Glow-in-the-Dark Infectious Disease Diagnostics Using CRISPR-Cas9-Based Split Luciferase Complementation

Nucleic acid detection methods based on CRISPR and isothermal amplification techniques show great potential for point-of-care diagnostic applications. However, most current methods rely on fluorescent or lateral flow assay readout, requiring external excitation or postamplification reaction transfer. Here, we developed a bioluminescent nucleic acid sensor (LUNAS) platform in which target dsDNA is sequence-specifically detected by a pair of dCas9-based probes mediating split NanoLuc luciferase complementation. LUNAS is easily integrated with recombinase polymerase amplification (RPA), providing attomolar sensitivity in a rapid one-pot assay. A calibrator luciferase is included for a robust ratiometric readout, enabling real-time monitoring of the RPA reaction using a simple digital camera. We designed an RT-RPA-LUNAS assay that allows SARS-CoV-2 RNA detection without the need for cumbersome RNA isolation and demonstrated its diagnostic performance for COVID-19 patient nasopharyngeal swab samples. Detection of SARS-CoV-2 from samples with viral RNA loads of ∼200 cp/μL was achieved within ∼20 min, showing that RPA-LUNAS is attractive for point-of-care infectious disease testing.


Cloning and protein expression
The S. pyogenes dCas9 coding sequence was copied from the pET-dCas9-VP64-6xHis plasmid gifted by David Liu (Addgene plasmid #62935) by means of overhang PCR and was cloned via traditional restriction/ligation into a pET28a(+) plasmid (ordered from GenScript) coding for the C-terminal flexible linker and small BiT (SB) as well as large BiT (LB) and a C-terminal Strep-tag II (see protein coding sequence in Supporting Information). E.coli BL21 (DE3) was transformed with the resulting plasmid directly for dCas9-SB expression, whereas the [SB -Strep-tag II -stop codon] sequence in-between the flexible linker and LB coding portions was removed by restriction/ligation for dCas9-LB expression. All cloning was confirmed successful by Sanger sequencing (BaseClear). Both proteins were expressed in E.coli BL21 (DE3), grown in LB-Miller medium with kanamycin (50 µg/mL) to OD600 = 0.6 -0.8 at 37°C before induction by 0.2 M IPTG. Following overnight incubation at 18°C, cells were harvested by centrifugation (8600×g, 15 min., 4°C) and resuspended in 12.5 mL pre-chilled lysis buffer (500 mM NaCl, 1 mM TCEP, 50 mM Tris-Cl pH 8.0) per gram of cell pellet, supplemented with benzonase (25 U/mL) (Merck) and a cOmplete EDTA-free protease inhibitor cocktail tablet (Merck). Cells were lysed by 3 passes through a high pressure homogenizer (Avestin Emulsiflex C3), at 15'000 -20'000 Psi. The proteins were purified using Strep-Tactin XT (Iba) purification. Protein purity was confirmed by reducing SDS-PAGE (see SI) and concentrations were determined by measurement of absorbance at 280 nm on a NanoDrop spectrometer using extinction coefficients calculated from the protein sequence. Proteins in Strep-Tactin XT elution buffer (150 mM NaCl, 100 mM Tris-Cl pH 8.0, 1 mM EDTA, 50 mM D-biotin, 1 mM TCEP) were aliquoted and snap frozen in liquid nitrogen and stored at −70°C.
The mNG-NL calibrator luciferase was expressed and purified as described previously 1 , and the same procedure was followed for expression and purification of NanoLuc from a plasmid available in our lab (for Figure S7).

Assay design and synthetic target nucleic acids
Guide RNA pairs were designed using the CRISPOR tool developed by Haeussler et al. 2,3 and a custom add-on tool (LUNAS_CRISPOR_tool) taking into account the on-target activity as predicted by the scoring algorithms described by Moreno-Mateos et al. 4 and Doench et al. 5 , as well as the specificity scores based on Hsu et al. 6 and Doench et al. 5 . For the initial LUNAS assay used for sensor characterization, we designed 2 gRNAs that target bacteriophage T7 protospacers. These target sites were included in synthetic target DNA fragments with varying interspace distance separating the two protospacers. gRNA was generated from crRNA + tracrRNA (IDT) by combining both 1:1 in IDT nuclease free duplex buffer (30 mM HEPES, pH 7.5; 100 mM potassium acetate) to 4 μM final gRNA duplex concentration and heating at 95°C for 5 min., followed by gradually cooling to room temperature. gRNAs were aliquoted and stored at −30°C. Target DNA fragments were ordered as gBlocks (IDT) or PCR amplified from plasmids containing multiple such fragments, and then gel purified. For initial RPA-LUNAS experiments (Figure 3), RPA primers were designed for these synthetic targets, following TwistAmp (TwistDx) primer design guidelines.
For the SARS-CoV-2 assay, the genome of the original Wuhan-Hu-1 isolate (GenBank: NC_045512.2) was scanned for suitable target site pairs using the CRISPOR tool 2,3 and our LUNAS add-on. Candidate gRNAs were screened for specificity against genomes of related common cold human coronaviruses OC43, NL63, HKU1 and 229E as well as SARS-CoV and MERS-CoV (NCBI RefSeq accessions NC_006213.1; NC_005831.2; NC_006577.2; NC_002645.1; NC_019843.3; NC_004718.3 respectively). Three gRNA pairs predicted to be highly specific and having individual Moreno-Mateos 4 activity scores >30/100 were selected. Complementary RPA primers were designed using the PrimedRPA tool developed by Higgins et al. 7 , aiming for small amplicon size. Final assay designs were checked for the absence of SNPs with >1% prevalence known at the time based on GISAID 8 /Nextstrain 9 data in the UCSC SARS-CoV-2 genome browser 10 .
Corresponding SARS-CoV-2 cDNA fragments, PCR amplified from positive control plasmids provided by the FreeGenes project, were used for gRNA screening in LUNAS assays and subsequent primer screening in RPA-LUNAS assays. Synthetic ORF1a target RNA fragment was produced from the corresponding cDNA fragment by in vitro transcription using the HiScribe T7 High Yield RNA Synthesis Kit (NEB), according to manufacturer's instructions. The IVT reaction was treated with DNAse I (Thermo Fisher) to degrade template DNA according to the manufacturer's instructions. The IVT RNA was purified using a Monarch RNA cleanup spin column kit (NEB) and aliquoted for storage at -70°C. Concentrations of all nucleic acids were determined based on NanoDrop absorbance measurement at 260 nm (with 1 A260 optical density unit equal to 40 μg/mL RNA or 50 μg/mL dsDNA).
The sequences of all crRNAs, RPA primers and synthetic targets are listed in Table S1. The gRNA design tool that functions as an add-on to the CRISPOR tool 2,3 is available at https://github.com/harmveer/LUNAS_CRISPOR_tool.

LUNAS assays
For LUNAS experiments, dCas9-SB:gRNA_A and dCas9-LB:gRNA_B complexes were preassembled separately using fresh protein and gRNA stock aliquots on the day of use by incubating dCas9-SB/LB (10 -100x final sensor concentration) with the corresponding gRNA (3-fold excess) in LUNAS buffer (20 mM Tris-Cl pH 7.5, 150 mM KCl, 5 mM MgCl 2 , 5% (v/v) glycerol, 1 mM DTT, 1 mg/mL BSA) for 15 min at 37°C. LUNAS assays (without RPA) were performed at sensor RNP complex concentrations of 1 -10 nM in a total volume of 20 µL, in Nunc 384-well non-treated flat-bottom white microplate (Thermo Fisher). Input DNA was prepared by serial dilution in LUNAS buffer, of which 1 µL was added to a LUNAS reaction. For the interspace variation and DNA titration assays ( Figure 2), 1 µL NanoGlo substrate (Promega, N1110) was added at a final dilution of 2100-fold after 30 min incubation at room temperature. Luminescence spectra (398 nm -653 nm, step size 15 nm, bandwidth 25 nm) were recorded on a Tecan Spark 10M plate reader with an integration time of 100 -200 ms and data was collected using Tecan SparkControl V2.1 / V3.1. 'Blue' luminescence refers to the luminescence intensity at 458 ± 12.5 nm. LODs were calculated in Microsoft Excel by linear regression of sensor response over a limited concentration range, using the standard deviation of the y-intercept 11 .
For the kinetic measurements ( Figure S6, S10), NanoGlo (1000-fold final dilution) was directly included upon mixing sensor RNP complexes and input DNA in a total reaction volume of 20 µL, and luminescence spectra were recorded over time.

(RT-)RPA-LUNAS assays
For the 2-step (RT-)RPA-LUNAS assays, RPA reactions were prepared on ice using the TwistAmp Basic kit (TwistDx), first making a master mix comprising 505.26 nM of both primers, 14.74 nM magnesium acetate and 62.11% (v/v) TwistAmp rehydration buffer. For RT-RPA reactions, the master mix additionally included 2.11 U/μL SuperScript IV reverse transcriptase (Invitrogen), 0.105 U/μL RNase H (NEB) and 1.05 U/μL murine RNase inhibitor (NEB). This master mix was used to resuspend lyophilized RPA reaction components (TwistAmp Basic kit, TwistDx) at 47.5 µL per pellet. RPA reactions were performed at 40°C for 40 min in a total volume of 20 µL, combining 1 µL sample with 19 µL reaction mixture per replicate in a 96-well white PCR plate (VWR). For amplicon detection, a 1 µL sample was added to LUNAS reactions prepared and performed as described above.
For one-pot RPA-LUNAS assays, reactions were prepared on ice by first making a master mix containing 505.26 nM primers, 1053-fold diluted NanoGlo substrate, 14.74 nM magnesium acetate, 62.11% (v/v) TwistAmp rehydration buffer (TwistAmp Basic kit, TwistDx), and 10.53% (v/v) LUNAS mix (100 nM dCas9-SB:gRNA_A and 10 nM dCas9-LB:gRNA_B in LUNAS buffer). For ratiometric assays, 120 pM mNG-NL calibrator luciferase was included in the LUNAS mix. For RT-RPA-LUNAS reactions, the master mix additionally included 2.11 U/μL SuperScript IV reverse transcriptase (Invitrogen), 0.105 U/μL RNase H (NEB) and 1.05 U/μL murine RNase inhibitor (NEB). This master mix was used to resuspend lyophilized RPA reaction components (TwistAmp Basic kit, TwistDx) at 47.5 µL per pellet. Reactions were performed in a total volume of 20 µL, combining 1 µL sample with 19 µL reaction mixture in a Nunc 384-well non-treated flat-bottom white microplate (Thermo Fisher). Luminescence spectra (413 -563 nm) were recorded over time at 40°C on a Tecan Spark 10M plate reader with an integration time of 100 ms. The blue/green emission ratio was calculated by dividing luminescence intensity at 458 ± 12.5 nm by the intensity at 518 ± 12.5 nm.

Clinical validation
For testing clinical eSwab (Copan, Italy) samples without RNA isolation, viral lysis and nuclease inactivation was performed by adding an inactivation buffer (200 mM TCEP, 2 mM EDTA, 2 U/µL murine RNase inhibitor, 20 mM Tris-HCl, pH 8.0) to the sample in 1:1 ratio, followed by incubation at 95°C for 5 min. The compatibility of RT-RPA-LUNAS with this sample pretreatment method, based on Arizti-Sanz et al. 12 and Qian et al. 13 , was tested using mock eSwab, as well as VTM (HiMedia HiViral medium) and saliva (obtained from a healthy donor) samples, which were prepared by spiking in IVT ORF1a target RNA. Target RNA was added before heating in order to simulate release of RNA from lysed virus before complete denaturation of RNases.
Clinical RNA isolate samples were extracted from nasopharyngeal swabs collected in either 1 mL eSwab or 2 mL PurePrep TL+ buffer (MolGen). For eSwabs, 150 µL sample was combined with 150 µL MP96 external lysis buffer (Roche) for viral lysis. RNA was extracted from 300 µL of this mixture or from 300 µL of samples collected in PurePrep TL+ directly using the STARMag 96x4 Viral DNA/RNA Universal kit (SeeGene, South-Korea) on a Hamilton Starlet, and was eluted in 100 µL.
The one-pot ratiometric RT-RPA-LUNAS assay was prepared as described above, with a higher calibrator luciferase (24 pM final concentration) and NanoGlo substrate concentration (400-fold final dilution). One or two NTC reactions (water input) were included per assay run of 18 -56 duplicate reactions. Samples (1 µL) were added to reaction mixtures (19 µL) in a 96-well white PCR plate (VWR) on ice, which was then briefly centrifuged in a lettuce spinner 14 before placing it in a heating block set at 40°C (monitored by external thermometer) within a black EPP box to exclude ambient light. Luminescence was recorded using a Sony DSC-RX100 III digital camera fitted through a hole in the lid of the box, with 30s exposure time, f/1.8 and ISO-6400. The camera was controlled by the Sony Imaging Edge Mobile app on an Android device with an auto clicker app (Click Assistant -Auto Clicker by Y.C. Studio) for continuous shooting over 1 hour. Resulting RAW images were converted to 16 bit TIFF files using Sony Imaging Edge Desktop, and mean blue (B) and green (G) intensities per well were extracted from split RGB channels in ImageJ (v1.53q). Reactions were regarded as positive for SARS-CoV-2 if the blue/green ratio (BGR(t i )) exceeded a threshold value BGR T (t i ) = BGR smoothNTC (t i ) + 6SD All,1-3min for t i to t i+2 , with BGR smoothNTC (t i ) the moving average blue/green ratio of the NTC over (t i-5 , t i+5 ), and SD All,1-3min the standard deviation of the blue/green ratio of all reactions between t = 1 min and t = 3 min. The reported LUNAS threshold time t T equals the first t i satisfying the condition BGR(t i ) > BGR T (t i ) for (t i , t i+2 ).

RT-qPCR and ddPCR testing of clinical samples
RNA was extracted as described above, followed by RT-PCR using the Allplex SARS-CoV-2 assay (SeeGene, South-Korea) on a CFX96 thermocycler (Biorad), simultaneously detecting four different target genes: E, N, RdRp and S (the latter two are combined in one fluorescent signal). Data was analyzed with the SARS-CoV-2 Viewer software (SeeGene). Some samples were initially tested using the BioFire Respiratory Panel 2.1+ (bioMérieux, France), enabling qualitative detection of multiple pathogens.
To quantify absolute ORF1a_1 RT-RPA-LUNAS target concentrations and correlate these with RT-qPCR C t values, droplet-digital PCR was performed for 16 extracted RNA samples using the 1-Step RT-ddPCR Advanced Kit for Probes (BioRad) and the CFX96 thermocycler (BioRad) and QX200 ddPCR system (BioRad). Three different reactions were performed per sample, targeting the E-and the N-gene as well as the ORF1a_1 region targeted by the RT-RPA-LUNAS assay. Samples were diluted 1 -1000fold to avoid overloading and 5 µL of diluted sample was added per 22 µL PCR reaction. Data was analysed using BioRad QX One (v1.2) and linear regression of the correlation between resulting concentrations and RT-qPCR C t values was performed in Origin 2020 (OriginLab). All results are listed in Table S2.

Thermodynamic model
To shed light on the dependence of the LUNAS response on the various thermodynamic parameters involved, we developed a model of the system ( Figure S1). In this model, the total concentration of the sensor RNP complexes was divided over an active (i.e. DNA-binding competent) and an inactive fraction in a 25:75 ratio, in accordance with previous reports 15, 16 . The DNA-binding incompetent sensor complexes (denoted 'Ld' and 'Sd' for the LB and SB complexes respectively) were still included in the model, to account for background signal derived from the split NanoLuc parts of these complexes, but conversion to active complexes (denoted 'La' and 'Sa') and vice versa was excluded. For the interaction of dCas9-SB:gRNA_A and dCas9-LB:gRNA_B with the target DNA (denoted 'T') the same single dissociation equilibrium constant was defined ('K D C'), assuming differences in affinity depending on the exact gRNA/protospacer sequence to be small for functional LUNAS RNP complex pairs. Moreover, dCas9-SB:gRNA_A and dCas9-LB:gRNA_B are considered to bind the target DNA in a non-cooperative fashion, hence dCas9-SB:gRNA_A binds to free target DNA with the same affinity as to target DNA already bound by dCas9-LB:gRNA_B (denoted 'LaT'). Upon formation of the ternary 'LaSaT' complex, the high local concentration (the effective molarity (EM)) of the split NanoLuc fragments promotes complementation, transitioning to the luminescent 'LaSaTa' complex. As non-templated split NanoLuc complementation could also result in luminescence, the total luminescent signal is modelled as the sum of the concentrations of LaSaTa, LaSa, LdSd, LdSa and LaSd, multiplied by a constant. From this, a model was generated in the framework, which is further detailed below for completeness: The equilibrium concentrations of the dependent species can be determined based on the concentrations of the independent species and the corresponding equilibrium constant, using the following relations: This model was fitted to the combined data presented in Figure 2C using 'KDNanoBiT' = 2.5E-6 as known parameter 18 . For dCas9:gRNA complex binding to target DNA, an upper limit in K D of ~0.5 nM was previously reported based on EMSA experiments 19 , and here 0.1 nM was taken as an initial value for 'KDdCas9'. The effective molarity of the split NanoLuc fragments (EM) depends on the length and flexibility of the linkers connecting them to the dCas9 proteins, as well as the distance that has to be bridged by the linkers for luciferase complementation. An initial value for the EM was estimated based on the wormlike chain model 20,21 . Considering the 21 residue linkers are made up mostly of GGS repeats, a persistence length of 3.7 Å was assumed 21 . Based on the structure of DNA bound dCas9:gRNA (PDB: 5F9R 22 ) and that of NanoLuc (PDB: 5IBO), the distance to be bridged between anchor points (dCas9 C-terminus and SB/LB N-terminus) for the 50 bp interspace target used is roughly 49 Å. From this, we estimated an EM of ~ 31 µM 20 . For the factor converting luminescent entity concentrations to luminescence signal intensity, the initial value was set as 'constant' = 1E+15. The following parameter estimates were obtained (fitted lines shown in Figure 2C): EM = 1.048e-05 KDdCas9 = 1.767e-11 constant = 1.939e+15 Root Mean Squared Error (10 nM dCas9-SB RNP + 1 nM dCas9-LB RNP) = 2.01e+04 Root Mean Squared Error (1 nM dCas9-SB RNP + 1 nM dCas9-LB RNP) = 2.02e+04 R 2 (10 nM dCas9-SB RNP + 1 nM dCas9-LB RNP)= 0,982 R 2 (1 nM dCas9-SB RNP + 1 nM dCas9-LB RNP)= 0,955 To gauge the dependence of the signal response on tunable sensor parameters, we performed simulations in which the concentration ratio of sensor RNP complexes was varied, and we simulated the effect of different split NanoLuc binding affinities ( Figure S2). For this, parameter estimates described above were used.

Ethics statement
The use of anonymised clinical samples for this study was evaluated and approved by the local medical ethics review committee of the Rijnstate Hospital (reference number: KCHL 2021-1950), and samples were acquired in accordance with the Declaration of Helsinki. The samples used here were obtained as part of standard clinical COVID-19 testing and patients did not object to the use of remnant sample material for quality purposes and research.

Safety statement
No unexpected or unusually high safety hazards were encountered.     : NanoLuc titration curve. A 10-fold serial dilution series of NanoLuc in LUNAS RNP buffer was combined with NanoGlo substrate (2000-fold final dilution) and blue luminescence intensity was measured. The inset zooms in on the portion of the main graph indicated in the dashed box. This data shows that the minimal active NanoLuc concentration that can be detected under LUNAS conditions in 20 µL is in the 10 -100 fM range. Since complemented split-NanoLuc (NanoBiT) has a relative luciferase activity of ~ 37% compared to that of full-length NanoLuc, the minimal concentration of complemented split-NanoLuc that can be detected under these conditions is presumably on the order of ~100 fM 18 . Individual replicates (n = 3) are shown as circles, the line represents a linear fit to the full data range (Pearson's r = 0.99969; R 2 (COD) = 0.99938).  Table S1). With these dCas9 RNPs (1 nM of both SB and LB), LUNAS assays were performed on matching target cDNA fragments (250 pM), as well as on the non-matching fragments (250 pM) used as non-target controls. Additionally, for the N-gene LUNAS assay, MERS and SARS N-gene cDNA fragments (250 pM) were used as non-target controls featuring strong sequence-homology with the target. Clearly, the ORF1a_1 and N-gene gRNA sets show the best LUNAS performance, with a 40-and 24-fold increase in blue luminescence over background level respectively, while the ORF1a_2 set only shows a moderate 4.5-fold increase and low absolute intensity. The low response of the SARS-CoV-2 N-gene LUNAS for MERS and SARS N-gene cDNA fragments compared to the target SARS-CoV-2 N-gene cDNA fragment demonstrates the assay specificity. Bars represent means of technical replicates (n = 3), which are indicated as black dots. Error bars show SD. B Screening RPA primer sets for best performance in an RPA-LUNAS assay. Two forward and reverse primers were designed for combination with the two best performing SARS-CoV-2 LUNAS assays in (A), one partly complementary to the PAM-distal protospacer sequence ('_S' primers), the other only binding to the sequence 3' from the protospacer ('_L' primers). All 4 combinations of primers per target region were tested. For both the N and ORF1a target region, the assays using the combination of only '_L' primers showed the highest signal. Clearly, the ORF1a assay performed best, showing the quickest onset of signal increase, and reaching the highest absolute intensity. The assays were performed at 42°C, using 200 cp of input cDNA fragment. Individual replicate traces are shown (n = 2).  , and the bottom panel shows the blue/green ratio over time. While an increase in target concentration from 25 pM to 250 pM can be seen to result in a rapid increase in blue signal, the increase from 250 pM to 2.5 nM does not result in substantial change in signal compared to that of the control left at 250 pM. This result confirms the extremely slow dissociation rate of dCas9 RNPs, as quick dissociation and redistribution of the dCas9 RNPs would result in a lower signal for 2.5 nM compared to 250 pM target (see Figure 2C). LUNAS assays were prepared as described for 1-pot RPA-LUNAS assays, excluding primers to preclude RPA, and were performed at 40°C to resemble RPA-LUNAS conditions. Individual replicate traces are shown (n = 3).  Figure 4D, 200 cp input) as measured over time (extended data version of Figure 4C). The right waterfall graph shows the backside of the left graph. The rapid increase in blue signal can be observed, followed by an overall gradually decreasing luminescence due to substrate depletion, which can be observed for the green signal already from the start. However, the blue/green ratio stays relatively constant over time after the initial rise in blue LUNAS signal.    Traces are grouped in panels per assay run, with data divided over multiple panels per run for clarity. In the legend, the sample ID is followed by the corresponding E-gene RT-qPCR C t value in brackets. Traces are grouped in panels per assay run, with data for some runs divided over multiple panels for clarity. In the legend, the sample ID is followed by the corresponding E-gene RT-qPCR C t value in brackets. Tables   Table S3: Comparison of RT-RPA-LUNAS SARS-CoV-2 assay performance with that of other recent CRISPR diagnostic methods applied to SARS-CoV-2 detection.

Method
Reference Steps

Protein coding sequences and translations dCas9-SB / dCas9-LB
The following combined sequence codes for dCas9-SB (default) and dCas9-LB (after SpeI restriction digest and subsequent self-ligation of the large fragment). Colour codes: dCas9 highlighted in yellow; SB in cyan; LB in green; SpeI restriction site in magenta, Strep-tag II in red font.