpheres) followed by application to the skin. Cryo-preserved skin sections were analyzed by fluorescent microscopy (representative images are shown in Figure 3E and F) to determine the MPA penetration by measuring to the depth which FluoSpheres were delivered. In order to have replicate devices from a single application, we diced the 16 mm2 110-MPAs into 9 identical pieces, which were spread over three wells of a plate for detection (i.e. n = 3/mouse). The “260-MPAs” are significantly longer, at 260 μm, and penetrate significantly deeper into the dermis to 119 ± 15 μm (Figure 1F) (p-value
