DNA-encoded immunoassay in picoliter drops: a minimal cell-free approach

Based on the remarkably specific antibody-antigen interaction, immunoassays have emerged as indispensable bioanalytical tools for both fundamental research and biomedical applications but necessitate long preliminary steps for the selection, production and purification of the antibody(ies) to be used. Here, we adopt a paradigm shift exploring the concept of creating a rapid and purification-free assay where the antibody is replaced by its coding DNA as a starting material, while exploiting a drop microfluidic format to dramatically decrease sample volume and accelerate throughput and sorting capability. The methodology consists in the co-encapsulation of a DNA coding for the variable domain of the heavy chain of heavy-chain only antibodies (VHH), a reconstituted cell-free expression medium, the target antigen and a capture scaffold where VHH:antigen accumulate to create a detectable signal, inside picoliter drop compartments. We first demonstrate successful synthesis of a functional hemagglutinin (HA)-tagged anti-GFP VHH, referred to as NanoGFP, at a high yield (15.3 ± 2.0 µg·mL-1) in bulk and in less than 3 h using PURExpress cell-free expression medium. We then use a microfluidic device to generate stable water-in-oil drops (30 pL) encapsulating NanoGFP-coding DNA, PURExpress medium, EGFP antigen and HA tag-specific magnetic nanoparticles prior to incubating at 37 °C the resulting emulsion under a magnetic field, inducing both in situ synthesis of NanoGFP and accumulation of NanoGFP:EGFP complexes on magnetically assembled particles. This allows us to assess, for the first time and in less than 3 hours, the binding of an antigen to a cell-free synthesized antibody, in a large number of picoliter drops down to a DNA concentration as low as 12 plasmids per drop. We also show that the drops of this immunoassay can be further sequentially analyzed at high throughput (500 Hz), thus offering capability for library screening, sorting and/or rare event detection. We finally demonstrate the versatility of this method by using DNA coding for different VHH (e.g., anti-mCherry protein), by characterizing VHH specificity in the presence of antigen mixtures, and by showing that antigens can be either inherently fluorescent or not. We thus anticipate that the ultraminiaturized format (pL), rapidity (3 h), programmability (DNA-encoded approach) and versatility of this novel immunoassay concept will constitute valuable assets for faster discovery, better understanding and/or expanded applications of antibodies.


Introduction
robust microfluidic approaches to implement antibody bioassays in highly miniaturized formats. 54 Drop microfluidics, which consists in handling, analyzing and sorting chemical and/or 55 biological components in nano to picoliter monodisperse drops in a high-throughput fashion, 56 has been identified as one of the most efficient methods to reach this objective. [5,6] For instance, 57 development of a microfluidic platform for compartmentalization, analysis and subsequent 58 sorting of individual cells [7] made possible to conduct studies on immune cells' antibody 59 secretome that was unexplorable by conventional flow cytometry. This breakthrough enabled 60 not only better understanding of antibody secretion dynamics [8] but also the characterization of 61 antibody binding properties, [8,9] both of which have facilitated the discovery of antibodies with 62 desired functionality. These achievements had in common to be based on single-cell incorporations [12,13] and new methods of extrinsic expression regulation, such as dynamic 75 photocontrol. [14] By simply using DNA coding for desired sequence, many protein types have 76 already been synthesized including enzymes, [15][16][17] membrane proteins [18][19][20][21] or large protein 77 assemblies. [22][23][24][25][26] Interestingly, cell-free expression systems can also be employed in 78 miniaturized format, successful examples including fluorescent proteins, [27][28][29] enzymes [30][31][32][33] and 79 transcriptional regulators. [34] In contrast, due to the large size and complex higher structure of 80 immunoglobulins (IgGs), their synthesis has so far only been achieved in bulk cell-free systems, 81 after substantial efforts to optimize both antibody-coding DNA sequence and the composition 82 of cell-free expression medium. [35][36][37] To achieve functional antibody synthesis in minimal 83 compartments without optimization steps, we could suggest instead to synthesize the variable 84 domains of the heavy chain of heavy-chain only antibodies (VHH) [38] engineered from naturally 85 occurring antibodies found in camelids. VHH are small-sized (~13 kDa), stable, single-unit and 86 easily-foldable proteins, originally used as tools for intracellular protein tracking [39] or reagents drops, the performance of this picoliter immunoassay in terms of minimum number of DNA 101 copies per drop, antigen detection limit and capture selectivity. antibody-antigen binding with minimal components, programmability and DNA instead of 107 purified antibody as starting material. The concept is based on water-in-oil drops of micrometric size 108 containing cell-free expression mix, fluorescent antigen, VHH-coding DNA and a capture scaffold 109 capable to attach the antibody (VHH) upon the synthesis. The resulting emulsion is incubated at 37°C 110 for 3 h during which the VHH is cell-free synthesized, is tethered onto the capture scaffold and binds 111 the antigen in case of sufficient antibody-antigen affinity. This binding alters the distribution of antigen 112 within the drop, from unbound antigen producing a homogeneous fluorescent signal (OFF) to antigen 113 accumulated on the capture scaffold, resulting in a local increase in fluorescence intensity (ON). associated it to an anti-GFP VHH, referred here to as NanoGFP [46] which was encoded in the 139 encapsulated DNA. For the cell-free VHH synthesis, we focused on reconstituted PURE 140 expression system, [47] a minimal set of recombinant components purified from E. coli, for their 141 well-known composition, fast protein synthesis and commercial availability. The template contains T7 promoter, lac operator (lacO) and ribosome binding site (RBS) located 147 upstream of the gene coding for NanoGFP (16.4 kDa), composed of VHH-coding gene separated by a 148 linker from HA epitope tag and 6xHistidine tag, followed by a T7 terminator. B) Capillary western blot 149 analysis of commercially available anti-GFP VHH (lane 1) and cell-free expressed NanoGFP (lane 2). 150 C) Dose-response curves of commercially available anti-GFP VHH (Standard) and cell-free expressed 151 NanoGFP. The curves were obtained by indirect ELISA, using EGFP antigen for VHH capture and an 152 anti-VHH peroxidase-conjugated IgG for detection. Dotted lines indicate the apparent dissociation 153 constant, for the standard: K D app = 0.49 ± 0.03 nM and for the NanoGFP: K D app = 0.26 ± 0.03 nM.

154
NanoGFP was expressed at [DNA] = 4 ngꞏµL -1 , 37°C, 3 h. 155 Prior to drop encapsulation, we characterized the amount of NanoGFP that could be synthesized 156 by PURExpress® in bulk and further tested its antigen recognition capability. To this end, we 157 designed a plasmid containing the necessary elements for transcription and translation in PURE 158 system (T7 promoter, ribosome binding site (RBS) and T7 terminator) around the gene coding 159 for VHH and hemagglutinin (HA) tag separated by a linker region ( Fig. 2A). For the VHH, we 160 used a sequence previously optimized for expression in E. coli. [46] The HA tag, designed for 161 tethering the synthesized protein onto the signal amplification scaffold, was positioned in the 162 C-terminus to minimize any possible effect on the antigen binding at the VHH paratope reported with the same VHH produced in bacteria. [39,49] All these results show that cell-free 181 synthesis in PURExpress® system produced functional NanoGFP at conventional yield and 182 expected binding affinity. We next exploited the DNA-encoding approach of our strategy and 183 applied it to explore the cell-free synthesis of other VHH sequences. We started with several  (Fig. 1), we had to devise a method to confine a small number of VHH-coding DNA 219 molecules in highly miniaturized containers and implement a method allowing to detect in situ 220 the functionality of the synthesized VHH. The strategy was to use a microfluidic device to 221 produce picoliter drops co-encapsulating VHH-coding DNA, the cell-free expression medium, 222 the EGFP antigen and a capture scaffold allowing to analyze the binding of the antigen to the 223 VHH synthesized in situ (Fig. 3). This binding was assessed using a capture nanoparticle made of a streptavidin-coated superparamagnetic bead previously functionalized with a biotinylated 225 IgG specific to the HA tag in the C-terminus of the target synthesized VHH (Fig. 3A, left). The 226 capture nanoparticles were assembled together with the VHH-coding DNA and the EGFP 227 antigen. This mix and PURExpress® expression medium were injected at 4 °C as the two 228 aqueous phases co-flowing in a microfluidic-device where drops were generated at a flow-229 focusing junction with a fluorinated oil supplemented with fluorinated surfactants to ensure 230 non-coalescence of the produced drops [50] (Fig. 3A, middle). The advantage of using capture  (Fig. 3A, right). 237 Microscopic observation on a large number of individual drops in parallel revealed that EGFP 238 signal was increasing at the position of the assembled capture particles, while vanishing in the 239 background (Fig. 3B, S5). The characteristic diffusion time for a NanoGFP:EGFP complex of 240 44.4 kDa was estimated to be 4.5 s (Text S1) using the drop diameter as a characteristic size, 241 allowing us to follow, with good temporal resolution, the capture of antigen along the course 242 of VHH expression (Fig. S6A) which is known to be sustained for about 3 h in bulk (Fig. S7).

243
The same experiment performed without DNA resulted in no signal evolution (Fig. S6B), the emulsion. We found that the signal from EGFP accumulated at the capture scaffold was 252 indeed correlated with the evolution of VHH level along the course of its expression (Fig. 3C).

253
Interestingly, significant signal concentration could already be observed in less than 30 min.

254
After 175 min of incubation at 37 °C, a concentration of 6.7 nM of synthesized VHH resulted 255 in a 5-fold increase of the EGFP signal per drop in average. Note that this assay involved a 256 number of DNA copies per drop λ = 300. Cell-free expression at the same DNA concentration 257 in bulk led to a yield of 4.9 nM of synthesized VHH after the same incubation time, emphasizing 258 that the encapsulation strategy did not hamper and may even favor the in situ protein synthesis.

259
All these results show that squeezing anti-GFP VHH-coding DNA, its target antigen and a 260 capture scaffold in microfluidic-generated picoliter drops allows one to quickly achieve VHH 261 synthesis and concomitantly assess its functional binding.

262
After characterizing a large ensemble of drops in parallel, we sought after sequential analysis 263 of individual drops at a high frequency offering, for instance, the capability to detect rare events 264 in real time. This method was implemented not only to characterize the performance of the 265 assay but also to devise a method that could be readily compatible with in situ sorting. To this 266 end, we injected the cell-free expression emulsion after 3 h incubation into another microfluidic 267 device integrating side oil channels to separate the drops and a laser-assisted in situ fluorescence 268 measurement (Fig. 4A, left). In this system, the whole width of the channel was laser- allowed us to discriminate between situations such as no detectable VHH expression (e.g., due 290 to absence of DNA or low-yield synthesis), expression of non-binding VHH (e.g., lack of 291 affinity), and successful synthesis of functional VHH (Fig. 4A, right). In the first two cases, which positive drops displayed particularly high signal compared to the background (Fig. S8). 309 We then assessed the minimum number of DNA copies per drop that could be used at this EGFP 310 concentration for the assay to remain applicable. We thus performed our analysis on different 311 emulsions produced with a varying VHH-coding DNA concentration (Fig. 4C, Figs. S9). We NanoGFP:EGFP binding (Fig. 5, top). Conversely, with the same amount of DNA (λ = 300) 334 but coding for anti-mCherry VHH (LaM-4) [51] , positive drops were detected with red channel 335 only (43%, Fig. 5, middle). The lower proportion of positive drops for red signal is likely 336 attributed to the lower yield of expression of LaM-4 (2.8 ± 0.6 µgꞏmL -1 in bulk, Fig. S3), which 337 is approximately 5 times less than that of NanoGFP, and the lower brightness of its mCherry 338 antigen. Without DNA (Fig. 5, bottom)   numbers of analyzed droplets are available in SI (Table S1). here mainly using a DNA coding for an anti-GFP VHH and the corresponding EGFP antigen.

367
Using DNA instead of a preliminarily purified antibody confers to the method a unique degree 368 of programmability that was demonstrated with the successful cell-free synthesis and 369 characterization of different mutants against the same antigen as well as VHH targeting other 370 antigens (mCherry). By simply adapting the DNA sequence, the method thus offers not only 371 the possibility to virtually implement any VHH but also to modify them in a highly tunable manner (e.g., tag addition, artificial amino acid incorporation, protein truncation/fusion). We 373 also showed that the antigen detection was operational with antigens that can be inherently 374 fluorescent (here, EGFP, mCherry) or not (secondary antibody labelling). For a fully DNA-375 encoded approach, we are also currently implementing the in situ cell-free expression of the 376 antigen itself (data not shown). The panel of antibody-antigen interactions that can be explored 377 with this method thus appears to be potentially extremely large. The microfluidic format of the 378 assay (drop generating device) only requires standard device fabrication and set-ups, thus being 379 implementable in a broad variety of environments, while offering the possibility to work with 380 minimal amounts of reagents at a high speed. We have shown in particular that the picoliter 381 drops containing the DNA-encoded immunoassay could be analyzed individually, in a parallel 382 or sequential manner, at a high frequency (500 Hz) and with amounts of DNA as low as 12