- Email: [email protected]
Microscale mesoarrays created by dip-pen nanolithography for screening of protein–protein interactions
Biosensors and Bioelectronics 26 (2011) 4667–4673
Contents lists available at ScienceDirect
Biosensors and Bioelectronics journal homepage: www.elsevier.com/locate/bios
Microscale mesoarrays created by dip-pen nanolithography for screening of protein–protein interactions David G. Thompson a , Ekaterina O. McKenna b , Andrew Pitt b , Duncan Graham a,∗ a b
Centre for Molecular Nanometrology, Department of Pure and Applied Chemistry, University of Strathclyde, Glasgow, G1 1XL, UK Institute of Molecular, Cell and Systems Biology, College of Medical, Veterinary and Life Sciences, University of Glasgow, Glasgow, G12 8QQ, UK
a r t i c l e
i n f o
Article history: Received 31 January 2011 Received in revised form 11 April 2011 Accepted 21 April 2011 Available online 16 June 2011 Keywords: Mesoarrays Protein–protein interaction Bionanotechnology DPN
a b s t r a c t Using microarrays to probe protein–protein interactions is becoming increasingly attractive due to their compatibility with highly sensitive detection techniques, selectivity of interaction, robustness and capacity for examining multiple proteins simultaneously. The major drawback to using this approach is the relatively large volumes and high concentrations necessary. Reducing the protein array spot size should allow for smaller volumes and lower concentrations to be used as well as opening the way for combination with more sensitive detection technologies. Dip-Pen Nanolithography (DPN) is a recently developed technique for structure creation on the nano to microscale with the capacity to create biological architectures. Here we describe the creation of miniaturised microarrays, ‘mesoarrays’, using DPN with protein spots 400× smaller by area compared to conventional microarrays. The mesoarrays were then used to probe the ERK2-KSR binding event of the Ras/Raf/MEK/ERK signalling pathway at a physical scale below that previously reported. Whilst the overall assay efficiency was determined to be low, the mesoarrays could detect KSR binding to ERK2 repeatedly and with low non-specific binding. This study serves as a first step towards an approach that can be used for analysis of proteins at a concentration level comparable to that found in the cellular environment. © 2011 Elsevier B.V. All rights reserved.
1. Introduction Unlocking the mechanism of action of the mammalian proteome is the key to truly understanding how human cells repair, regulate and replicate. These highly complex processes are controlled by a vast number of cellular protein signalling pathways and gaining even a small insight into some of the protein–protein interactions that comprise the pathways could lead to new medical treatments for a large range of diseases. The methods currently used to analyse the proteome are diverse and numerous. Assays to determine protein–protein interactions require volumes and concentrations many orders of magnitude larger than are actually present within a single cell or even a small cluster of cells. Within their approximate 1 pL volume a single cell can have up to 8 × 109 protein molecules but this value covers the entire cellular proteome including ultra-low concentration receptors and initiators (<100 molecules) through to mid-range signalling enzymes (1000–10,000 molecules) and up to high concentration structural proteins (108 ) (Cooper and Hausman, 2007; Lodish et al., 2000; Sims and Allbritton, 2007). Proteins of interest need to be expressed, amplified and purified so that their inter-
∗ Corresponding author. Tel.: +44 141 548 4701; fax: +44 141 548 4787. E-mail address: [email protected] (D. Graham). 0956-5663/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.bios.2011.04.040
actions can be detected, a process that can take weeks with no guarantee of success. Most protein pathways take the form of a cascade, so whilst an initial interaction may be minor and only involve a few proteins, the effects of this are multiplied and can have a direct effect upon a huge range of cellular functions. Direct analysis of the cellular environment for the ultra-low concentration signalling cascade initiators could allow for extremely early diagnosis of diseases, even before the appearance of any symptoms, allowing treatment to begin significantly earlier than is currently possible. One of the methods that is becoming more widely used for proteome analysis is protein microarray analysis (Haab et al., 2001; Kukar et al., 2002; MacBeath and Schreiber, 2000; Patterson and Aebersold, 2003; Schena, 2005). This technique uses a microarrayer to drop subnanolitre volumes of protein onto a functionalised surface to which the protein adheres, creating a spot 150–300 m in diameter. The surface-bound protein is then probed by the addition of a possible interactor, the presence of which can be determined using antibodies followed by fluorescence detection with a microarray scanner. The key advantage of the microarray-based approach is the ability to detect interactions of the probe components to multiple distinct interactors immobilised in parallel on the surface, but suffers from the same drawbacks as other techniques, namely the relatively large volumes and concentrations necessary for the assays (approximately 100 L volume, 1 × 10−5 M concen-
4668
D.G. Thompson et al. / Biosensors and Bioelectronics 26 (2011) 4667–4673
tration). By miniaturising the microarray spot size it should be possible to use smaller total amounts and sample volumes of array probe protein, conserving the precious biological stocks, lowering costs, facilitating incorporation into microfluidic chip systems and ultimately achieving increased sensitivity. Ultra-small surface immobilised protein features have been created by several nanofabrication techniques: electron beam lithography (EBL) (Christman et al., 2008), nanoimprint lithography (Cai and Ocko, 2005), micro- (or nano-) contact printing (Bernard et al., 2000; Renault et al., 2002) and dip-pen nanolithography (DPN) (Lee et al., 2006) have all been used for this purpose (Wingren and Borrebaeck, 2007). Among all the listed approaches DPN arguably holds the most promise as a technique for controlled ultra-small array formation. DPN is a scanning probe microscopic technique based upon atomic force microscopy (AFM)(Piner et al., 1999) and involves the coating of an AFM cantilever with an ‘ink’ of interest and bringing the cantilever into contact with a surface. In the infancy of this technique the ink was a simple thiol-functionalised molecule written onto a planar gold surface (Piner et al., 1999). Now this has evolved to encompass a large range of capture chemistries and ink molecules including DNA, peptides and antibodies written into nanoarray formats (Basnar and Willner, 2009; Bellido et al., 2010; Lee et al., 2002, 2006; Nam et al., 2004; Piner et al., 1999; www.nanoink.net). Primarily reported as a method for nanoscale architecture generation, DPN is a remarkably flexible technique also allowing the creation of microscale features. Critically, the technique is suitable for creating arrays containing libraries of multiple different agents, such as proteins, with the use of multi-tip pens and inkwells (www.nanoink.net). DPN requires a surface that is both amenable to the mode of writing and suitable for a particular biological application. Thin, microscopically smooth nitrocellulose is a promising candidate because array-based monitoring of protein–protein interactions on such surfaces is possible (www.invitrogen.com). These surfaces are characterised as having a low background fluorescence and a high binding capacity combined with the ability to maintain protein functionality. Binding of the protein to the surface is non-specific, but strong when the proteins, delivered to the surface in an aqueous buffer, are subsequently dried into the nitrocellulose matrix. There is no requirement for the immobilised protein to have any additional functionality or tag to aid surface adsorption. The aim of the approach reported here was to work in the scale dimension between microarrays and nanoarrays and optimise it for bioanalysis. The term ‘mesoarray’ has been devised to describe array spots that are 1–20 m in size. This intermediate size dimension of 1–20 m is of interest as it can be analysed without using specialised optical techniques or other high-powered microscopy, which is necessary for nanoarrays, but still represents a significant size decrease compared with the conventional microarray. Here we report the creation of mesoarrays on microscopically smooth nitrocellulose using DPN to write the proteins into features 400× smaller than the conventional microarray spot and their use in an assay consisting of part of the Ras/Raf/MEK/ERK signalling pathway. This work is designed to be a step towards the creation of active globular protein nanoarrays on nitrocellulose and their use in directly probing the cellular proteome.
2. Materials and methods 2.1. Chemicals and reagents 11-Mercaptoundecanoic acid, analytical grade ethanol, PBS and Tween20 were purchased from Sigma Aldrich, UK. Bovine serum albumin (BSA) Fraction V was purchased from Roche Diagnostics, UK. PATHTM slides were purchased from Fischer Scientific,
Fig. 1. Schematic diagram representing the plane calculation points (P1, P2, and P3) and their placement compared to the planned protein mesoarray (represented by circles). The diagram is approximately to scale.
UK. Monoclonal mouse anti-GST and Alexafluor546 goat antimouse IgG were purchased from Invitrogen, UK. NanoInk® AFM probes and inkwells were purchased from LOT Oriel, UK. Proteins ERK2-His and KSR-GST were expressed in E. coli and affinity purified.
2.2. Creation of protein mesoarrays All protein mesoarrays were created using a NanoInk nanolithography platform 2000 instrument (NLP). Prior to arraying the AFM probes used were cleaned in an oxygen plasma for 20 s followed by passivation via immersion in an ethanolic 1 mmol/L mercaptoundecanioc acid solution for 30 min to minimize the risk of the gold coated tips denaturing the protein. Whilst this was underway a new PATH slide was placed in the environmental chamber of the NLP and the humidity raised to ∼65%. The passivated tip was then washed in ethanol for 30 s and dried using dry nitrogen and positioned in the NLP. A fresh 1:1 solution was made up consisting of the protein to be written and a commercial protein transfer buffer supplied by NanoInk. 0.4 L of this solution was then added to an inkwell already within the chamber. Before protein writing could commence it was necessary to determine the 3 dimensional plane of the surface. This was calculated by bringing the tip into contact at three points, designated P1, P2 and P3, at the approximate edges of the planned mesoarray (Fig. 1). The degree of tip contact was determined by observing the change in reflectivity of the gold cantilever upon touching the surface. The tips used were the ‘A-frame’ NanoInk single pen A tips and B tips. These tips had a sufficiently high spring constant (0.100 N/m and 0.046 N/m, respectively) to allow writing without damaging the surface or deforming upon inking with protein. Once the plane was calculated the tip was maneuvered into contact with the inkwell followed by arraying onto the slide. The tip contact time was 10 ms. For each 5 × 5 mesoarray created a single row of five spots was written after each inkwell dip. The tip was then dipped in the inkwell again and the same spots were rewritten to ensure maximum protein content before moving to the next row. The arraying process was undertaken at room temperature. Once arraying was complete the mesoarrays were allowed to equilibrate for 1 hour in the environmental chamber. The array slide was then removed and placed in a vacuum desiccator for 12 h to maintain protein stability prior to assaying.
D.G. Thompson et al. / Biosensors and Bioelectronics 26 (2011) 4667–4673
4669
2.3. Protein–protein interaction assay The mesoarray slides were placed in a SIMplexTM (16 MultiArray system, Gentel® Biosciences) or AHM-64 (Gel Company) microarray probing frame. Block solution (2% BSA in PBS) was applied at 150 L per well at room temperature without agitation with the wells sealed with Mylar ate sealer. After 1 h of incubation the block solution was shaken out of the frame, and the probe protein solution was applied immediately. Probe protein ERK2 binding fragment of KSR, a GST tagged version (Jacobs et al., 1999) (KSRGST) (purified stock concentration 0.8 mg/mL) was prepared to the desired concentration in protein probe buffer (PBS with 2% BSA). Samples were added at volumes of 70 L to the arrays in frame wells in accordance with the desired order and incubated for 1 h at room temperature without agitation under a seal. A zero probe control (probing buffer not containing any probe protein) was always applied to at least one array well. After probe protein incubation the bulk of the sample was removed by shaking it out of the frame and PBST (PBS with 0.1% Tween20) was applied to the wells and quickly shaken off as an initial rinse. Subsequently PBST was replenished in the wells and the frame was placed on a shaker for 5 min. This was repeated 5 times, followed by removal of the wash buffer. For detection, primary antibody monoclonal mouse anti-GST (stock at 0.5 mg/mL) was prepared at 1:1000, and added at 70 L per well, and incubated with the arrays for 45 min. The array was washed by the same procedure as previous. Reporter antibody, AlexaFluor® 546 labelled goat anti-mouse IgG (2 mg/mL stock) prepared at a 1:1000 dilution was applied in the dark for 30 min; application was again followed by the same wash procedure. The final wash solution was shaken out of the frame, and the sample was allowed to dry in the dark in the frame overnight, removed from the frame the next day and analysed.
Fig. 2. Schematic diagram of an ERK2-KSR binding microarray. In this format a capture protein, ERK2 (A) is arrayed onto a nitrocellulose surface. A probe protein, a KSR-GST (glutathione-s-transferase) fusion protein (B) is added which binds through its KSR part to the ERK2. The KSR used is the ERK2 binding fragment of KSR and is referred to as KSR throughout. This binding is detected by the addition of a monoclonal mouse anti-GST antibody (C) followed by the subsequent addition of a reporter antibody (AlexaFluor® 546 labelled goat anti-mouse IgG) (D).
The mesoarrays were analysed using a Nikon Eclipse E600 microscope fitted with TRITC filter set (540 nm excitation, 565 nm emission) and the required microscope objective. The data was collected using Metamorph® software and background corrected before image analysis. Averaged spot image analysis was performed using ImageJ, specifically the Microarray Analysis plugin. Using this software the intensity for each spot was calculated and the average and standard deviation calculated using MS Excel.
successfully written onto the nitrocellulose support in a controlled, reproducible manner and that the protein adhered to the surface, a fluorescently modified protein was written and analysed using fluorescence microscopy. Fig. 3 shows a visible image of the AlexaFluor® 546 labelled goat anti-mouse IgG spots taken immediately after writing (the AFM cantilever is present in the foreground) and a series of fluorescence images of the same spots. These results show that the fluorescent protein was successfully written as well as demonstrating the consistency of spot size, circularity, array dimensions and fluorescence intensity created by this methodology. All spots were approximately 10 m in diameter with the fluorescence intensity evenly distributed to every spot of the array. The spot size is controlled by a number of factors including tip dwell time, temperature and humidity (Piner et al., 1999). With regular microarray spots having a diameter of approximately 200 m these meso spots are 400× smaller by area.
3. Results and discussion
3.2. Probing the ERK2-KSR interaction with mesoarrays
3.1. Creation and characterisation of fluorescent mesoarrays
Once the technique for successfully writing proteins by this approach was established, the next step was to write ERK2 and perform a successful assay with KSR. In this protein–protein interaction screening assay the aim was to ascertain whether two proteins bind together. This was accomplished by performing a series of experiments, increasing the concentration of the binding protein until the spot intensity reached a maximum. Using concentrations higher than that necessary for maximum signal diminishes the spot intensity achieved due to the increase in non-specific binding of probe protein to the nitrocellulose surface. Fig. 4 shows the results of a functional protein–protein interaction mesoarray. The fluorescence microscopy images in Fig. 4 show defined protein spot features. The spots are approximately 10 m in diameter, circular and although there is some intra-spot ‘patchiness’, intraand inter-spot variation does not exceed that observed in a conventional microarray. No ‘doughnut effect’ features where the signal intensity is concentrated at the edges of the spot is present. The plot of the average normalized spot intensity showed a consistent increase in spot intensity with the concomitant increase in
2.4. Analysis of the mesoarrays
The mitogen-activated protein kinase (MAPK) cascades have been identified as vital for cell proliferation and regulation whilst their aberrant behaviour has been linked to a range of cancer types (Friday and Adjei, 2008; McCubrey et al., 2007; Roberts and Der, 2007). One of the primary focal points for cancer research is the Ras/Raf/MEK/ERK pathway which consists of multiple interactions and regulators. Several proteins of interest are components of this pathway, two of which are extracellular signal-regulated kinase (ERK2) and kinase suppressor of Ras (KSR) where KSR acts as a scaffold for ERK2 and other proteins (Kolch, 2005). The interaction of ERK2 and KSR can be detected using a microarray formatted assay (see supplementary material), the binding architecture of which is shown in Fig. 2. This interaction has a reported dissociation constant of 1.2 mol/L (Fantz et al., 2001). The main aim was to use DPN to miniaturise this microarray format where the capture protein (ERK2) is written onto a nitrocellulose surface. However, to ensure that capture protein was being
4670
D.G. Thompson et al. / Biosensors and Bioelectronics 26 (2011) 4667–4673
Fig. 3. Visible (A) and fluorescence (B, C, D) images of reporter antibody (AlexaFluor® 546 labelled goat anti-mouse IgG) written onto a nitrocellulose surface. The arrays consist of 25 spots that were written in a 5 × 5 array format (B, C) and 4 spots written in a 2 × 2 format (B, D) with 20 m spacing between spots in both x and y dimensions (pitch). The visible image was taken using the nanolithography platform 2000 (NLP) viewing camera fitted with a Mituloyo M Plan 20× objective lens. The fluorescent images were taken with a Nikon Plan fluor 20× objective lens (B) and a Nikon 60× objective lens (C, D) both with a 100 ms accumulation time.
KSR-GST concentration. The graph indicates that a concentration of 5 M is sufficient to adequately probe ERK2-KSR binding. At this concentration the spot intensity is at is maximum compared to the background fluorescence caused by non-specific binding. At higher KSR-GST concentrations non-specific binding to the nitrocellulose surface increases which in turn causes an increase in the background fluorescence. This is observed in the results as the decrease in averaged intensity at the highest KSR-GST concentration (Fig. 4F). The dissociation constant was calculated to be 5.6 mol/L using this data which is in good agreement with the reported value (1.2 mol/L) (Fantz et al., 2001) and the dissociation constant calculated using a protein microarray (2.07 mol/L (see supplementary material). As well as the zero probe control shown in Fig. 4A, a second control was performed where a noninteracting protein (BSA) was written and assayed with KSR-GST to ensure the writing method was not causing any non-specific binding. That control did not display any detectable features. These results prove that the mesoarrays created using DPN can be used in a protein–protein interaction screening assay format. Recently (Bellido et al., 2010) reported a mathematical method to calculate the number of protein molecules in a DPN-created feature. Using this equation it should be possible to quantify the absolute number of ERK2 molecules detected by the assay which can be compared to the number written into each spot to give an
indication of the efficacy of the assay. To do this an array comprised solely of the same fluorescent reporter antibody used to detect the ERK2-KSR interaction with a fluorescent intensity equivalent to that of an ERK2 mesoarray was written. The fluorescent mesoarray was subjected to the same assay conditions to ensure any non-specific binding was also equivalent. Fig. 5 shows the comparison between the positive mesoarray assay and the mesoarray of adsorbed fluorescent antibody. The data in Fig. 5 shows that the positive assay and fluorescent antibody array have approximately the same fluorescence intensity. Using the reported equation it was possible to calculate that each ERK2 spot contained 1.10 × 106 protein molecules (concentration = 0.515 g/L, MW = 45 kDA, contact angle = 70◦ (Starov et al., 2003)) whilst each reporter antibody spot contains 3.21 × 104 protein molecules (concentration = 0.05 g/L, MW = 150 kDA, contact angle = 70◦ ). Comparing the values for the theoretically estimated initial protein numbers of the ERK2 spots and fluorescent antibody spots that have equivalent florescent intensities leads to the calculation of an overall assay efficiency value of 3%. This figure means that for every 100 ERK2 molecules originally written, only 3 are actually detected. Whilst the efficiency value appears low with regards to an analytical assay it is sufficient for a screening assay where the desired result is either positive or negative (i.e. binding or no binding). There is a number of possible reasons for this low value
D.G. Thompson et al. / Biosensors and Bioelectronics 26 (2011) 4667–4673
4671
Fig. 4. Fluorescence images and plot of the average normalized spot intensity of a successful protein–protein interaction screening mesoarray assay. The fluorescence images correspond to the different concentrations of probe protein KSR-GST used: (A) zero probe (control), (B) 1 mol/L, (C) 2.5 mol/L, (D) 5 mol/L, (E) 7.5 mol/L and (F) 10 mol/L. The plot shows the averaged spot intensity of the 25 spots in each 5 × 5 array with error bars corresponding to twice the standard deviation. For a full list of spot intensity values see supplementary material. The fluorescence images were taken with a Nikon PlanFluor 40× objective lens at with an accumulation time of 5 s.
including: over saturation of the nitrocellulose surface with ERK2 upon writing to such an extent that the nitrocellulose cannot retain all the protein molecules written; steric crowding factors at each probing step limiting the number of interactions possible, especially at the antibody addition step due to the antibody being much larger than ERK2 and KSR; and the variable protein–surface and protein–protein binding efficiencies. To truly harness the full potential of this successful miniaturisation of the protein–protein interaction screening microarray
alterations would have to made to the assay to completely utilise the benefits of the ‘meso’ aspects of the arrays as although the protein spots are significantly smaller they are still assayed and analysed using ‘micro’ techniques that use large volumes and have low sensitivities. Synergizing these mesoarrays with nano techniques and/or microfluidic systems to deliver ultra-low volumes of sample directly to the arrays and analysis with extremely sensitive detection technologies would allow detection at near proteomic levels.
4672
D.G. Thompson et al. / Biosensors and Bioelectronics 26 (2011) 4667–4673
Fig. 5. Fluorescent images and intensity plots of a positive ERK2 mesoarray after probing (A), and a fluorescent antibody mesoarray written at a concentration of 50 g/mL and subjected to identical assay conditions (B). The averaged spot intensity for the probe mesoarray was 42.8 ± 7.1 and for the fluorescent mesoarray was 43.1 ± 9.8. For a full list of intensity values see supplementary material.
4. Conclusion In conclusion, protein–protein interaction mesoarrays have been created and successfully used to probe an interaction in the Ras/Raf/MEK/ERK cascade on a physical scale never previously reported. Although the assay had a low efficiency the results demonstrated that the capture molecule was not denatured by the writing process, the binding functionality of the protein was maintained, the created arrays were reproducible and the results obtained from a positive assay were clear and quantifiable. With full exploitation of the positive attributes of nanoscale writing and detection leading to further advances in this area, this work serves as a first step to pushing analytical limits towards ultra-low (proteomic) concentration detection.
Acknowledgements This work was funded through the joint BBSRC/EPSRC Radical Solutions for Researching the Proteome (RASOR) funding award. DG acknowledges the Royal Society for support from a Wolfson Research Merit Award. The authors gratefully appreciate Prof. Walter Kolch for helpful discussions. The authors wish to thank Dr Sarah Cumming and Dr. Susan Gannon for expression and purification of the recombinant proteins used in this study, Gary L. Johnson and Kerry Kornfeld for supplying the protein construct plasmids, Eleanor Irvine and Dr. Aaron Hernandez-Santana for assistance with the NLP instrument, Dr. Nikolaj Gadegaard and Dr. Robert Stokes for their contribution to the initial stage AFM analysis, and the Strathclyde Institute of Biophotonics for use of the Nikon micro-
scope and David Blatchford for his assistance and expertise with the microscope. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.bios.2011.04.040. References Basnar, B., Willner, I., 2009. Small 5 (1), 28–44. Bellido, E., Miguel, R., Ruiz-Molina, d., Lostao, D., Maspoch, A.D., 2010. Adv. Mater. 22 (3), 352–355. Bernard, A., Renault, J.P., Michel, B., Bosshard, H.R., Delamarche, E., 2000. Adv. Mater. 12 (14), 1067–1070. Cai, Y.G., Ocko, B.M., 2005. Langmuir 21 (20), 9274–9279. Christman, K.L., Schopf, E., Broyer, R.M., Li, R.C., Chen, Y., Maynard, H.D., 2008. J. Am. Chem. Soc. 131 (2), 521–527. Cooper, G.M., Hausman, R.E., 2007. The Cell, 4th ed. ASM Press, Washington D.C. Fantz, D.A., Jacobs, D., Glossip, D., Kornfield, K., 2001. J. Biol. Chem. 276 (29), 27256–27265. Friday, B.B., Adjei, A.A., 2008. Clin. Cancer Res. 14, 342–346. Haab, B., Dunham, M., Brown, P., 2001. Genome Biol. 2 (2), 1-4–4.13. Jacobs, D., Glossip, D., Xing, H., Muslin, A.J., Kornfeld, K., 1999. Gene. Dev. 13, 163–175. Kolch, W., 2005. Nat. Rev. Mol. Cell Biol. 6, 827–837. Kukar, T., Eckenrode, S., Gu, Y., Lian, W., Megginson, M., She, J.-X., Wu, D., 2002. Anal. Biochem. 306 (1), 50–54. Lee, K.-B., Park, S.-J., Mirkin, C.A., Smith, J.C., Mrksich, M., 2002. Science 295, 1702–1705 (5560). Lee, M., Kang, D.-K., Yang, H.-K., Park, K.-H., Choe, S.Y., Kang, C., Chang, S.-I., Han, M.H., Kang, I.-C., 2006. Proteomics 6 (4), 1094–1103. Lodish, H., Berk, A., Zipursky, L.S., Matsudaira, P., Baltimore, D., Darnell, J., 2000. Molecular Cell Biology, 4th ed. W.H. Freeman, New York.
D.G. Thompson et al. / Biosensors and Bioelectronics 26 (2011) 4667–4673 MacBeath, G., Schreiber, S.L., 2000. Science 289, 1760–1763 (5485). McCubrey, J.A., Steelman, L.S., Chappell, W.H., Abrams, S.L., Wong, E.W.T., Chang, F., Lehmann, B., Terrian, D.M., Milella, M., Tafuri, A., Stivala, F., Libra, M., Basecke, J., Evangelisti, C., Martelli, A.M., Franklin, R.A., 2007. BBA-Mol. Cell Res. 1773 (8), 1263–1284. Nam, J.-M., Han, S.W., Lee, K.-B., Liu, X., Ratner, M.A., Mirkin, C.A., 2004. Angew. Chemie. Int. Ed. 43 (10), 1246–1249. Patterson, S.D., Aebersold, R., 2003. Nat. Genet. 33, 311–323. Piner, R.D., Zhu, J., Xu, F., Hong, S., Mirkin, C.A., 1999. Science 283, 661–663 (5402).
4673
Renault, J.P., Bernard, A., Bietsch, A., Michel, B., Bosshard, H.R., Delamarche, E., Kreiter, M., Hecht, B., Wild, U.P., 2002. J. Phys. Chem. B 107 (3), 703–711. Roberts, P.J., Der, C.J., 2007. Oncogene 26, 3291–3310. Schena, M., 2005. Protein Microarrays. Jones and Bartlett. Sims, C.E., Allbritton, N.L., 2007. Lab Chip 7 (4), 423–440. Starov, V.M., Zhdanov, S.A., Kosvintsev, S.R., Sobolev, V.D., Velarde, M.G., 2003. Adv. Colloid Interface 104 (1–3), 123–158. Wingren, C., Borrebaeck, C.A.K., 2007. Drug Discov. Today 12 (19–20), 813–819, www.invitrogen.com, www.nanoink.net, Multiplexed protein arrays application notes.
Contents lists available at ScienceDirect
Biosensors and Bioelectronics journal homepage: www.elsevier.com/locate/bios
Microscale mesoarrays created by dip-pen nanolithography for screening of protein–protein interactions David G. Thompson a , Ekaterina O. McKenna b , Andrew Pitt b , Duncan Graham a,∗ a b
Centre for Molecular Nanometrology, Department of Pure and Applied Chemistry, University of Strathclyde, Glasgow, G1 1XL, UK Institute of Molecular, Cell and Systems Biology, College of Medical, Veterinary and Life Sciences, University of Glasgow, Glasgow, G12 8QQ, UK
a r t i c l e
i n f o
Article history: Received 31 January 2011 Received in revised form 11 April 2011 Accepted 21 April 2011 Available online 16 June 2011 Keywords: Mesoarrays Protein–protein interaction Bionanotechnology DPN
a b s t r a c t Using microarrays to probe protein–protein interactions is becoming increasingly attractive due to their compatibility with highly sensitive detection techniques, selectivity of interaction, robustness and capacity for examining multiple proteins simultaneously. The major drawback to using this approach is the relatively large volumes and high concentrations necessary. Reducing the protein array spot size should allow for smaller volumes and lower concentrations to be used as well as opening the way for combination with more sensitive detection technologies. Dip-Pen Nanolithography (DPN) is a recently developed technique for structure creation on the nano to microscale with the capacity to create biological architectures. Here we describe the creation of miniaturised microarrays, ‘mesoarrays’, using DPN with protein spots 400× smaller by area compared to conventional microarrays. The mesoarrays were then used to probe the ERK2-KSR binding event of the Ras/Raf/MEK/ERK signalling pathway at a physical scale below that previously reported. Whilst the overall assay efficiency was determined to be low, the mesoarrays could detect KSR binding to ERK2 repeatedly and with low non-specific binding. This study serves as a first step towards an approach that can be used for analysis of proteins at a concentration level comparable to that found in the cellular environment. © 2011 Elsevier B.V. All rights reserved.
1. Introduction Unlocking the mechanism of action of the mammalian proteome is the key to truly understanding how human cells repair, regulate and replicate. These highly complex processes are controlled by a vast number of cellular protein signalling pathways and gaining even a small insight into some of the protein–protein interactions that comprise the pathways could lead to new medical treatments for a large range of diseases. The methods currently used to analyse the proteome are diverse and numerous. Assays to determine protein–protein interactions require volumes and concentrations many orders of magnitude larger than are actually present within a single cell or even a small cluster of cells. Within their approximate 1 pL volume a single cell can have up to 8 × 109 protein molecules but this value covers the entire cellular proteome including ultra-low concentration receptors and initiators (<100 molecules) through to mid-range signalling enzymes (1000–10,000 molecules) and up to high concentration structural proteins (108 ) (Cooper and Hausman, 2007; Lodish et al., 2000; Sims and Allbritton, 2007). Proteins of interest need to be expressed, amplified and purified so that their inter-
∗ Corresponding author. Tel.: +44 141 548 4701; fax: +44 141 548 4787. E-mail address: [email protected] (D. Graham). 0956-5663/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.bios.2011.04.040
actions can be detected, a process that can take weeks with no guarantee of success. Most protein pathways take the form of a cascade, so whilst an initial interaction may be minor and only involve a few proteins, the effects of this are multiplied and can have a direct effect upon a huge range of cellular functions. Direct analysis of the cellular environment for the ultra-low concentration signalling cascade initiators could allow for extremely early diagnosis of diseases, even before the appearance of any symptoms, allowing treatment to begin significantly earlier than is currently possible. One of the methods that is becoming more widely used for proteome analysis is protein microarray analysis (Haab et al., 2001; Kukar et al., 2002; MacBeath and Schreiber, 2000; Patterson and Aebersold, 2003; Schena, 2005). This technique uses a microarrayer to drop subnanolitre volumes of protein onto a functionalised surface to which the protein adheres, creating a spot 150–300 m in diameter. The surface-bound protein is then probed by the addition of a possible interactor, the presence of which can be determined using antibodies followed by fluorescence detection with a microarray scanner. The key advantage of the microarray-based approach is the ability to detect interactions of the probe components to multiple distinct interactors immobilised in parallel on the surface, but suffers from the same drawbacks as other techniques, namely the relatively large volumes and concentrations necessary for the assays (approximately 100 L volume, 1 × 10−5 M concen-
4668
D.G. Thompson et al. / Biosensors and Bioelectronics 26 (2011) 4667–4673
tration). By miniaturising the microarray spot size it should be possible to use smaller total amounts and sample volumes of array probe protein, conserving the precious biological stocks, lowering costs, facilitating incorporation into microfluidic chip systems and ultimately achieving increased sensitivity. Ultra-small surface immobilised protein features have been created by several nanofabrication techniques: electron beam lithography (EBL) (Christman et al., 2008), nanoimprint lithography (Cai and Ocko, 2005), micro- (or nano-) contact printing (Bernard et al., 2000; Renault et al., 2002) and dip-pen nanolithography (DPN) (Lee et al., 2006) have all been used for this purpose (Wingren and Borrebaeck, 2007). Among all the listed approaches DPN arguably holds the most promise as a technique for controlled ultra-small array formation. DPN is a scanning probe microscopic technique based upon atomic force microscopy (AFM)(Piner et al., 1999) and involves the coating of an AFM cantilever with an ‘ink’ of interest and bringing the cantilever into contact with a surface. In the infancy of this technique the ink was a simple thiol-functionalised molecule written onto a planar gold surface (Piner et al., 1999). Now this has evolved to encompass a large range of capture chemistries and ink molecules including DNA, peptides and antibodies written into nanoarray formats (Basnar and Willner, 2009; Bellido et al., 2010; Lee et al., 2002, 2006; Nam et al., 2004; Piner et al., 1999; www.nanoink.net). Primarily reported as a method for nanoscale architecture generation, DPN is a remarkably flexible technique also allowing the creation of microscale features. Critically, the technique is suitable for creating arrays containing libraries of multiple different agents, such as proteins, with the use of multi-tip pens and inkwells (www.nanoink.net). DPN requires a surface that is both amenable to the mode of writing and suitable for a particular biological application. Thin, microscopically smooth nitrocellulose is a promising candidate because array-based monitoring of protein–protein interactions on such surfaces is possible (www.invitrogen.com). These surfaces are characterised as having a low background fluorescence and a high binding capacity combined with the ability to maintain protein functionality. Binding of the protein to the surface is non-specific, but strong when the proteins, delivered to the surface in an aqueous buffer, are subsequently dried into the nitrocellulose matrix. There is no requirement for the immobilised protein to have any additional functionality or tag to aid surface adsorption. The aim of the approach reported here was to work in the scale dimension between microarrays and nanoarrays and optimise it for bioanalysis. The term ‘mesoarray’ has been devised to describe array spots that are 1–20 m in size. This intermediate size dimension of 1–20 m is of interest as it can be analysed without using specialised optical techniques or other high-powered microscopy, which is necessary for nanoarrays, but still represents a significant size decrease compared with the conventional microarray. Here we report the creation of mesoarrays on microscopically smooth nitrocellulose using DPN to write the proteins into features 400× smaller than the conventional microarray spot and their use in an assay consisting of part of the Ras/Raf/MEK/ERK signalling pathway. This work is designed to be a step towards the creation of active globular protein nanoarrays on nitrocellulose and their use in directly probing the cellular proteome.
2. Materials and methods 2.1. Chemicals and reagents 11-Mercaptoundecanoic acid, analytical grade ethanol, PBS and Tween20 were purchased from Sigma Aldrich, UK. Bovine serum albumin (BSA) Fraction V was purchased from Roche Diagnostics, UK. PATHTM slides were purchased from Fischer Scientific,
Fig. 1. Schematic diagram representing the plane calculation points (P1, P2, and P3) and their placement compared to the planned protein mesoarray (represented by circles). The diagram is approximately to scale.
UK. Monoclonal mouse anti-GST and Alexafluor546 goat antimouse IgG were purchased from Invitrogen, UK. NanoInk® AFM probes and inkwells were purchased from LOT Oriel, UK. Proteins ERK2-His and KSR-GST were expressed in E. coli and affinity purified.
2.2. Creation of protein mesoarrays All protein mesoarrays were created using a NanoInk nanolithography platform 2000 instrument (NLP). Prior to arraying the AFM probes used were cleaned in an oxygen plasma for 20 s followed by passivation via immersion in an ethanolic 1 mmol/L mercaptoundecanioc acid solution for 30 min to minimize the risk of the gold coated tips denaturing the protein. Whilst this was underway a new PATH slide was placed in the environmental chamber of the NLP and the humidity raised to ∼65%. The passivated tip was then washed in ethanol for 30 s and dried using dry nitrogen and positioned in the NLP. A fresh 1:1 solution was made up consisting of the protein to be written and a commercial protein transfer buffer supplied by NanoInk. 0.4 L of this solution was then added to an inkwell already within the chamber. Before protein writing could commence it was necessary to determine the 3 dimensional plane of the surface. This was calculated by bringing the tip into contact at three points, designated P1, P2 and P3, at the approximate edges of the planned mesoarray (Fig. 1). The degree of tip contact was determined by observing the change in reflectivity of the gold cantilever upon touching the surface. The tips used were the ‘A-frame’ NanoInk single pen A tips and B tips. These tips had a sufficiently high spring constant (0.100 N/m and 0.046 N/m, respectively) to allow writing without damaging the surface or deforming upon inking with protein. Once the plane was calculated the tip was maneuvered into contact with the inkwell followed by arraying onto the slide. The tip contact time was 10 ms. For each 5 × 5 mesoarray created a single row of five spots was written after each inkwell dip. The tip was then dipped in the inkwell again and the same spots were rewritten to ensure maximum protein content before moving to the next row. The arraying process was undertaken at room temperature. Once arraying was complete the mesoarrays were allowed to equilibrate for 1 hour in the environmental chamber. The array slide was then removed and placed in a vacuum desiccator for 12 h to maintain protein stability prior to assaying.
D.G. Thompson et al. / Biosensors and Bioelectronics 26 (2011) 4667–4673
4669
2.3. Protein–protein interaction assay The mesoarray slides were placed in a SIMplexTM (16 MultiArray system, Gentel® Biosciences) or AHM-64 (Gel Company) microarray probing frame. Block solution (2% BSA in PBS) was applied at 150 L per well at room temperature without agitation with the wells sealed with Mylar ate sealer. After 1 h of incubation the block solution was shaken out of the frame, and the probe protein solution was applied immediately. Probe protein ERK2 binding fragment of KSR, a GST tagged version (Jacobs et al., 1999) (KSRGST) (purified stock concentration 0.8 mg/mL) was prepared to the desired concentration in protein probe buffer (PBS with 2% BSA). Samples were added at volumes of 70 L to the arrays in frame wells in accordance with the desired order and incubated for 1 h at room temperature without agitation under a seal. A zero probe control (probing buffer not containing any probe protein) was always applied to at least one array well. After probe protein incubation the bulk of the sample was removed by shaking it out of the frame and PBST (PBS with 0.1% Tween20) was applied to the wells and quickly shaken off as an initial rinse. Subsequently PBST was replenished in the wells and the frame was placed on a shaker for 5 min. This was repeated 5 times, followed by removal of the wash buffer. For detection, primary antibody monoclonal mouse anti-GST (stock at 0.5 mg/mL) was prepared at 1:1000, and added at 70 L per well, and incubated with the arrays for 45 min. The array was washed by the same procedure as previous. Reporter antibody, AlexaFluor® 546 labelled goat anti-mouse IgG (2 mg/mL stock) prepared at a 1:1000 dilution was applied in the dark for 30 min; application was again followed by the same wash procedure. The final wash solution was shaken out of the frame, and the sample was allowed to dry in the dark in the frame overnight, removed from the frame the next day and analysed.
Fig. 2. Schematic diagram of an ERK2-KSR binding microarray. In this format a capture protein, ERK2 (A) is arrayed onto a nitrocellulose surface. A probe protein, a KSR-GST (glutathione-s-transferase) fusion protein (B) is added which binds through its KSR part to the ERK2. The KSR used is the ERK2 binding fragment of KSR and is referred to as KSR throughout. This binding is detected by the addition of a monoclonal mouse anti-GST antibody (C) followed by the subsequent addition of a reporter antibody (AlexaFluor® 546 labelled goat anti-mouse IgG) (D).
The mesoarrays were analysed using a Nikon Eclipse E600 microscope fitted with TRITC filter set (540 nm excitation, 565 nm emission) and the required microscope objective. The data was collected using Metamorph® software and background corrected before image analysis. Averaged spot image analysis was performed using ImageJ, specifically the Microarray Analysis plugin. Using this software the intensity for each spot was calculated and the average and standard deviation calculated using MS Excel.
successfully written onto the nitrocellulose support in a controlled, reproducible manner and that the protein adhered to the surface, a fluorescently modified protein was written and analysed using fluorescence microscopy. Fig. 3 shows a visible image of the AlexaFluor® 546 labelled goat anti-mouse IgG spots taken immediately after writing (the AFM cantilever is present in the foreground) and a series of fluorescence images of the same spots. These results show that the fluorescent protein was successfully written as well as demonstrating the consistency of spot size, circularity, array dimensions and fluorescence intensity created by this methodology. All spots were approximately 10 m in diameter with the fluorescence intensity evenly distributed to every spot of the array. The spot size is controlled by a number of factors including tip dwell time, temperature and humidity (Piner et al., 1999). With regular microarray spots having a diameter of approximately 200 m these meso spots are 400× smaller by area.
3. Results and discussion
3.2. Probing the ERK2-KSR interaction with mesoarrays
3.1. Creation and characterisation of fluorescent mesoarrays
Once the technique for successfully writing proteins by this approach was established, the next step was to write ERK2 and perform a successful assay with KSR. In this protein–protein interaction screening assay the aim was to ascertain whether two proteins bind together. This was accomplished by performing a series of experiments, increasing the concentration of the binding protein until the spot intensity reached a maximum. Using concentrations higher than that necessary for maximum signal diminishes the spot intensity achieved due to the increase in non-specific binding of probe protein to the nitrocellulose surface. Fig. 4 shows the results of a functional protein–protein interaction mesoarray. The fluorescence microscopy images in Fig. 4 show defined protein spot features. The spots are approximately 10 m in diameter, circular and although there is some intra-spot ‘patchiness’, intraand inter-spot variation does not exceed that observed in a conventional microarray. No ‘doughnut effect’ features where the signal intensity is concentrated at the edges of the spot is present. The plot of the average normalized spot intensity showed a consistent increase in spot intensity with the concomitant increase in
2.4. Analysis of the mesoarrays
The mitogen-activated protein kinase (MAPK) cascades have been identified as vital for cell proliferation and regulation whilst their aberrant behaviour has been linked to a range of cancer types (Friday and Adjei, 2008; McCubrey et al., 2007; Roberts and Der, 2007). One of the primary focal points for cancer research is the Ras/Raf/MEK/ERK pathway which consists of multiple interactions and regulators. Several proteins of interest are components of this pathway, two of which are extracellular signal-regulated kinase (ERK2) and kinase suppressor of Ras (KSR) where KSR acts as a scaffold for ERK2 and other proteins (Kolch, 2005). The interaction of ERK2 and KSR can be detected using a microarray formatted assay (see supplementary material), the binding architecture of which is shown in Fig. 2. This interaction has a reported dissociation constant of 1.2 mol/L (Fantz et al., 2001). The main aim was to use DPN to miniaturise this microarray format where the capture protein (ERK2) is written onto a nitrocellulose surface. However, to ensure that capture protein was being
4670
D.G. Thompson et al. / Biosensors and Bioelectronics 26 (2011) 4667–4673
Fig. 3. Visible (A) and fluorescence (B, C, D) images of reporter antibody (AlexaFluor® 546 labelled goat anti-mouse IgG) written onto a nitrocellulose surface. The arrays consist of 25 spots that were written in a 5 × 5 array format (B, C) and 4 spots written in a 2 × 2 format (B, D) with 20 m spacing between spots in both x and y dimensions (pitch). The visible image was taken using the nanolithography platform 2000 (NLP) viewing camera fitted with a Mituloyo M Plan 20× objective lens. The fluorescent images were taken with a Nikon Plan fluor 20× objective lens (B) and a Nikon 60× objective lens (C, D) both with a 100 ms accumulation time.
KSR-GST concentration. The graph indicates that a concentration of 5 M is sufficient to adequately probe ERK2-KSR binding. At this concentration the spot intensity is at is maximum compared to the background fluorescence caused by non-specific binding. At higher KSR-GST concentrations non-specific binding to the nitrocellulose surface increases which in turn causes an increase in the background fluorescence. This is observed in the results as the decrease in averaged intensity at the highest KSR-GST concentration (Fig. 4F). The dissociation constant was calculated to be 5.6 mol/L using this data which is in good agreement with the reported value (1.2 mol/L) (Fantz et al., 2001) and the dissociation constant calculated using a protein microarray (2.07 mol/L (see supplementary material). As well as the zero probe control shown in Fig. 4A, a second control was performed where a noninteracting protein (BSA) was written and assayed with KSR-GST to ensure the writing method was not causing any non-specific binding. That control did not display any detectable features. These results prove that the mesoarrays created using DPN can be used in a protein–protein interaction screening assay format. Recently (Bellido et al., 2010) reported a mathematical method to calculate the number of protein molecules in a DPN-created feature. Using this equation it should be possible to quantify the absolute number of ERK2 molecules detected by the assay which can be compared to the number written into each spot to give an
indication of the efficacy of the assay. To do this an array comprised solely of the same fluorescent reporter antibody used to detect the ERK2-KSR interaction with a fluorescent intensity equivalent to that of an ERK2 mesoarray was written. The fluorescent mesoarray was subjected to the same assay conditions to ensure any non-specific binding was also equivalent. Fig. 5 shows the comparison between the positive mesoarray assay and the mesoarray of adsorbed fluorescent antibody. The data in Fig. 5 shows that the positive assay and fluorescent antibody array have approximately the same fluorescence intensity. Using the reported equation it was possible to calculate that each ERK2 spot contained 1.10 × 106 protein molecules (concentration = 0.515 g/L, MW = 45 kDA, contact angle = 70◦ (Starov et al., 2003)) whilst each reporter antibody spot contains 3.21 × 104 protein molecules (concentration = 0.05 g/L, MW = 150 kDA, contact angle = 70◦ ). Comparing the values for the theoretically estimated initial protein numbers of the ERK2 spots and fluorescent antibody spots that have equivalent florescent intensities leads to the calculation of an overall assay efficiency value of 3%. This figure means that for every 100 ERK2 molecules originally written, only 3 are actually detected. Whilst the efficiency value appears low with regards to an analytical assay it is sufficient for a screening assay where the desired result is either positive or negative (i.e. binding or no binding). There is a number of possible reasons for this low value
D.G. Thompson et al. / Biosensors and Bioelectronics 26 (2011) 4667–4673
4671
Fig. 4. Fluorescence images and plot of the average normalized spot intensity of a successful protein–protein interaction screening mesoarray assay. The fluorescence images correspond to the different concentrations of probe protein KSR-GST used: (A) zero probe (control), (B) 1 mol/L, (C) 2.5 mol/L, (D) 5 mol/L, (E) 7.5 mol/L and (F) 10 mol/L. The plot shows the averaged spot intensity of the 25 spots in each 5 × 5 array with error bars corresponding to twice the standard deviation. For a full list of spot intensity values see supplementary material. The fluorescence images were taken with a Nikon PlanFluor 40× objective lens at with an accumulation time of 5 s.
including: over saturation of the nitrocellulose surface with ERK2 upon writing to such an extent that the nitrocellulose cannot retain all the protein molecules written; steric crowding factors at each probing step limiting the number of interactions possible, especially at the antibody addition step due to the antibody being much larger than ERK2 and KSR; and the variable protein–surface and protein–protein binding efficiencies. To truly harness the full potential of this successful miniaturisation of the protein–protein interaction screening microarray
alterations would have to made to the assay to completely utilise the benefits of the ‘meso’ aspects of the arrays as although the protein spots are significantly smaller they are still assayed and analysed using ‘micro’ techniques that use large volumes and have low sensitivities. Synergizing these mesoarrays with nano techniques and/or microfluidic systems to deliver ultra-low volumes of sample directly to the arrays and analysis with extremely sensitive detection technologies would allow detection at near proteomic levels.
4672
D.G. Thompson et al. / Biosensors and Bioelectronics 26 (2011) 4667–4673
Fig. 5. Fluorescent images and intensity plots of a positive ERK2 mesoarray after probing (A), and a fluorescent antibody mesoarray written at a concentration of 50 g/mL and subjected to identical assay conditions (B). The averaged spot intensity for the probe mesoarray was 42.8 ± 7.1 and for the fluorescent mesoarray was 43.1 ± 9.8. For a full list of intensity values see supplementary material.
4. Conclusion In conclusion, protein–protein interaction mesoarrays have been created and successfully used to probe an interaction in the Ras/Raf/MEK/ERK cascade on a physical scale never previously reported. Although the assay had a low efficiency the results demonstrated that the capture molecule was not denatured by the writing process, the binding functionality of the protein was maintained, the created arrays were reproducible and the results obtained from a positive assay were clear and quantifiable. With full exploitation of the positive attributes of nanoscale writing and detection leading to further advances in this area, this work serves as a first step to pushing analytical limits towards ultra-low (proteomic) concentration detection.
Acknowledgements This work was funded through the joint BBSRC/EPSRC Radical Solutions for Researching the Proteome (RASOR) funding award. DG acknowledges the Royal Society for support from a Wolfson Research Merit Award. The authors gratefully appreciate Prof. Walter Kolch for helpful discussions. The authors wish to thank Dr Sarah Cumming and Dr. Susan Gannon for expression and purification of the recombinant proteins used in this study, Gary L. Johnson and Kerry Kornfeld for supplying the protein construct plasmids, Eleanor Irvine and Dr. Aaron Hernandez-Santana for assistance with the NLP instrument, Dr. Nikolaj Gadegaard and Dr. Robert Stokes for their contribution to the initial stage AFM analysis, and the Strathclyde Institute of Biophotonics for use of the Nikon micro-
scope and David Blatchford for his assistance and expertise with the microscope. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.bios.2011.04.040. References Basnar, B., Willner, I., 2009. Small 5 (1), 28–44. Bellido, E., Miguel, R., Ruiz-Molina, d., Lostao, D., Maspoch, A.D., 2010. Adv. Mater. 22 (3), 352–355. Bernard, A., Renault, J.P., Michel, B., Bosshard, H.R., Delamarche, E., 2000. Adv. Mater. 12 (14), 1067–1070. Cai, Y.G., Ocko, B.M., 2005. Langmuir 21 (20), 9274–9279. Christman, K.L., Schopf, E., Broyer, R.M., Li, R.C., Chen, Y., Maynard, H.D., 2008. J. Am. Chem. Soc. 131 (2), 521–527. Cooper, G.M., Hausman, R.E., 2007. The Cell, 4th ed. ASM Press, Washington D.C. Fantz, D.A., Jacobs, D., Glossip, D., Kornfield, K., 2001. J. Biol. Chem. 276 (29), 27256–27265. Friday, B.B., Adjei, A.A., 2008. Clin. Cancer Res. 14, 342–346. Haab, B., Dunham, M., Brown, P., 2001. Genome Biol. 2 (2), 1-4–4.13. Jacobs, D., Glossip, D., Xing, H., Muslin, A.J., Kornfeld, K., 1999. Gene. Dev. 13, 163–175. Kolch, W., 2005. Nat. Rev. Mol. Cell Biol. 6, 827–837. Kukar, T., Eckenrode, S., Gu, Y., Lian, W., Megginson, M., She, J.-X., Wu, D., 2002. Anal. Biochem. 306 (1), 50–54. Lee, K.-B., Park, S.-J., Mirkin, C.A., Smith, J.C., Mrksich, M., 2002. Science 295, 1702–1705 (5560). Lee, M., Kang, D.-K., Yang, H.-K., Park, K.-H., Choe, S.Y., Kang, C., Chang, S.-I., Han, M.H., Kang, I.-C., 2006. Proteomics 6 (4), 1094–1103. Lodish, H., Berk, A., Zipursky, L.S., Matsudaira, P., Baltimore, D., Darnell, J., 2000. Molecular Cell Biology, 4th ed. W.H. Freeman, New York.
D.G. Thompson et al. / Biosensors and Bioelectronics 26 (2011) 4667–4673 MacBeath, G., Schreiber, S.L., 2000. Science 289, 1760–1763 (5485). McCubrey, J.A., Steelman, L.S., Chappell, W.H., Abrams, S.L., Wong, E.W.T., Chang, F., Lehmann, B., Terrian, D.M., Milella, M., Tafuri, A., Stivala, F., Libra, M., Basecke, J., Evangelisti, C., Martelli, A.M., Franklin, R.A., 2007. BBA-Mol. Cell Res. 1773 (8), 1263–1284. Nam, J.-M., Han, S.W., Lee, K.-B., Liu, X., Ratner, M.A., Mirkin, C.A., 2004. Angew. Chemie. Int. Ed. 43 (10), 1246–1249. Patterson, S.D., Aebersold, R., 2003. Nat. Genet. 33, 311–323. Piner, R.D., Zhu, J., Xu, F., Hong, S., Mirkin, C.A., 1999. Science 283, 661–663 (5402).
4673
Renault, J.P., Bernard, A., Bietsch, A., Michel, B., Bosshard, H.R., Delamarche, E., Kreiter, M., Hecht, B., Wild, U.P., 2002. J. Phys. Chem. B 107 (3), 703–711. Roberts, P.J., Der, C.J., 2007. Oncogene 26, 3291–3310. Schena, M., 2005. Protein Microarrays. Jones and Bartlett. Sims, C.E., Allbritton, N.L., 2007. Lab Chip 7 (4), 423–440. Starov, V.M., Zhdanov, S.A., Kosvintsev, S.R., Sobolev, V.D., Velarde, M.G., 2003. Adv. Colloid Interface 104 (1–3), 123–158. Wingren, C., Borrebaeck, C.A.K., 2007. Drug Discov. Today 12 (19–20), 813–819, www.invitrogen.com, www.nanoink.net, Multiplexed protein arrays application notes.