1. Introduction
A microarray is typically defined as a collection of microscopic spots of biological solutions attached and arranged in a defined location on a solid surface that allows a massive number of parallel genotyping and/or phenotyping measurement. The fundamental principles of parallelized microspot (microarray) technology were conceived of on the basis of the ambient analyte model of Roger Ekins and colleagues in the late 1980s. However, the enormous interest evoked by microarray-based assays came in the late 1990s, when researchers were first able to utilize high-quality and reproducible results using DNA chips [1]. Since then, microarray technology, involving the direct transfer of biological molecules to a substrate using molecular printing techniques, has been extensively explored; there have been over 186,000 publications about the potential applications of microarray technology for scientific discoveries in a broad range of biological disciplines (Figure 1). The exponential growth in the number of publications in its first decade (1995–2005) has resulted in the maturation and potential application of microarray technology for biological research. However, the growth was less rapid in the following decade, known as the postgenomic era, which revealed the limitations, pitfalls and redesign considerations that must be addressed for the microarray approach to meet expectations in the field of proteomics [2,3]. Protein microarrays are not only emerging as an alternative tool to DNA chips to profile protein products, but are also expected to link both genomics and proteomics. The purpose of this review is to discuss the key innovations and technological limitations that must be overcome to develop a breakthrough platform for microarrays, which are no longer determined by technical advances, but by the expected experimental requirements.
Figure 1  Growth in the number of publications on microarrays. The number of papers published in the last two decades utilizing the term “microarray” or “protein microarray” in the title (searched using Thomson Scientific Web of Knowledge) has increased continuously. The number in the first decade (1995–2005) increased exponentially. However, the increase was less significant in the second decade (2006–2014).   The breakthrough in the fabrication of microarrays came through the development of two parallel key approaches (Figure 2). The first was the “delivery-to-chip”-type approach using a robotic spotter, where biological molecules, such as DNAs, are dispensed by spotting and physically attaching presynthesized biological molecules to a solid substrate using robotic printing technologies. In this approach, source plates (such as microtiter plates with a 96- or 384-well format) are first filled with a presynthesized biological solution, which is followed by the transfer of a few nanoliters of the solution per spot onto the microarray chip surface using pins or piezoelectric dispensers to produce an array of submicrometer features [4]. The probe spots can be applied by either contact printing (e.g., microstamping [5] or microcontact printing [6]) or noncontact printing (e.g., microspotting [7], inkjet printing [8], laser writing [9], photochemical printing [10] or nanosphere lithography [11,12]), as reviewed in [13]. The key technical challenges in creating high-quality microarrays include the efficient transfer of biological molecules onto substrates and achieving a high coupling efficiency. Microspotting has been the most straightforward and leading platform, where biological molecules are synthesized off a chip then transfer-printed on a chip using a robotic spotter. A multi-pin mode (typically 12 × 4 pins) instead of a low-throughput single-pin mode can produce arrays in a highly parallel manner, but has limitations in terms of the size and density of arrays, because the pitch between pins must match the pattern of the microtiter plates. Therefore, the spotter-based robotic arraying approach is practically limited to producing low-speed and low-density arrays of spots. Second, this approach requires tedious and time-consuming multistep reactions to spot individual biomolecules from a separately-synthesized and purified solution followed by their immobilization on the surface of the array. The successive loading and dispensing of the biological solution are time-consuming and can lead to cross-contamination and even the loss of functionality of biological molecules, such as proteins, owing to denaturation or dehydration. Thirdly, robotic spotting has relatively high start-up costs and is thus less affordable. Core facilities typically charge $80–$300 per spotted array [14].
Figure 2  Key approaches in the historical development and future outlook of microarrays.   The soft lithography method developed by Whitesides’ group [15] is a complementary molecular printing technique. Specifically, microcontact printing (µCp), a form of soft lithography, has flourished as a simple, reproducible, cost-effective patterning method. In µCp, elastomeric stamps are prepared by pouring an elastomer, such as poly(dimethylsiloxane) (PDMS), in a mold. The polymer is then cured in the mold, thereby generating a relief structure in the elastomeric stamp. The stamp can then be used to directly transfer molecules to a surface of interest. PDMS polymer has the advantages of low toxicity, high flexibility and low cost. Although PDMS molds have been utilized with great success in preparing microarrays, alternative mold materials, such as cross-linked perfluoropolyether (PFPE), can be used with some distinct advantages over PDMS-based materials [16]. However, µCp also has several drawbacks. Although the stamps can print over large areas, they can only deposit a single predetermined pattern, and thus, a new mold must be fabricated each time for each printing pattern. Furthermore, nonuniformity is another major concern, as the patterning resolution is affected and limited by stamp swelling and shrinkage. In addition, µCp has limitations in printing multicomponent materials, such as bioarrays, and therefore, a convenient strategy for the massive patterning of biomolecules using soft lithography is yet to be developed.
The second major approach to microarray fabrication was the development of photolithography by Stephen Fodor and colleagues at Affymax Research Institute, which has made an important contribution to the fabrication of “synthesis-on-chip”-type microarrays [17]. In this approach, arrays are constructed in situ by actually producing the biological molecules directly on the substrate with extremely high density using photochemistry and solid-phase synthesis [17,18]. Affymetrix Corporation has been a pioneer in this field by developing the GeneChip® [19]. Parallel on-chip gene synthesis has also recently been reported [20]. However, because of the complex nature of the chemical synthesis and the intrinsically expensive process involved in production, this method is usually inflexible and expensive. The capital investment required to build a clean room makes this method inaccessible to most researchers. Thus, this method is only feasible when applied to a limited set of materials. Therefore, the challenge to date has been to overcome the most pressing and ever-growing issue, i.e., increasing the current low capacity (a typical density of thousands of array elements) of microarrays to and ultradense (mega-giga) capacity. As shown in Figure 2, the next forthcoming issue in microarray technology is the inclusion of a key feature that can facilitate a link between individual DNA sequences (genetic information) and their protein products (functional information) for gene expression profiling and related applications. To meet these requirements, we have recently introduced and developed a basic technology termed “microintaglio printing” (µIP) for next-generation, high-density microarray production [21,22]. This method is a conceptually different low-cost and spotter-free arraying approach to the rapid prototyping of microscale patterns, where biological molecules are intaglio-printed on a substrate using a microfabricated printing mold. Furthermore, the incorporation of our original work on fabricating high-density DNA microreactor array chips makes this approach very simple and robust for achieving high-density DNA-linked protein microarrays [23,24]. Therefore, the µIP approach can provide a direct link between recombinantly-expressed proteins and the corresponding DNA sequence information. Consequently, genetic information in clone libraries can be integrated with their biological functional information and potential interaction partners (Figure 2).