Generation of miniaturized planar ecombinant antibody arrays using a microcantilever-based printer

Miniaturized (Ø 10 μm), multiplexed (>5-plex), and high-density (>100 000 spots cm−2) antibody arrays will play a key role in generating protein expression profiles in health and disease. However, producing such antibody arrays is challenging, and it is the type and range of available spotters which set the stage. This pilot study explored the use of a novel microspotting tool, BioplumeTM—consisting of an array of micromachined silicon cantilevers with integrated microfluidic channels—to produce miniaturized, multiplexed, and high-density planar recombinant antibody arrays for protein expression profiling which targets crude, directly labelled serum. The results demonstrated that 16-plex recombinant antibody arrays could be produced—based on miniaturized spot features (78.5 um2, Ø 10 μm) at a 7–125-times increased spot density (250 000 spots cm−2), interfaced with a fluorescent-based read-out. This prototype platform was found to display adequate reproducibility (spot-to-spot) and an assay sensitivity in the pM range. The feasibility of the array platform for serum protein profiling was outlined.


Introduction
Technology platforms which provide rapid, semi-/quantitative, multiplex, and high-resolution protein expression profile maps of crude samples, such as serum, in the microliter scale, are still, de facto, a rate-limiting step for biomarker discovery and the subsequent clinical development of personalized medicine [1,2]. Adopting parallelized and miniaturized microspot assays might resolve, as postulated by the ambient analyte theory [3,4], these technical limitations. To this end, high-performing antibody arrays, predominantly in microscale, have been successfully developed [5][6][7][8] and applied to various biomarker discovery endeavours [9][10][11][12]. However, producing antibody arrays is challenging, and it is the type Ten human recombinant single-chain Fragment variable (scFv) antibodies directed against six high-abundant human serum proteins, including C1q, C3, C4, C5, factor B (FB), and properdin (FP), and four low-abundant human serum proteins, including interleukin (IL)-6, IL-8, IL-12, and vascular endothelial growth factor (VEGF), were included in this study (table 1). The scFv antibodies were stringently selected from a phage display library designed in-house. The specificity, affinity (normally in the 1-10 nM range), and on-chip functionality of scFv antibodies derived from phage display were ensured by using: (i) stringent phage-display selection protocols [38], and (ii) a molecular design, adapted for microarray applications [5,39]. The specificity of these antibodies has previously been validated by using pure analytes; mixtures of pure analytes; well-characterized, standardized crude serum; and orthogonal methods, such as mass spectrometry (serum/tissue extract pull-down assays), ELISA, MesoScaleDiscovery (MSD) assay, immunohistochemistry, and/or cytometric bead assay; as well as by spiking and blocking experiments in crude sample formats (e.g. serum) [9,[40][41][42][43][44][45][46][47].
Purified C1q was purchased from ElectraBox Diagnostica (Tyresö, Sweden) and anonymized human serum samples were obtained from Skåne University Hospital (Lund, Sweden).

Production and purification of scFv antibodies
All scFv antibodies were produced in E. coli. Briefly, soluble scFvs, all carrying a C-terminal his 6 -tag, were purified from expression supernatants or periplasmic preparations by affinity chromatography on Ni 2+ -NTA (Qiagen, Hilden, Germany). Bound molecules were eluted with 250 mM imidazole (pH = 8), dialyzed against phosphate buffered saline (PBS) (pH = 7.4), and stored at 4°C until further use. The integrity and degree of purity of the produced scFvs were evaluated by 10% sodium dodecylsulfate polyacrylamide gel electrophoresis (SDS-PAGE) (Invitrogen, Carlsbad, CA, USA). The protein concentrations were determined by measuring the absorbance at 280 nm.

Labeling of samples
The serum samples were biotinylated using EZ-Link Sulfo-NHS-LC-Biotin (Pierce Protein Biology Products, Thermo Scientific, Inc., Rockford, IL, USA) based on a previously  , at a protein:dye ratio of 10:1 mg, following the protocol provided by the supplier, except for using 40% (v v −1 ) glycerol in PBS (pH 7.4) as a labelling buffer, and for performing the labelling on ice for 5 h due to the impaired stability of C1q in pure PBS.

Production of miniaturized antibody arrays
The microcantilever-based spotter, the Bioplume TM arrayer (Patent EP1509324) (NanoSen, Vichères, France) [34], was used to produce miniaturized antibody arrays. The spotting principle relies on microcantilever-based contact deposition. The microfabricated silicon cantilevers are 1500 μm long, 120 μm wide, and 5 μm thick, so the diameters of the obtained spots are in the 5-20 μm range (depending on contact time and surface hydrophobicity). The microcantilevers were filled by immersion in a 1 μL droplet of spotting reagent, thanks to capillary forces. The washing of the microcantilevers was achieved by dipping them into water for 30 s, and then allowing water evaporation to occur within a few seconds. The instrument was equipped with 3 cantilevers for printing.
The Bioplume TM was equipped with an automated xyz motion-control system, enabling arrays of spotted scFv antibodies (60-560 attomole antibody per spot) to be produced on black Polymer Maxisorp slides (NUNC A/S, Roskilde Denmark). The contact time was set to 300 min, resulting in 10 μm sized spots. The pitch-to-pitch distance was 20 (or 30) μm, giving a spot density of about 250 000 (or 111 100) spots per cm −2 . Three array layouts were printed. First, 1-plex arrays (labelled streptavidin) were printed in 15 3 × 3 subarrays in four separate printing areas. Second, 3-plex arrays (anti-C1q, anti-C3, and anti-VEGF) were printed in 5 30 × 30 subarrays (each antibody was printed in 27 replicates) in four separate printing areas. Third, 10 identical slides were printed, where each slide contained six identical arrays, composed of 16 spotting reagents (i.e. 16-plex), each spotted in 27 replicates (9 spots/cantilever × 3 cantilevers). The spotted reagents consisted of 10 different scFv, two of them (anti-C3 and anti-C5) being printed at 3 dilutions, as well as positive control (labelled streptavidin) and negative control (PBS). All samples were prepared in PBS containing 10% glycerol in order to prevent evaporation during the spotting procedure.

Processing of miniaturized antibody arrays
The quality of the miniaturized antibody arrays was first examined by evaluating the pattern of the deposited positive control, Alexa-647 labelled streptavidin, at 5 μm resolution, using a confocal microarray scanner, ScanArray Express (PerkinElmer Life & Analytical Sciences, Wellesley, MA, USA). A hydrophobic pen (DakoCytomation Pen, DakoCytomation, Glostrup, Denmark) was then used to draw a hydrophobic barrier around the arrays. Next, the arrays were blocked with 90 μl 5% (w v −1 ) fat-free milk powder (Semper AB, Sundbyberg, Sweden) in PBS for 1 h. All incubations were performed in a humidity chamber at room temperature. The arrays were washed three times with 90 μl 0.05% (v v −1 ) Tween-20 in PBS (PBS-T) and then incubated with 90 μl biotinylated serum sample, diluted from 1:2 to 1:75 (resulting in a final dilution of from 1:90 to 1:3375) or directly labelled C1q (0.01 nM to 100 nM) for 1 h. All samples were diluted in 1% (w v −1 ) fat free milk powder and 1%Tween-20 in PBS (PBS-MT). The arrays were then washed three times with 90 μl 0.05% (v v −1 ) PBS-T. Any subarrays incubated with biotinylated serum samples were also incubated with 90 μl 1 μg mL −1 Alexa-647 conjugated streptavidin diluted in PBS-MT for 1 h. Finally, the arrays were washed three times with 90 μl 0.05% (v v −1 ) PBS-T and one time with 90 μl PBS, directly dried under a stream of nitrogen gas, and scanned.

Analysis of scFv Nanoarrays
The arrays were scanned using a confocal microarray scanner (ScanArray Express, Perkin-Elmer Life & Analytical Science) with 5 μm resolution. The signal intensity of each spot was quantified using ScanArray Express software V4.0 (Perkin Elmer), using a spot diameter of 10 μm. The mean value of each set of nine replicate spots is reported (negative control subtracted). The limit of detection was defined two standard deviations above the negative control.

Results
In this conceptual study, we have explored the use of a novel microspotting tool, the Bioplume TM arrayer equipped with three printing needles, for producing miniaturized, high-density planar recombinant antibody arrays for protein expression profiling ( figure 1).
First, we tested the feasibility of the printing approach by producing 1-plex (labelled streptavidin) (figure 2(A)) and 3plex arrays (anti-C1q, anti-C3, and anti-VEGF) on black polymer Maxisorb slides ( figure 2(B)). The nozzle-surface contact time was set to 300 min, resulting in 10 μm-sized spots. The pitch-to-pitch distance was set to 20 μm, giving a spot density of about 250 000 spots per cm −2 . The data showed that 1-plex protein arrays readily could be produced ( figure 2(A)). The miniaturization was highlighted by the fact that about 57 spots could be printed within the same footprint as that of a conventional spot (Ø 150 μm) from an ordinary antibody microarray. The 3-plex antibody array showed that two of three antigens could be detected in a crude, biotinylated human serum sample (diluted 450 times) ( figure 2(B)). The fact that the low-abundant serum protein VEGF could not be detected might be sample-dependent (the serum concentration varies from serum to serum) and/or assay-dependent (below limit of detection). Still, the experiments outlined   (300 min contact time) were printed at a pitch-to-pitch distance of 30 μm, giving a spot density of about 111 100 spots per cm 2 (figure 3(C)). At first, we examined the amount of antibody needed for printing to yield detectable array signals (figure 4). The 10 scFv antibodies were printed at 0.5 mg ml −1 (560 attomole) (6 antibodies) or at the maximum concentration at hand, 0.38 mg ml −1 (430 attomole) (1 antibody) and 0.14 mg ml −1 (160 attomole) (3 antibodies) ( figure 4(A)). Targeting a crude, biotinylated serum sample (diluted 450 times), 7 of 10 antibodies gave detectable signals, including antibodies against both high-(C3, C4, C5, and C1q) and low-abundant (IL-8, IL-12, and VEGF) serum analytes ( figure 4(A)). Of note, targeting another serum samples, the antibodies directed against Factor B and IL-6 were also found to give detectable signals (figure 4(A)), indicating that printing ⩾160 attomole scFv might be sufficient. Next, we titrated the amount of printed antibodies (560/220/60 attomole) for two of the clones, anti-C3 (figures 4(A) and (B)) and anti-C5 ( figure 4(A)). In both cases, the data showed that 220 attomole of antibodies were required to give a detectable signal, indicating an appropriate amount of scFv antibody per spot required to produce functional arrays.
Next, we determined the printing efficiency of functional scFv by evaluating the spot-to-spot (figure 5), array-to-array (figure 6), and slide-to-slide (figure 7) reproducibility for three representative scFv antibodies after capturing both high-(C1q and C3) and low-abundant (IL-12) biotinylated targets in crude serum samples (diluted 450 and 1125 times). Being a pilot study, we chose to use the raw compiled array data without filtration (for e.g. spot imperfections due to e.g. dust particles) and normalization, in order to display the underlying reproducibility. The spot-to-spot reproducibility was based on nine replicate spots/needle, and expressed in terms of coefficient of variation (CV) ( figure 5). The results showed that the CV values varied in a needle-, antibody-, and/or analyte-dependent manner, and were in the range of 1-36% (mean CV 17%) (serum diluted 450 times) and 15-50% (mean CV 31%) (serum diluted 1125 times). It should be noted that the preferred serum dilution is 450 times [41,44], for which acceptable mean CV value was observed, especially considering that this is the 1st generation of array platform produced with the Bioplume TM microdispenser.  The array-to-array reproducibility, i.e. arrays analyzed on the same slide, was based on 27 spot replicates per subarray (using three needles) targeting serum diluted 450 times, and expressed in terms of CV values ( figure 6). Of note, the CV value within each set of 27 replicate spots was 13-61%, further supporting the spot-to-spot reproducibility observed above. The data showed that the array-to-array reproducibility, expressed as CV-values, was found to be in the range of 16-51% (mean CV 34%) ( figure 6). This range of variability was expected, considering that the data was not arrayto-array normalized, further indicating the feasibility of the printing approach.
The slide-to-slide reproducibility, i.e. arrays analyzed on different slides, was based on in total 3 slides, 6 subarrays, and 162 spots for each antibody, and expressed in terms of CV-values ( figure 7). The results showed that the CV values were in the range of 36-46%. Again, this range of variability was expected, and should be handled (minimized) in future efforts by adopting slide-to-slide normalization. Next, we determined the limit of detection (LOD) by titrating pure, directly labelled analyte (C1q) and/or wellcharacterized crude, biotinylated serum targeting 4 high-(C1q, C3, C4, and Factor B) and 1 low-abundant (IL-12) analytes ( figure 8). In the case of C1q, the LOD was found to be 10 pM for pure C1q ( figure 7(A)), while the LOD was reduced 12.5 times (125 pM/1:2250 dilution) when targeting crude, biotinylated serum ( figure 8(B)). In comparison, the OD was also found to be represented by a 2250-times dilution of crude, biotinylated serum for both C3 (3100 pm) (figure 8(C)) and IL-12 (70 fg ml −1 ) ( figure 8(D)). Further, the LOD was found to be 330 pM for C5, 880 pM for C4, and 6600 pM for Factor B. Hence, the set-up was found to display adequate LODs in the pM range when targeting various analytes in crude, directly biotinylated serum samples.  Finally, the feasibility of the 16-plex set up for serum protein profiling was evaluated by analyzing a well-characterized crude, biotinylated serum sample ( figure 9). The results showed that low non-specific background binding and dynamic specific (spot) signals were observed ( figure 9(A)). In more detail, the results showed that 7 of 10 antibodies, including both high-(C1q, C3, C4, and C5) and low-abundant (IL-8, IL-12, and VEGF) analytes, gave detectable signals when analyzing this particular serum sample. Hence, the data supported the feasibility of the printing process for producing miniaturized, planar recombinant antibody microarrays for protein expression profiling.

Discussion
In this pilot study, we explored the use of the new microspotting tool, the Bioplume TM arrayer, equipped with three printing needles, for producing antibody arrays. We demonstrated the feasibility of the printer for producing miniaturized, high-density planar recombinant antibody arrays for protein expression profiling. Hence, the early design of this printing device should be added to the short list of currently available instruments capable of generating such miniaturized arrays [25,30,32,33,34,48].
In the current configuration, the Bioplume TM instrument was equipped with three needles, so that multiplexity was achieved in incremental steps of 3, including a washing step of the needles in between the printing of new reagents. In comparison, using the DPN-technology (NLP3000 TM ), the multiplexity is extended in incremental steps of 6, 12, 24, or 48 by loading the printer with a new 6-, 12-, 24-, or 48-pen and a matching inkwell prepared with the next set of reagents. While possible, in practice it might be challenging to succeed with this due to the reorientation, etc that will be required after mounting a new pen. Hence, Bioplume TM works more like standard ink-jet printers used to produce conventional antibody microarrays (18 × 10 3 μm 2 (Ø ∼ 150 μm)-sized spots) that deposit the proteins, either one-by-one or ⩽8 simultaneously, depending on the printing-mode capability and number of nozzles; and the multiplexity is increased by washing the nozzles and reloading them with new reagents.
Considering that the set-up represents the 1st prototype, i.e. non-optimized, miniaturized antibody arrays produced using Bioplume TM , the spot-to-spot reproducibility was found to be satisfactory (mean CV 17%) when targeting directly biotinylated serum samples at the preferred dilution (450 times) using non-filtered data [41,44]. When reducing the spot size, the quality of each miniaturized replicate spot is more vulnerable to imperfections caused by, for example, dust particles. Filtering the data for such defects will improve the spot-to-spot reproducibility in future efforts. In addition, by exploiting the use of clean room facilities and by optimizing the platform (e.g. choice of solid support, spotting conditions) for recombinant antibody array production in future efforts, it is very likely that this key parameter can be even further improved [5]. In comparison, we observed mean/ median CV-values of 12% (range 5-27%) and 4% (range 0-11%) for the 1st and second generation of miniaturized recombinant antibody arrays produced based on DPN technology (NLP3000) [33,34]. Noteworthy, the CV-value has been found to be <30% for conceptual antibody nanoarrays [14-23, 25, 30], while conventional antibody microarrays (established platforms) regularly display CVs of <10% (recombinant antibody microarrays) [49] and <20% (polyclonal and monoclonal antibody microarrays [50,51].   The array-to-array and slide-to-slide reproducibility were found to be higher, which could be expected in evaluating non-normalized data [5,33,34]. Being a pilot study, we identified the efforts to apply such normalization strategies [5,33,34] to be more suited to follow-up studies, where the aim would be to evolve the platform from a proof-of-concept set-up to an optimized (established) methodology aiming to run (clinical) applications. In this context, it might still be of interest to note that CV values in the range of 5% was observed for the second generation of miniaturized antibody arrays (DPN technology) [33,34], while CV values in the same range [5][6][7]13] or better (1.6%) (Wingren et al unpublished data) regularly have been observed for established conventional antibody microarray set-ups. In comparison, the inter-and intra-assay CV values for well-optimized conventional ELISAs are frequently ⩽15%.
The set-up interfaced with a confocal fluorescent-based scanner for sensitive sensing, also compatible for reading high-density arrays. At the time when the wet-lab experiments were performed, we only had access to a scanner with 5.0 μm resolution. This explains the somewhat rough view of the array images; however, this should not impair our evaluation and conclusion in a significant manner. Of note, we have recently acquired a high-resolution (0.5 μm resolution) scanner, to be adopted for future work. The LOD was found to be in the pM range, which is highly adequate for tentative proteomic applications [5,52], and in the same range as that observed for the miniaturized recombinant antibody arrays produced using DPN technology [33,34]. Hence, most of the targeted analytes could also be successfully detected in directly labelled, crude serum samples, representing one of the most common clinical sample format. In comparison, LODs in the pM to fM range have been observed for high-end conventional antibody microarrays [5,11,41,44,53] (Wingren et al unpublished observations), but these set-ups have then been subjected to significant optimization and method refinement.
Taken together, in this pilot study, we have demonstrated for the first time the feasibility of the novel microdispensing tool Bioplume TM for producing miniaturized, multiplexed, high-density planar recombinant antibody arrays-displaying adequate performance-for protein expression profiling of crude, directly labelled proteomes. In this, we have successfully extended the range of candidate printers that can be used to produce miniaturized arrays. Further, the Bioplume TM printer provides a key edge towards manufacture of more high-density antibody array designs. Fully automated platforms based on printers such as Bioplume TM could represent the next generation of printers, setting the stage for highthroughput screening efforts. Miniaturized (recombinant) antibody microarrays will be an essential tool for large-scale, multiplexed profiling of crude proteomes, in both health and disease research, opening up novel opportunities for biomarker discovery, while consuming minimal amounts of both reagents and samples.