Medical Device Link.

Originally Published IVD Technology June 2002

Processing Technologies

Improving performance in protein-based microarrays

Steven W. Metzger, Michael J. Lochhead, and David W. Grainger

Surface chemistries enable application of protein microarray technologies to diagnostics.

Fluorescence image of mouse DNA expression microarray. The transfer of DNA microarray methods to protein microarray diagnostics is currently under development. However, the technology transfer from DNA to protein arraying creates new challenges when working in complex sample environments typically found in diagnostic settings.

The evolution of protein microarrays from laboratory tools to clinical diagnostic products has been a natural technological progression, driven by the promise of improved healthcare and corresponding commercial potential. The microarray format provides a method for simultaneously probing thousands of biomolecular interactions using extremely small sample volumes and user-friendly instrumentation. The IVD industry has envisioned, for example, a diagnostic "protein chip" in which a microarray of antibodies or capture ligands are printed on a test piece. In theory, after such a chip has been innoculated with a patient sample (e.g., a drop of blood, urine, saliva), proteins would bind to highly specific antibodies or capture ligands and would be quickly measured to determine whether they are at normal or abnormal levels. Such a technology would enable healthcare practitioners to diagnose and administer appropriate therapies rapidly and with greater confidence.

Unfortunately, numerous technical hurdles must be overcome before such a protein-based diagnostic device becomes widely available. While protein microarray technology borrows extensively from DNA microarray analysis, it is still in its infancy due to fundamental differences between proteins and DNA. Unlike DNA, proteins are chemically and physically heterogeneous, have a three-dimensional structure that is critical to their function, and have no analogous amplification technique. Many proteins are also known to lose activity when bound to a solid surface, and most will adsorb nonspecifically to commonly used substrate materials.

When combined, all of these factors can compromise the performance of protein microarrays and limit specificity and sensitivity in diagnostic devices, particularly those that will handle real-world biological fluids. In human serum, for example, proteins that are targeted by diagnostics may have clinical significance at nanogram or even picogram per milliliter concentrations. Yet the serum itself contains milligrams per milliliter of background proteins, several of which notoriously adsorb to surfaces (e.g., fibrinogen, antibodies). The challenge of detecting clinically relevant analytes in a solution containing more than a millionfold confounding proteins places enormous demands on the assay surface.

Figure 1. Thematic illustration of the OptiChem coatings. (Click to enlarge)

Additional challenges include transferring traditional diagnostic technology to the microarray format. For example, transferring sandwich-format enzyme-linked immunosorbent assays (ELISAs) to a high-density microarray format mandates an increase in the number and total concentration of detection antibodies, since each analyte requires a unique detection antibody for quantification. With ELISAs, each detection antibody has a finite amount of nonspecific surface binding contributing to the background noise level. In the microarray format, however, the cumulative effect of numerous detection antibodies will eventually increase background noise levels until the noise is significant enough to limit the size of the array. Therefore, a critical component of the diagnostic protein microarray will be a sophisticated surface chemistry that efficiently couples the capture ligands and preserves their activity, while minimizing nonspecific adsorption of all other biomolecules. Another challenge will be developing a set of highly efficient and selective capture ligands, such as antibodies.

Industry has recognized that surface chemistry is a key technology for protein microarrays.¹ In addition to providing good specific binding and low background, surface chemistries must be chemically and physically robust, and must be compatible with existing and future microarray printing and scanning instrumentation. For widespread commercial applications such as diagnostics, the chemistry must be scalable (i.e., amenable to large-volume production) and cost-effective. This article examines several important protein microarray surface chemistry design parameters and provides performance data for a new surface chemistry that is being developed.

Protein versus Nucleic Acid Microarrays

Figure 2. Protein nonspecific binding was assessed by incubating the surfaces with rabbit anti-sheep immunoglobulin G-horseradish peroxidase (IgG-HRP). Tetramethylbenzidine peroxidase substrate was used to detect nonspecifically bound IgG-HRP in a colorimetric assay. Bare substrates and substrates blocked with bovine serum albumin (BSA) were run as controls. (Click to enlarge)

The most significant differences between protein microarrays and nucleic acid microarrays are found in the physical properties and stability of soluble proteins and DNA at interfaces. Because of these differences, vastly altered surface interactions emerge, which have important implications for designing protein microarray surfaces. Nucleic acid arrays tend to be more robust than protein arrays, partly as a result of intrinsic DNA stability and structure as well as sample preparation in assay designs. DNA or RNA can be isolated and then amplified from biological samples at high purity levels, and are frequently analyzed in protein-free solutions. In this scenario, a single-stranded DNA or RNA is homogeneously negatively charged and hydrated, resisting many of the intermolecular forces that plague proteins at surfaces. This resistence to surface perturbing forces provides a convenient basis for nucleic acid electrostatic surfaces immobilizing on cationic surfaces (e.g., poly-L-lysine or aminosilane), without compromising recognition of soluble single-stranded DNA targets.

By contrast, proteins have tremendous structural variability with a plethora of metastable native folded states and an associated hierarchy of long- and short-range forces that dictate and complicate their interactions with microarray surfaces.^2–4 Most proteins have both positively and negatively charged regions that interact with surfaces via long-range electrostatic forces. Globular proteins have multiple folded intramolecular domains interacting with each other via weak hydrophobic forces, and hydrogen bonds yielding tertiary and quaternary structures that are readily recognized by antibodies and other capture ligands. Upon encountering a surface, protein intramolecular folding forces are frequently exceeded by adsorption energies and entropies compelling these domains to unfold, or denature, in the vicinity of surfaces. This adsorption process may proceed despite the presence of electrostatic forces repelling proteins and surfaces of similar charge via ion bridging, hydration forces, and hydrophobic forces.⁴ Furthermore, the magnitude of intermolecular forces operating between soluble proteins and any surface is influenced by the adsorption affinity and the protein surface-adsorbed state, including weak dipole-dipole forces, dispersion forces, ion-induced dipoles, ion bridging, image-charge attraction or repulsion, and short-range hydration forces.

The result is that all proteins exhibit nonspecific surface activity distinct from DNA surface activity. Protein assays using serum, physiological fluids, or tissue/cell homogenate contain hundreds of proteins with intrinsic interfacial activity that produces loss of structure, surface-binding specificity, and analyte sensitivity. Furthermore, this intrinsic interfacial activity produces uncontrolled adsorption that is commonly referred to as nonspecific binding. In most diagnostic applications, a process of concentrating and amplifying protein analytes from complex protein solutions is currently not possible. The desire to detect specific analytes from cell lysate, serum, and tissue combined with nonspecific protein adsorption complicate the development of protein-based assay formats. Hydrophobic surfaces tend to exacerbate the protein adsorption problem, while certain hydrophilic surface chemistries (e.g., grafted hydrophilic polymers, hydrogels) reduce this problem.

Despite the inherent difficulties that are associated with proteins, researchers have demonstrated functional laboratory protein microarrays.^5–9 Such systems typically use covalent attachment or affinity binding of capture ligands followed by a blocking step (i.e., adsorption of bovine serum albumin) to limit nonspecific binding. The challenge is to extend this work and develop diagnostic tools that are robust, sensitive, easy to use, manufacturable, and cost-effective.

Surface Chemistry Challenges

Protein microarray-based molecular diagnostics present unique surface chemistry challenges that are not present in nucleic acid microarray applications. From a performance perspective, protein microarray surface chemistries must:

• Enable immobilization of specific recognition, or capture, elements (signal) in a manner that minimizes perturbations to native structure and maintains bioactivity over time.
• Minimize nonspecific protein surface binding (noise).

Surface Physical Adsorption

Bioimmobilization (e.g., affinity chromatography, biosensors, enzyme-immobilized bioreactors) and bioconjugation techniques have been common starting points for developing protein microarray surface chemistries.¹¹ One common method for microarray surface fabrication is immobilization to the assay substrate via physical adsorption.

Figure 3. Demonstration of covalent attachment of streptavidin to amine-reactive OptiChem through amines on the protein surface. Bound streptavidin is probed using biotin-horseradish peroxidase. In the control, reactive groups are chemically quenched with ethanolamine prior to incubating the streptavidin. The lack of binding on the control surface demonstrates that the attachment is covalent through the amino linkage. Duplicate slides demonstrate reproducibility. (Click to enlarge)

Signal. As stated previously, several nucleic acid platforms use electrostatic adsorption between cationic surfaces and anionic nucleic acid phosphate backbone groups for high-density immobilization. This technique has generated sufficient signal and specificity to satisfy current applications. Even though issues involving washoff, orientation, and target-capture efficiency have not been assessed in the scientific literature, nucleic acid immobilization seems to survive and function under these conditions.

Electrostatic adsorption may work for some classes of proteins, but it is not a universal approach, given the inherent variability of charge states in proteins. While some highly charged protein models (e.g., lysozyme, ribonuclease), can use electrostatically mediated adsorption, they also demonstrate a significant loss of native structure with such substrate interaction.⁴ Hydrophobically driven protein adsorption via spotting of libraries onto apolar, nonwetting polymer membranes, such as PET or PVDF, has also been reported.^8,12 However, this method is not desirable for protein arrays since proteins and capture ligands (e.g., antibodies) readily denature, lose bioactivity and active sites, and exhibit relatively poor shelf life on hydrophobic supports. Hydrophilic polymer environments also rely on passive adsorption, but shelf life and preservation of bioactivity are much less of an issue. However, the same adsorptive forces that are exploited for surface immobilization can lead to an exchange or a displacement with other surface active components in serum and from other processing steps.

Noise. Physical adsorption problems with capture ligands can be turned into an advantage for reducing nonspecific binding on array surfaces. The intrinsic tendency of globular protein to adsorb on surfaces can be exploited to prereact and mask, or block, surfaces with surface-active proteins—typically serum albumin or casein—from buffer solutions. Surface-active water-dispersible polymers, including block and other copolymers, can also be formulated into masking agents. These systems adsorb anywhere a spotted protein is not surface resident, limiting the ability of the surface to adsorb further proteins nonspecifically during assay steps.

Disadvantages with physical blocking include the undesirable use of serum-derived proteins (e.g., albumin), possible exchange or degradation (e.g., proteolysis) of the material during assay steps, the extra processing steps that are required for such surface passivation, and potential steric hindrance of small molecules.⁵ Additionally, while surface blocking effectively reduces noise by reducing nonspecific binding, it is limited in this capacity because noise always remains. Nevertheless, most current protein microarray methods employ such masking techniques to reduce nonspecific binding background.

Covalent Surface Attachment

Other factors on which surface chemistries are chosen and evaluated are consistency, efficiency, side reactions, and robustness of the attachment. Considering these factors, another method for microarray surface fabrication is direct covalent attachment.

Signal. Successful nucleic acid immobilization has been achieved using several different strategies for covalent attachment of modified synthetic nucleic acid oligomers to reactive surfaces. Most of these strategies use well-developed reactions between functional groups deliberately placed on the oligonucleotide terminus via synthetic methods and complementary chemistry on surfaces. Surface functional groups can be coated, photoimmobilized, or chemically grafted to many support chemistries using well-established coating methods (e.g., silanization, functional polymer coating). These methods are substrate-specific by design, and no covalent chemistry currently has the ability to collectively coat plastics, metals, metal oxides, and glass supports.

Direct covalent approaches to protein arrays represent a rational extension of this same strategy since several amino acids (e.g., lysine, glutamate, aspartate, histidine, tyrosine, cysteine, or any N- or C-terminus acids) provide suitable functional capacity for covalent immobilization and modification chemistries. Several commercial microarray slide chemistries are currently available, including aldehyde and epoxide silane films and reactive polymer coatings. While covalent chemistries are a viable approach for protein immobilization, they still have inherent limitations. For example, covalent approaches are hampered by competing side reactions (e.g., hydrolysis, elimination) and evaporation of carrier solvent throughout the coupling process. Less-reactive functional groups and protection chemistries have been utilized to minimize the effects of competing reactions, and optimized arraying procedures have been developed to mitigate evaporation effects.

The differences between coupling reactions in a solid-liquid interface and a bulk solution—and the disadvantages that result—are also underappreciated. In a solid-liquid interface, interfacial transport issues, steric constraints, altered kinetics, and reactivity access for the functional groups may contribute to consistently poorer surface reaction yields and immobilization efficiencies compared with bulk coupling.

Additionally, since this covalent technique is a volume-conservative mass-production process involving microliters of reagent solution producing thousands of nanoliter spots, it enables a "brute force" approach that can finesse the more subtle issues of coupling efficiencies. As a result, improved coupling efficiency can be realized by increasing the concentration of arrayed biomolecules, but it comes with an increase in reagent cost.

Covalent chemistries remain a suitable approach for protein arrays despite the technical challenges. Few thorough studies have focused on yield and consistency issues in various immobilization schemes for either DNA or protein array surfaces. In fact, for many surfaces there is little evidence to distinguish nonspecific capture ligand binding from specific immobilization. A significant fraction of array signal is likely generated from capture molecules that are intended for covalent immobilization but are nonspecifically adsorbed instead. It is also possible that current assay-signal intensity is generated from a minority of capture molecules that remain active and bioavailable after immobilization and that a large portion of the capture molecule population is inactivated by the immobilization process.

Figure 4. Performance of OptiChem surface chemistry compared with conventional polystyrene in an ELISA sandwich assay. Blocking procedures were utilized with the hydrophobic polystyrene surface. The target analyte is staphylococcal enterotoxin B. (Click to enlarge)

Noise. The presence of functional surface groups correlates with increased noise from nonspecific binding, physically or chemically, of undesired proteins to these groups. That is, these groups are both physically and chemically reactive with proteins. Therefore, after immobilization, the remaining unreacted surface groups are deactivated with suitable passivating small-capping molecules (e.g., NHS with ethanolamine) to reduce this source of noise.

Another strategy involves mixing reactive group chemistries as pendant side chains or terminal groups with otherwise nonreactive polar polymers in the surface matrix. Such nonfouling surface chemistries that exploit this approach have been available in the biomedical device industry for years.^13–15 Such chemistries are comprised of a hydrophilic surface-grafted polymer base matrix that is mixed in various fractions with reactive analogs capable of covalent chemistry. Polyethylene glycol (PEG), polyvinylpyrrolidone (PVP), polyvinyl alcohol (PVA), dextrans, and hyaluronic acid are commonly used for such applications. They can be derivatized or copolymerized with reactive groups that are appropriate for immobilizing proteins, while maintaining their intrinsically low nonspecific binding capacity.¹⁵ Performance in bioassays is not perfect, however, because deactivation of the reactive group can be problematic, and nonspecific binding is still finite and measurably unacceptable.

Surface Affinity Capture Ligands

Signal. Affinity chromatography has recently seen a number of developments brought about by the creation of a series of effective, noncovalent, highly specific interaction chemistries that can separate and bind biomolecules. For example, while the biotin-streptavidin pair has been highly touted, other affinity-capture pairings are equally interesting and potentially valuable in providing strong but noncovalent associations through solution-phase pairing with excellent built-in specificity. The microarray industry has also adopted such capture ligands to exploit these same specific interactions for surface immobilization. Affinity ligands require either physical or covalent immobilization of one partner in the affinity pair onto microarray supports, and are subject to the same constraints that are discussed above for these methods.

Figure 5. Signal-to-noise (S/N) ratios of OptiChem-coated substrates compared with traditional ELISA substrates. The solid black line indicates equivalent signal-to-noise ratios for the two approaches. Each data point represents one analyte concentration. Data points above the black line demonstrate superior signal-to-noise ratios for the OptiChem surface chemistry relative to conventional ELISA substrates. (Click to enlarge)

Additionally, because affinity ligands maintain specificity for their particular binding partner, these requisite chemistries must then logically exist for capture. Affinity ligands that are commonly exploited to immobilize biomolecules can be put into two categories: large-protein and small-molecule couplers. Protein ligands (e.g., streptavidin, protein A, and protein G) are conceptually at odds with surface chemistries that are designed to reduce nonspecific binding. Large protein molecules mask the underlying supporting film performance that is intended to reduce nonspecific binding, increasing the possibilities of generating surface nonspecific noise.

Small-molecule receptors such as biotin, phenylboronic and salicylhydroxamic acid, glutathione, and nickel ion (Ni²+–NTA) do not suffer as greatly from this problem. Once surface attached, these affinity ligand pairs represent a viable strategy for the immobilization of arrayed proteins and nucleic acids. The incorporation of recombinant tags during large-scale protein expression has further advanced this approach as a potential method for surface immobilization.¹⁶ However, the impact of these immobilization methods on nonspecific, nonanalyte binding and on possibly weakened interactions between recombinantly modified proteins and resultant surfaces has not been adequately assessed or reported.

Noise. While affinity capture is elegant in many regards, its strengths in signal generation, through effective capture, also result in weaknesses due to noise generation on surfaces. For example, protein-based affinity binding surfaces are not inert to further protein adsorption. Nonspecific binding limits assay sensitivity, and weakened substrate interaction will increase the likelihood of desorption and the removal of ligand-functionalized protein during subsequent microarray processing steps.

Applying Strategies

The objective in molecular diagnostics is to provide surface chemistry strategies to solve the variety of problems that are currently precluding the development of diagnostic applications for the detection and analysis of proteins using microarray technologies. Such chemistries should follow the generic requirements now common to DNA arrays, which are dictated by the unique challenges important to protein microarray development in the protein-containing milieu.

Accelr8 Technology Corp. (Denver) has developed a proprietary surface chemistry platform to address design criteria specific to protein microarray applications. These OptiChem coatings are formed with a single-step solvent casting process. The coating chemistry is composed of a multicomponent cross-linked matrix of organic materials that resists protein nonspecific binding and minimizes interactions that degrade bioactivity (see Figure 1). The coating does not require lengthy blocking steps to achieve low nonspecific binding, and affinity binding or reactive groups on tethered supports enable specific or covalent immobilization of proteins on surfaces. The chemistry is robust and capable of coating numerous surfaces such as plastics, glass, Si/SiO₂, and some metal substrates.

Performance

The low nonspecific binding of OptiChem coating chemistry was experimentally determined relative to bare substrates and bovine serum albumin (BSA)–blocked substrate (see Figure 2). The nonspecific binding of the coating chemistry was measured by probing the surface with a polyclonal horseradish peroxidase (HRP)–labeled antibody. A polyclonal antibody was chosen as a model protein with a propensity for surface nonspecific binding. The HRP label provided a means to detect the relative amount of nonspecifically bound antibody on the substrates.

The bare surfaces showed high nonspecific binding as expected, while the blocking procedure effectively reduced nonspecific binding. The blocking agent in an aqueous buffered system is expected to be susceptible to competitive displacement with the HRP-labeled antibody. Competitive displacement limits the effectiveness of the blocking strategy and can result in considerable nonspecific binding. By contrast, OptiChem is not susceptible to competitive displacement and has significantly lower nonspecific binding than the BSA-blocked surfaces. Such performance is achieved without blocking steps, a required practice for most commercially available microarray chemistries.

Another experiment demonstrated the covalent attachment of a 50-kD streptavidin capture protein to an amine-reactive OptiChem through primary amines on the protein surface (see Figure 3). The activity and density of bound streptavidin was probed using biotin-HRP. In the control, reactive groups were chemically deactivated with ethanolamine prior to incubating the streptavidin. The lack of binding on the control surface showed that the attachment is covalent through the amino linkage, and duplicate slides demonstrated reproducibility.

The specific binding signal also increases with protein load and is capable of generating signal strengths that are equivalent to an adsorbed control surface at the highest protein load. The correlation between signal and protein loading illustrates the ability to control protein density on the OptiChem surface.

Improved signal-to-noise ratios translated into increased sensitivity on the OptiChem surface relative to conventional substrates in sandwich ELISA applications (see Figure 4). The increase in signal is attributed to a higher density of capture antibody that is immobilized on the surface and improved presentation of antibody to analyte epitopes. The OptiChem coating does not require lengthy adsorption times, blocking steps, or incubations at elevated temperatures.

In addition, an ELISA application on the OptiChem surface chemistry resulted in gains in signal-to-noise ratios over traditional ELISA substrates (see Figure 5). This data set illustrates consistent performance across a number of experimental data points.

Conclusion

Extension of the microarray protein methodology to clinical diagnostics represents a natural technological progression. Ultimately, the objective of molecular diagnostics is to provide information at the protein level allowing for more-accurate detection and diagnosis of disease. In order to enable wide application of protein microarray technologies to diagnostics, a number of major surface chemistry challenges must be addressed.

Numerous companies and academic laboratories are addressing these challenges using different surface chemistry approaches. For example, OptiChem is a surface chemistry designed to address issues unique to protein microarray technology and represents an alternative to microarray surface technologies that rely on blocking strategies to reduce nonspecific binding. OptiChem enables the immobilization of specific recognition (capture) ligands in their native state and minimizes surface interactions that perturb structure and degrade bioactivity on surfaces. The rugged chemistry supports a variety of receptor-ligand and covalent immobilization strategies and is amenable to coating a variety of substrate materials. OptiChem combines a protein-favorable environment, optimized specific binding, and low nonspecific binding that translates into an enhanced signal-to-noise ratio and improved assay sensitivity.

References

1. P Mitchell, "A Perspective on Protein Microarrays," Nature Biotechnology 20, no. 3 (2002): 225–229.

2. Surface and Interfacial Aspects of Biomedical Polymers, vol. 2, ed. JD Andrade (New York: Plenum Press, 1985), 1–80.

3. V Hlady and J Buijs, "Protein Adsorption on Solid Surfaces," Current Opinion in Biotechnology 7 (1996): 72–77.

4. CA Haynes and W Norde, "Globular Proteins at Solid/Liquid Interfaces," Colloids Surfaces B 2 (1994): 517–566.

5. G MacBeath and SL Schreiber, "Printing Proteins as Microarrays for High-Throughput Function Determination," Science 289 (2000): 1760–1763.

6. H Zhu et al., "Global Analysis of Protein Activities Using Proteome Chips," Science 293 (2001): 2101–2105.

7. LG Mendoza et al., "High-Throughput Microarray-Based Enzyme-Linked Immunosorbent Assay (ELISA)," BioTechniques 27, no. 4 (1999): 778–788.

8. M Lueking et al., "Protein Microarrays for Gene Expression and Antibody Screening," Analytical Biochemistry 270, no. 1 (1999): 103–111.

9. B Schweitzer et al., "Multiplexed Protein Profiling on Microarrays by Rolling-Circle Amplification," Nature Biotechnology 20, no. 3 (2002): 359–365.

10. D Janssen, "Amplifying Alternatives," Genomics and Proteomics 2, no.1, (2002): 46–48.

11. GT Hermanson, Bioconjugate Techniques (London: Academic Press, 1996).

12. K Bussow et al., "A Method for Global Protein Expression and Antibody Screening on High-Density Filters of an Arrayed cDNA Library," Nucleic Acids Research 26, no. 21 (1998): 5007–5008.

13. AS Hoffman, "Non-Fouling Surface Technologies," Journal of Biomaterials Science: Polymer Edition 10, no. 10 (1999): 1011–1014.

14. E Ostuni, "A Survey of Structure-Property Relationships of Surfaces That Resist the Adsorption of Protein," Langmuir 17, no. 18 (2001): 5605–5620.

15. N Xia et al., "Functionalized Poly(ethylene Glycol)-Grafted Polysiloxane Monolayers for Control of Binding," Langmuir 18, no. 8 (2002): 3255–3262.

16. J Nilsson et al., "Affinity Fusion Strategies for Detection, Purification, and Immobilization of Recombinant Proteins," Protein Expression and Purification 11, no.1 (1997): 11–16.

Steven W. Metzger and Michael J. Lochhead, PhD, are principal scientists at Accelr8 Technology Corp. (Denver); they can be reached via smetzger@accelr8.com and mlochhead@accelr8.com, respectively. David W. Grainger, PhD, is a professor of chemistry at Colorado State University (Fort Collins, CO) and a chief technical advisor to Accelr8 Technology Corp.; he can be reached via grainger@lamar.colostate.edu.

Top Photo Courtesy of Brian Soriano/University of Colorado Health Sciences Center, Microarray Core Facility

Copyright ©2002 IVD Technology