Introduction

Many laboratories and companies around the world have deployed their resources to generate fast and cheap diagnostic tools [1,2,3] since the beginning of the coronavirus pandemic. In this regard, cell-free protein synthesis (CFPS)-based tools emerged as attractive technologies, because CFPS is a sustainable platform [4] that is relatively simple, inexpensive and straightforward to engineer [5,6,7,8]. Consequently, multiple proof-of-concept and market-oriented studies have been conducted to test the feasibility of deploying a test kit using CFPS [9], rendering such efforts quite timely. This is mainly because the market for global point-of-care diagnostics is expected to reach 72 billion dollars by 2027 [10] and grow even further. CFPS attracted particular attention also because the freeze-drying of cell-free components enables biomanufacturing since activation only requires “a drop of sample” [11,12,13]. In parallel, other efforts were diverted to adapt CFPS platforms for efficient and virus-specific detection of RNA, in particular with CRISPR/Cas nucleases [14,15,16,17]. Likewise, multiple rapid and enzymatic RNA amplification methods were developed with a colorimetric output [18,19,20,21]. While many of these assays can detect RNA down to a few copies, they often suffer from relatively high costs per run, rely on commercial reagents, need trained workers or require instruments for readout, even if the methods are deployable in the field [22, 23]. Most notably, temperature cycling or incubation above room temperature hampers the capacity of cell-free biosensors for field applications, especially in develo** countries with limiting access to resources and instrumentation. Despite recent promising advances to reduce the costs and render cell-free systems widely available [24, 25], the field is still in need of genuinely end-user friendly diagnostics at room temperature, similar to antigen-based lateral flow kits; rather than instrument-dependent readouts above room temperature.

If a massive scale-up of rapid diagnostic kits is necessary for population-wide screening to cope with emerging pandemics, relying on commercial sources of reagents can be unsustainable, thereby limiting diagnostic capacity. To this end, the unmet need for accessible, sustainable and inexpensive diagnostic tools that are easy-to-deploy and are affordable still remains a challenge. With this strategic goal in mind, we explored the experimental conditions and report the capacity of viral RNA detection at ambient temperature by repurposing an E. coli cell-free transcription/translation system coupled to an isothermal nucleic acid amplification tool, such as Reverse Transcriptase-Recombinase Polymerase Amplification (RT-RPA) (Scheme 1).

Scheme 1.
scheme 1

Conceptual summary of viral RNA detection by cell-free assays

Full size image

We adapted toehold-switch-mediated riboregulatory elements for the activation of gene expression with a colorimetric output, whereby the isothermal RNA amplification step was solely aided by body heat. Using minimal equipment and cell-free reactions operating at room temperature, high-attomolar (ca. 110 aM) concentrations of viral RNA were detected from synthetic samples. The colorimetric output was generated by α-complementation of the β-galactosidase ω-fragment using an X-gal analogue as a color-changing substrate with enzyme activity. In principle, this platform can be coupled to magnetic-bead-driven RNA isolation, where a proof-of-principle was demonstrated from saliva. We estimate the total cost of the assay to be ca. 0.70 euro/test or less, even without industrial scale-up steps. Our assay can serve as a starting point for deploying an inexpensive and sensitive nucleic acid diagnostics that operates at ambient temperature, with further optimizations performed as we recommended.

Results and discussion

As a proof-of-concept system, we chose to work with RNA sequences from SARS-CoV-2. To this end, we in silico designed and functionally verified 11 different toehold-switch-mediated riboregulatory constructs that are complementary to the 5’ and 3’ untranslated regions (5’ UTR and 3’ UTR) of the SARS-CoV-2 genome (Figure S1). The 5’ UTR was specifically chosen because during the coronavirus replication cycle, discontinuous RNA synthesis generates higher copy numbers of UTR sections than the rest of the genome (Figure S1A) [26,27,28], potentially enabling sensitive diagnostics. For initial characterization of the riboregulatory elements and cell-free extracts, we used superfolder GFP (sfGFP), controlled by T7 promoter (Fig. 1 and Figure S2). The activation of cell-free gene expression with different toehold-switches was triggered by a complementary DNA oligonucleotide for initial screening. All riboregulatory elements activated the cell-free gene expression only in the presence of complementary oligonucleotides, but showed different performances (Fig. 2 and S3A). The most promising construct, TH001, gave ca. 15-fold increase in gene expression, thus it was chosen for subsequent experiments. In E. coli cell-free systems, the innate nucleic acid detection limit, that is without any pre-amplification step, was found to be ca. 1 nM (Fig. 3 and S3B). Such levels of toehold-switch activation are on par with the detection limit of PURExpress-based cell-free systems [7, 25]. Thus, we concluded that a pre-amplification step was necessary to detect clinically-relevant concentrations of viral RNA.

Fig. 1
figure 1

Toehold-switch mediated activation of cell-free gene expression

Full size image
Fig. 2
figure 2

Activation of gene expression by target viral sequences. A. Secondary structures of regulatory elements. B. Green Fluorescent Protein (GFP) production by cell-free reaction

Full size image
Fig. 3
figure 3

Isothermal amplification of viral RNA at various conditions. 'NT' stands for No template cell-free reaction assembly without any template DNA for amplification

Full size image

Point-of-care (POC) diagnostic tools may need exclusively room temperature operations, especially if instrumentation is scarce. To this end, cell-free reaction conditions and extract preparation protocols were tested for room temperature work-up (Figure S4). That is, E. coli extracts were prepared following post-log-phase growth at 23 °C. E. coli extracts were found to retain 50% activity compared to regular 37 °C growth (Figure S4A). To minimize the dependency to expensive instruments, the functionality of cell-free reactions was tested and verified without freeze-drying but only after drying (Figure S4B). This way, we potentially eliminated the need for expensive freeze-drying equipment for preparation of lyophilized cell-free reactions. POC diagnostics can also be adapted to use one-step viral inactivation from bodily fluids, followed by the cell-free reactions. To this end, the cell-free reactions were tested for compatibility with human saliva. As a reaction additive, saliva did not inhibit the reactions, as long as dilution was used between 1/20th and 1/100th of the reaction volume (Figure S4C).

Following these optimization efforts of the cell-free extract, we focused on develo** cell-free-compatible isothermal RNA amplification strategies. To this end, nucleic acid sequence-based amplification (NASBA) [29], reverse transcriptase-recombinase polymerase amplification (RT-RPA) [30, 31] and reverse transcriptase rolling circle amplification (RT-RCA) [32] reactions were tested (see supplementary methods). In these assays, we used T7 RNA polymerase promoter containing forward primers. The resulting double-stranded DNA (dsDNA) can act as a single-stranded DNA (ssDNA) trigger to the toehold switches due to single-strand DNA-binding protein (SSB) and serve as transcriptional templates to generate single-stranded RNA (ssRNA) to amplify the signals (Scheme 1).

Of all pre-amplification methods, RT-RPA proved to be the most efficient technique, both at recommended temperatures and at 30 °C (Fig. 3 and S5). Nevertheless, we still decided to put further effort in optimization of NASBA reactions, which can be assembled in-house, used in high-throughput sequencing-based diagnostics [33], and accessible without a patent-wall. Our goal was to minimize the dependency of diagnostic tools to commercial components and kits – proven to be detrimentally scarce at the early days of COVID-19 pandemic. This way, the entire work-up would be as sustainable as possible to deploy the kits in develo** countries. We assembled NASBA reactions using past reports as a guide [34, 35] and additionally tested different parameters. Our major goal was two-fold: (1) to couple NASBA to cell-free reactions; (2) to minimize temperature cycling and preferably operate solely at room temperature. Initially, we screened 10 different reaction additives inside a unique buffer composition that is suitable for all enzymes in the cocktail. (Figure S5A, supplementary methods). The additives that elicited the most positive impact were 10% (v/v) DMSO and 1 M betaine (Figure S5B).

Previous reports showed that “unoccupied” T7 bacteriophage RNA polymerase (T7 RNAP) can trigger transcription of random, non-specific RNA duplexes [36]. In order to alleviate this issue, random DNA oligonucleotide duplex was added to the mixture to minimize non-specific transcription by T7 RNAP (Figure S5B) [37]. The NASBA reaction conditions were further tested with decreasing enzyme concentrations at room temperature (Figure S5C). None of the methods allowed for efficient RNA amplification at room temperature, including RT-RPA (Figure S5C and S5D). However, conditions for optimal enzyme activity in pre-amplification steps were conducive to a putative single-pot lysis of human cells and extraction of viral particles. Coupling with lysis buffer, the presence of 0.5% (v/v) Triton X-100 in the final NASBA reaction mixture improved the amplification efficiency (Figure S6A). Nevertheless, the process of cell lysis from human saliva did not overlap with RNA amplification in NASBA reaction, as a “ready-to-use” reagent (Figure S6B).

To minimize the dependency of temperature cycling, we further tested NASBA without pre-heating at 65 °C for primer annealing. A pre-heating step proved to be non-detrimental for amplification, albeit at a cost of decreased efficiency (Figure S6C). Given the known intrinsic toxicity of low concentrations of glycerol for E. coli cell-free expression system, and the sensitivity of T7 RNAP to reducing storage conditions, we conclude that a system independent of externally provided T7 RNAP is the best choice for reproducible results. In the end, the lysis buffer composition was also more compatible with RT-RPA than NASBA (Figure S6C) for single-pot purification and amplification of viral RNA from saliva samples. Moreover, since accessibility to instruments in field conditions is limited, we confirmed that RT-RPA works well, when 8-strip tubes were hand-held and warmed by body heat as good as reactions at 37 °C (Scheme 1 and Fig. 3). We find this approach significant, since healthcare workers or the patients themselves can perform the test without any instrumentation apart from micro pipettors or disposable Pasteur pipettes.

Having found the most conducive pre-amplification mode as RT-RPA, we set out to modify our reporter system from fluorometric to colorimetric output. To this end, maltose-binding protein (MBP)-tagged β-galactosidase ω-fragment (LacZω) was expressed in E. coli NEB5α, purified with Amylose Resin and separated from MBP by TEV protease digestion (Figure S7A). For α-complementation, LacZα subunit gene expression was placed under the control of endogenous E. coli promoters (Figure S7B and Table S1). To test the complemented LacZ activity, 2 mM (final or 12 µg/µL) Chlorophenol Red-β-D-galactopyranoside (CPRG) was used as a substrate to give an expected color change from yellow to reddish-purple. We note that the precise color change is heavily dependent on the solid support [38].

E. coli BL21-derivative strains are regularly employed for cell-free extract preparation and contain an endogenous copy of LacZ, which was not suitable for α-complementation (Figure S8A). To this end, we set out to prepare cell-free extracts from E. coli strains of JM109 and DH10β, which have genotypes with a LacZΔ15 mutation. The cell-free gene expression from the crude extracts of JM109 and DH10β was verified and, in the presence of toehold-switch triggering complementary DNA oligonucleotide, the colorimetric assays showed the color change from yellow/orange to orange/red within 1 h, demonstrating the functionality of our reporter system (Figure S8C).

Having shown the cell-free reactions generated colorimetric output at room temperature, we next set out to couple RT-RPA amplification to cell-free reactions. First, RT-RPA was run for 30 min at 37 °C and then added to the cell-free reaction at a 1:20 dilution. Within 8 h after rehydration, the reddish color was developed down to ca. 110 attomolar (aM) final concentration of RNA, which corresponded to ca. 667 copies of RNA per reaction. This value was verified by qRT-PCR having a cycle threshold (Ct) value higher than 28.79 ± 0.36 (Fig. 4 and S9). In other words, as long as the RT-RPA amplified RNA generated a Ct value above 10–12, we were able to detect a color change (Figure S10A).

Fig. 4
figure 4

Colorimetric RNA detection on paper-disks

Full size image

Subsequently, we wanted to render all steps compatible at ambient temperature, in a proof-of-concept experiment: (1) cell lysis + viral inactivation, (2) RNA capture in saliva, (3) RT-RPA and (4) cell-free colorimetric assay. We chose saliva as a biofluid for the test, since saliva can be self-tested by the users in a non-invasive manner, just as in rapid antibody-based kits. Synthetic RNAs were spiked-in to human saliva and isolated by magnetic bead separation. Nevertheless, the isolated RNA could not be efficiently amplified with RT-RPA, suggesting an incompatibility in buffer compositions, as saliva on RT-RPA was not completely inhibitory (Figure S6C). In the end, we examined the clinical relevance of the proposed magnetic-bead isolation. That is, we tested whether we can reveal the presence of viral RNA in patient samples. First, positive patient samples were identified by 5’ UTR-specific qRT-PCR (1 hit out of 7 samples) and later samples were amplified by RT-RPA (Figure S10B).

The amplified RNA was isolated from saliva spiked-in samples, verifying our approach in-principle, but requiring further optimization efforts to combine all the steps together in one-pot. The clinical translation of our findings, in part, can be achieved if the subsequent research maintains the goal to reduce the duration of the cell-free reactions as well as the target reaction costs at low levels after scale-up. Since our data revealed that the room temperature performance of the CFPS was the bottleneck, we propose following experimental approaches to optimize the E. coli cell-free reactions for room temperature operations, either individually or in combination: (1) a genome-wide, combinatorial mutation screen, (2) Adaptive Laboratory Evolution (ALE) [39], (3) Multiplex Automated Genome Engineering (MAGE) [40, 41], or (4) other accelerated evolution strategies [42] to engineer E. coli strains with competitive growth rates at room temperature.

Conclusions

Here, we reported colorimetric detection of clinically relevant concentrations of RNA with a low-cost cell-free assay, with all operating conditions at ambient temperature. In theory, this assay does not require any instrumentation apart from the disposable pipettors (and/or a magnet). The current version may suffer from suboptimal reaction conditions, particularly with the composition of the E. coli extract. Such focused efforts to improve the assay efficiency are likely to decrease the detection limit to low-attomolar concentrations and reduce the incubation times down to minutes rather than hours. While our assay cannot compete with antigen-detecting lateral flow kits in terms of assay times for the moment, our assay clearly falls within the formal definition of point-of-care biosensors [43]. Our assay is compatible with ‘overnight’ diagnosis, which is faster than the average turnaround times of current state-of-the-art RT-qPCR diagnostics at peak demand [44]. The type of toehold-mediated switches that we exploited, are incapable of distinguishing mismatches less than 4 nucleotides. However, other engineered systems show promise to identify single mismatches in a given RNA sequence [45, 46].

We estimate the overall cost of the single test to be as low as ~ 0.26 euro (see Supplementary Text and Table S1), including the labor costs to prepare and perform the test. Given that future efforts can be diverted to the optimization of this assay as an end-user-friendly diagnostic, even further cost reductions can be anticipated at industrial scale. At the moment, another limiting factor appears to be the commercial dependence to RT-RPA reactions. Nevertheless, low microliter volume RT-RPA and future scale-up efforts for E. coli extracts can also bring the assay costs substantially lower than our estimates, up to 50% reduction [13].

Finally, we excitedly point out that the colorimetric detection can be coupled to cell phone applications [14] or wearable devices [20], bringing cell-free synthetic biology technologies in our everyday lives. The clinical and translational potential of paper-based biosensors has once more highlighted with our work, given the apparent limitations were addressed and extensive optimization studies were performed, potentially opening new avenues for sustainable diagnostics.

Methods

E. coli cell-free extract preparation

Homemade E. coli cell-free extracts that are used for in-solution reactions were prepared from Rosetta 2(DE3) Singles strain (Novagen), using published protocols as a guide [47]. For on-paper reactions, the E. coli strains JM109 and DH10β were used. This crude extract preparation did not include the dialysis step. For details, see supplementary materials.

Isothermal nucleic acid amplification

Reverse Transcriptase Recombinase Polymerase Amplification (RT-RPA) reactions were by from TwistAmp® Basic kit, and assembled according to manufacturer’s instructions (TwistDx, UK). The reactions were assembled in 10 µL volumes and contained 1 µL of template (either clinical samples or synthetic RNA) and 0.2 µL (40 U) of RevertAid Reverse Transcriptase (ThermoScientific). The final RT-RPA volume was 0.525 µL in a 10.5 µL cell-free reaction (1:20 dilution). For details on the NASBA and RT-RCA, see supplementary materials. For hand-warm protocol, the estimated ambient temperature was 32 ± 2 °C.

In vitro transcription/translation

All in vitro transcription/translation reactions were performed in a final volume of 10.5 µL. For in-solution reactions, the sfGFP fluorescence was measured by Rotor–Gene Q qPCR machine (Qiagen) or by a multi-well plate reader (Varioskan, ThermoFisher). For on-paper reactions, the amino acid solution mix and energy solution compositions were taken from the literature [47]. The supplementary solution additionally contained 10 mM maltose, glutamate salts (Mg2+ and K+) and 2% (w/v) PEG8000 as a molecular crowding agent. Purified LacZω (50 ng/µL) and LacZ substrate CPRG (0.12 µg/µL). Template plasmid DNA was at a final concentration of 30 nM. Whenever supplemented, the T7 RNA polymerase was provided with a final glycerol concentration of 0.1% (v/v).

General procedure for colorimetric assay

Prior to on-paper reaction, 1 µg of MBP-LacZω was digested by 1 µL TEV protease (New England Biolabs) in a 50 µL at 30 °C for 5 h. Then, 5% (w/v) BSA-blocked, air-dried (16 h) paper-disks were put in in clear bottom 96–well plate (Costar) and the E. coli cell-free reactions were added on top. Both synthetic RNA (pre-amplified) and clinical samples (purified and pre-amplified) were provided at a 1:20 final dilution. The reactions were let air-dry and monitored for color change at uncontrolled room temperature (20 ± 2 °C) up to 10 h.

Synthetic and clinical RNA sample preparation

The sequence of 5’ UTR( +) region of the wild-type SARS-CoV-2 was taken from NCBI (265 bp), obtained as a dsDNA gene fragment (GenScript) and cloned into an expression cassette under the control of consensus T7 promoter. In vitro transcription was performed by T7 RNA polymerase for 12 h at 37 °C from PCR-amplified template. Synthesized RNAs were initially cleaned up by TRIzol, ethanol precipitated and subsequently purified by NucleoSpin RNA Clean-up kit (Macherey–Nagel). The RNA from infected patients was derived from anonymized saliva samples collected with ethical committee approval of the Azienda Provinciale per i Servizi Sanitari (A.P.S.S.) of the Autonomous Province of Trento (P.A.T.). Prior to the proof-of-concept RNA spike-in tests, saliva was diluted up to 5 or 10 times with PBS to reduce its viscosity. Magnetic-bead-mediated RNA isolation was by biotinylated complementary DNA oligonucleotides to capture, and streptavidin-coated magnetic beads to isolate with a magnetic rack (Invitrogen). The experiments were performed using manufacturer’s instructions as a general guide, i.e., high-salt buffer to bind and low-salt buffer to wash; with significant modifications as specified in supplementary methods.

Real-Time Quantitative Reverse Transcriptase PCR (qRT-PCR)

Prior to qRT-PCR, cDNAs were synthesized by reverse transcription with iScript™ cDNA synthesis kit (Bio-Rad). For clinical samples, the template was 14 µL (max. amount in 20 µL). qRT-PCR was performed by SsoAdvanced™ SYBR® Supermix (BioRad). qPCR was run at Bio-Rad CFX96 Real-Time machine and acquisition was at FAM channel. qRT-PCR primer pairs were designed by online software Primer3, optimized for Tm = 57–60 °C, to generate an amplicon size of ca. 100 bp. Standard curve was generated by serial dilutions with 1:10 and primer efficiency was calculated as the slope of cycle threshold (Ct) vs dilution factor.

Genetic constructs and protein purification

The toehold switches were designed using previously published principles (details in the supplementary methods) [48]. All clonings were performed by homemade Gibson Assembly mix [49]. All expression plasmids were with pSB1A3 backbone from iGEM Parts Registry. Double-stranded DNAs were obtained by PCR, gBlocks or in-house assembly of DNA primer-stitched templates. LacZω was cloned into a modified pMAL-c4X backbone containing N-terminus Maltose Binding Protein (MBP) with TEV protease recognition site flanked by (GS)2 linker sequence. E. coli NEB5α strain transformed with MBP-TEV-LacZω expressing plasmid was grown in Terrific Broth at 37 °C. The fusion protein was overexpressed with Autoinduction Medium [50] with overall growth of 24 h. The MBP-fusion protein was purified by Amylose Resin (New England Biolabs), eluted with 10 mM maltose.