General reagents
Oligonucleotides were ordered from Integrated DNA Technologies (IDT). Unless otherwise noted, chemical reagents were ordered from Sigma. Oligonucleotide sequences are listed in Supplementary Table 1.
Ethics: clinical samples
IAV seasonal clinical samples consisted of nasopharyngeal swabs taken from hospitalized participants who were able to give consent during routine testing at Addenbrookes Hospital during the 2016–2020 flu seasons. Routine reverse transcription (RT)–qPCRs were performed to confirm that persons were positive for either H1N1 or H3N2 and negative for other respiratory viruses. Samples were anonymized after routine laboratory testing and no identifiable information generated with the leftover material as determined by the East of England Cambridge Central Research Ethics committee. The study protocol was prepared by A.J.W.t.V. and H.J. and reviewed and approved by the Health Research Authority (IRAS ID 258438; REC reference 19/EE/0049).
COVID-19 clinical samples consisted of self-collected saliva samples submitted with informed consent by Princeton University students, faculty and staff members for surveillance testing43. The presence of SARS-CoV-2 RNA was confirmed using RT–qPCR; samples were subsequently anonymized before further testing and no identifiable information was generated during subsequent analyses of leftover material. The study protocol was prepared by C.M and A.J.W.t.V. and reviewed and considered exempt from human subject research by the Princeton University Institutional Review Board.
Cambodian H5N1 clinical samples consisted of nasopharyngeal and/or oropharyngeal swabs taken during suspected H5N1 infections or routine testing from persons through the Cambodian influenza-like illness (ILI) and severe acute respiratory illness (SARI) surveillance systems44,45. The Cambodian ILI and SARI surveillance systems are integral components of the World Health Organization (WHO) global influenza surveillance and response system. These public health initiatives are managed by Cambodia’s Ministry of Health (MoH) and the Communicable Disease Control (CCDC) Department. For each human H5N1 case, a collaborative One Health investigation was carried out by the CCDC and MoH, alongside the National Animal Health and Production Research Institute and provincial authorities. This effort was supported by the US Centers for Disease Control and Prevention, the WHO and the Food and Agriculture Organization of the United Nations.
The oseltamivir-resistant virus samples were identified through the Dutch national sentinel general practitioner surveillance for acute respiratory infections, the Dutch national A(H1N1)pdm09 influenza case finding program and the Amsterdam and Groningen University Medical Centers. Reference numbers for sequenced viruses are provided in Supplementary Table 2.
crRNA design
Most crRNA spacers were designed to be perfectly complementary to their 28-nt protospacer region. For SARS-CoV-2-targeting sequences, a single synthetic mismatch was inserted at position 5 to improve baseline specificity. Spacers were appended to the 3′ end of the consensus direct repeat sequence for the Cas13 ortholog used (Supplementary Table 1) and ordered from IDT as Alt-R guide RNA.
Target design
For our tiling experiment, we designed an RNA molecule with a length of 961 nt and minimal internal secondary structure. After an initial G nucleotide, the molecule comprises ten target blocks, each defined by a 34-nt buffer region, a 28-nt protospacer and a second 34-nt buffer region. We sought to have as many as possible of the 28-nt protospacers resemble natural sequences.
To this end, we started with a set of 18,508 28-nt-long protospacer sequences compiled from the ADAPT dataset, which has a sequence composition representative of viral diversity10. A total of 3,391 sequences with poly(A), poly(C) or poly(U) stretches ≥ 5 nt or poly(G) stretches ≥ 4 nt were removed. Of the remaining sequences, we removed 6,459 that had low average measured activity, defined as ‹out_log_k› ≤ −2 (on a logarithmic scale from −4 to 0, where 0 is high activity) using the activity definitions and measurements from a previous study10. We used LandscapeFold46 with parameter m = 2 (m represents the minimum allowed stem length), disallowing pseudoknots, to predict the structure landscapes of the remaining sequences. LandscapeFold predicted that 1,287 of these remaining sequences had extremely low intramolecular structure, defined as all nucleotides having a ≥40% probability of being unpaired in equilibrium.
We then aimed to find a set of these sequences that were all dissimilar from one another. First, given a sequence s, we found all those sequences with a Hamming distance ≤ 15 from s. A pair of sequences with a Hamming distance of h share all but h nucleotides. Of these sequences, we chose the one with the least secondary structure to keep and removed the others, with total amount of secondary structure quantified as \({\varSigma }_{n}{p}_{n}\) where the sum is over nucleotides and \({p}_{n}\) is the probability of the nucleotide being paired in equilibrium. We repeated this step for each sequence s we did not already remove. Next, we used a Smith–Waterman alignment47 to check for sequence similarity in nonidentical nucleotide positions, repeating the same procedure as above but, instead of Hamming distance, using the criterion of an alignment score ≥ 9 to define sequence similarity, where the alignment score parameters were (+1, −2, −2) for (match, mismatch, gap). This procedure resulted in a set of 20 sequences all distant from one another in sequence space.
Lastly, although we ensured each of these sequences had low secondary structure, we wanted to minimize binding between these sequences. For each pair of sequences, we used LandscapeFold with parameter m = 3 to predict the structure of the two strands, allowing for both intramolecular and intermolecular interactions. We defined two sequences to be incompatible if the resulting prediction had any nucleotide on either sequence with a ≤40% probability of being unpaired in equilibrium. We exhaustively enumerated the possible ordered sets of mutually compatible sequences, finding 60 ordered sets of five mutually compatible sequences and no set of six mutually compatible sequences. Of these 60 sets, we chose the one with the least structure. Under the assumption that entropic loop closure costs will create a barrier to non-neighbor sequence pairing (that is, where each sequence is less likely to pair to a sequence that is not its neighbor), we defined structure here as the sum, over the four pairs of neighboring sequences, of the maximum probability of a nucleotide being paired in that sequence pair. Thus, we arrived at a set of five distinct sequences from ADAPT with minimal intramolecular and intermolecular structure. These five sequences became the protospacer sequences corresponding to crRNAs 2, 4, 6, 8 and 9.
The other five protospacer sequences and the buffer regions were compiled out of 64 16-nt-long DNA sequences with minimal internal structure from Shortreed et al.23. Seven of these sequences with poly(A) or poly(T) stretches ≥ 5 nt were removed. Concatenating these sequences resulted in a long sequence with minimal structure, which we used to construct the rest of the 961-nt-long RNA target. We used NUPACK 3 (ref. 48) to predict the structure of the resulting target, finding various predicted stems. We then made individual point mutations in the buffer regions and non-ADAPT-derived protospacers to minimize the probabilities of the resulting stems (ensuring that NUPACK predicted no base pair forming with probability ≥ 60% in equilibrium) and to remove sequence similarity between targets (ensuring that there were no more than five identical consecutive nucleotides between the protospacer regions, no more than six identical consecutive nucleotides between two regions spanning a protospacer and a buffer and no more than eight identical consecutive nucleotides in buffer regions).
Lastly, we created a ‘shuffled’ version of the target, placing the target blocks (numbered 1–10 from 5′ to 3′ in the original target) in the following order: 1, 4, 2, 7, 5, 3, 9, 6, 8 and 10. We ensured that NUPACK 3 did not predict any base pair forming with probability ≥ 60% in the resulting sequence.
For our initial experiments (Fig. 1), we filtered the ADAPT dataset sequences to those with high activity (‹out_log_k› > −2) and perfect complementarity between target and crRNA in the ADAPT dataset. We then measured LandscapeFold’s prediction of the secondary structure of each candidate protospacer sequence. For each nucleotide, we calculated the total probability that the nucleotide is unpaired in equilibrium. The protospacer chosen had each nucleotide with at least a 92% probability of being unpaired in equilibrium.
RNA preparation (including structured targets)
RNA targets were ordered from IDT as DNA containing a T7 promoter sequence. Targets were then transcribed to RNA using the T7 HiScribe high-yield RNA synthesis kit in 55-μl reactions (New England Biolabs) with a 16-h incubation step at 37 °C and purified with 1.8× volume AMPure XP beads (Beckman Coulter) with the addition of 1.6× isopropanol, before eluting into 20 μl of nuclease-free (NF) water. All RNAs were then quantified using a NanoDrop One (Thermo Fisher Scientific) or Biotek Take3Trio (Agilent) and then stored in NF water at −80 °C for later use.
Occluded targets and crRNAs were prepared by mixing DNA and RNA oligo occluders with target RNA or crRNA in 60 mM KCl (Invitrogen) in NF water at a ratio of 2:1 (BioMark assays) or 10:1 (plate reader assays) and put through an annealing cycle consisting of a high-temperature melting step at 85 °C for 3 min, followed by gradual cooling to 10 °C at 0.1 °C s−1 followed by cooling to 4 °C. For massively multiplexed assays, occluders were first pooled by length and start position within the target block (Extended Data Fig. 1d) such that each resulting oligo pool contained all eight n-mers binding to a given position within each of the experimental target blocks. Targets and crRNAs were then used for detection assays immediately as described below.
Targets were input into detection reactions at various concentrations. For experiments in Fig. 1, targets were input at 7.5 × 108 cp per μl. For experiments in Fig. 2a–f, targets were input at 8 × 108 cp per μl. For Fig. 2h, targets were input at 5 × 109 cp per μl. For experiments in Fig. 3a–f, input concentrations of 7.5 × 109 cp per μl were used unless otherwise noted in figure caption. For Fig. 3g, targets were spiked in at the indicated allele frequency into a background of 5 × 1010 cp per μl (for occluded conditions) or 5 × 108 cp per μl (for nonoccluded conditions). For Fig. 3h, occluded conditions used an input concentration of 5 × 1010 cp per μl, whereas nonoccluded conditions used a concentration of 5 × 108 cp per μl.
Virus strains and seedstocks
Viral strains used are listed in Supplementary Table 2. Target controls were amplified from plasmids. Extracted viral genomic RNA samples were acquired from BEI Resources (hCoV-19/USA/MD-HP05285/2021 (B.1.617.2) Delta and hCoV-19/USA/GA-EHC-2811C/2021 (B.1.1.529) Omicron). Amplification reactions using 1 or 2 μl of viral RNA as the input (total reaction volume: 50 μl) were performed using the Qiagen One-Step RT–PCR kit according to the manufacturer’s specifications.
Clinical sample amplification
Total RNA was extracted from clinical samples using the trizol–chloroform method. Extracted RNA was then amplified using Qiagen One-Step RT–PCR (UK seasonal influenza samples, US SARS-CoV-2 samples and Cambodia H5N1 samples and isolates) or RT–recombinase polymerase amplification (TwistDx; Netherlands seasonal influenza samples, select SARS-CoV-2 samples) using either 1 or 2 μl of input material.
Cas13 detection assays
Standard bulk detection assays were performed by mixing target RNA or complementary DNA (cDNA) at a ratio of 10% v/v with 90% Cas13 detection mix. The detection mix consisted of 1× RNA detection buffer (20 mM HEPES pH 8.0, 54 mM KCl and 3.5% PEG-8000 in NF water), supplemented with 45 nM purified LwaCas13a (Genscript; stored in 100 mM Tris-HCl pH 7.5 and 1 mM DTT), 1 U per μl murine RNAse Inhibitor (New England Biolabs), 62.5 nM fluorescent reporter (/5FAM/rUrUrUrUrUrU/IABkFQ/; IDT), 22.5 nM processed crRNA (IDT) and 14 mM magnesium acetate. In experiments using crRNA occlusion, crRNAs were preannealed to DNA occluders as described above and used at a final concentration of 22.5 nM. Experiments using cDNA as the input included 0.3 mM ribonucleotide triphosphates (rNTPS; New England Biolabs) and 1 U per μl T7 polymerase (Biosearch). Minor adjustments to the detection mix were made for experiments using other orthologs of Cas13. For RfxCas13d, the Cas13 concentration was set to 90 nM and crRNA concentration was set to 45 nM. For LbuCas13a, Cas13 concentration was set to 10 nM and crRNA concentration to 5 nM. The 15-μl reactions were loaded in technical triplicate (Fig. 1) or duplicate (Figs. 2h and 3d–h and Extended Data Figs. 5–9) onto a Greiner 384-well clear-bottom microplate (Greiner, 788096) and measured on an Agilent BioTek Cytation 5 or Synergy H1 microplate reader for 3 h with excitation at 485 nm and detection at 528 nm every 5 min.
For tiling assays and the mCARMEN KRAS assay, Standard Biotools genotyping IFC (192.24 format) was used in a BioMark HD for multiplexed detection. Assay mix (10% of final reaction volume) contained 1× assay detection mix (Standard Biotools) supplemented with 100 nM crRNA, 100 nM LwaCas13a (Genscript; stored in 100 mM Tris-HCl pH 7.5 and 1 mM DTT). Sample mix (90% of final reaction volume) contained 1× sample buffer (44 mM Tris-HCl pH 7.5, 5.6 mM NaCl, 10 mM (tiling experiment) or 2 or 14 mM (KRAS) MgCl (comparison of Mg concentrations in Extended Data Fig. 6d), 1.1 mM DTT and 1.1% w/v PEG-8000), supplemented with murine RNAse Inhibitor (1 U per μl; New England Biolabs), fluorescent reporter (500 nM; IDT), 1× ROX reference dye (used for normalization of random fluctuations in fluorescence between chambers; Standard Biotools), 1× GE Buffer (Standard Biotools), 20 mM KCl and occluded RNA target (9 × 108 cp per μl). Experiments using cDNA as the input included 0.9 mM rNTPS (New England Biolabs) and 0.125 U per μl T7 polymerase (Biosearch).
Sample volumes of 3.5 μl and assay volumes of 3.5 μl, in addition to appropriate volumes of control line fluid, actuation fluid, and pressure fluid (Standard Biotools), were loaded onto the 192.24 genotyping IFC chip (Standard Biotools). Chips were then placed into the Fluidigm Controller and loaded and mixed using the Load Mix 192.24 GE script (Standard Biotools).
After mixing, reactions (two technical replicates each) were run on BioMark HD at 37 °C for 8 h, with measurements taken in the FAM and ROX channels every 5 min. Normalized and background-subtracted fluorescence for a given time point was calculated as (FAM − FAM background)/(ROX − ROX background).
For fluorescence in-tube detection assays, Cas13 detection mix was prepared as in bulk detection assays with fluorescent reporter raised to 250 nM and crRNA raised to 45 nM in the final reaction. The 33-μl reactions were incubated at 37 °C for 3 h. Every 30 min, reactions were visualized with ultraviolet light on a transilluminator and captured with a smartphone camera.
Nondenaturing acrylamide gel electrophoresis
Two separate mixtures were prepared for the target RNA with occluders and the ribonucleoprotein (RNP) complex. Target RNA and occluders were first annealed at 1:1 ratio in 7.5 μl of tube A (final concentration 2 μM each in 7.5 μl) containing 1× binding buffer (24 mM KCl, 4 mM Tris-HCl pH 8.0, 0.4 mM DTT, 10% glycerol, 0.1 mg ml−1 BSA and 5 mM MgCl2) at 85 °C for 10 min followed by gradual cool down at a rate of 0.1 °C s−1. To prepare the RNP complex, dLwaCas13a was mixed with crRNA at 1:1.33 ratio in 7.5 μl of tube B (final concentration 2 μM and 2.67 μM, respectively, in 7.5 μl) containing 1× binding buffer and incubated at 37 °C for 15 min. Next, 7.5 μl of tube A was added to tube B to bring the final volume to 15 μl, followed by another 37 °C incubation for 30 min. Samples were then loaded into 5% Mini-PROTEAN TBE gel (Bio-Rad) and run in 0.5× TBE buffer. The gel was analyzed by imaging the FAM and Cy5 channels on an Azure Biosystems 600 imager before staining with SYBR gold for total nucleic acid visualization.
Activity fits
Fluorescence curves were converted to activity scores by fitting the curves to effectively first-order reactions. With a certain amount of active Cas13, the concentration of uncleaved reporter is expected to decrease exponentially according to the reaction
$${E}^{\star }+U\to {E}^{\star }+P$$
where \({E}^{\star }\) is the concentration of active Cas13, U is the concentration of uncleaved reporter and P is the concentration of cleaved reporter RNA. Labeling the (second-order) rate constant of this reaction as r, the concentration of P changes over time according to
$$P
