FLASH: a next-generation CRISPR diagnostic for multiplexed detection of antimicrobial resistance sequences

Abstract

The growing prevalence of deadly microbes with resistance to previously life-saving drug therapies is a dire threat to human health. Detection of low abundance pathogen sequences remains a challenge for metagenomic Next Generation Sequencing (NGS). We introduce FLASH (Finding Low Abundance Sequences by Hybridization), a next-generation CRISPR/Cas9 diagnostic method that takes advantage of the efficiency, specificity and flexibility of Cas9 to enrich for a programmed set of sequences. FLASH-NGS achieves up to 5 orders of magnitude of enrichment and sub-attomolar gene detection with minimal background. We provide an open-source software tool (FLASHit) for guide RNA design. Here we applied it to detection of antimicrobial resistance genes in respiratory fluid and dried blood spots, but FLASH-NGS is applicable to all areas that rely on multiplex PCR.

Figure 1.

Overview of the FLASH method. Genomic DNA or cDNA is first blocked with phosphatase treatment and then digested with Cas9 complexed to a set of guide RNAs targeting genes of interest. Ligation of sequencing adaptors, amplification and sequencing follows.

Figure 2.

Results of FLASH on cultured isolates. (A) Proportion of targeted genes detected in FLASH-NGS and NGS libraries of six S. aureus isolates. Average of three replicates. chr: chromosomal gene; acq: acquired gene. Numbers in boxes represent, in order for each isolate: i. number of targets present (based on NGS); ii. number of expected targets observed in FLASH-NGS; iii. number of false-positive targets observed in FLASH-NGS (false positive not depicted in pie chart). (B) For isolate 1 with FLASH, <2000 sequencing reads were needed to achieve coverage of each targeted gene by at least 10 reads. Over 100 000 reads were needed to achieve the same coverage with NGS alone. (C) The fraction of targeted reads relative to background decreases substantially below 100 pg of DNA input; however, with as little as 100 fg input (∼35 copies of the S. aureus genome), the full set of targeted genes was detected at 10 rpM or greater. Bars and error bars represent mean and standard deviation of three replicates.

Figure 3.

Results of FLASH on respiratory samples. Number of reads aligning to targeted genes in NGS and FLASH-NGS sequencing experiments on respiratory fluid samples from (A) patient 1, (B) patient 2, (C) patient 3 and (D) patient 4. Average of three or six replicates. Bars and error bars represent mean and standard deviation of three to six replicates (see Supplementary Table S5).

Figure 4.

Results of FLASH on dried blood spots. (A) Dried blood spots (DBSs) from malaria lab strains were sequenced using either NGS or FLASH-NGS. Reads per million is plotted for each window for nine FLASH-NGS samples (three strain mixtures, each in triplicate, blue) and nine equivalent NGS samples (pink). Top panel indicates fold enrichment (average FLASH-NGS rpM divided by average NGS rpM). Light green bars in the upper panel represent windows for which no haplotype-determining read pairs were found. (B) Both lab strain mixtures and patient samples were evaluated for the presence of drug resistant haplotypes. Bars indicate mean of three replicates. (C) Target windows with sequences that distinguish the lab strains D10, HB3 and U659 were used to estimate strain ratios in the three mixtures. For the clinical DBS samples, the number of sequencing reads attributable to different haplotypes at each window was determined using SeekDeep. The average proportion of haplotypes at each of the maximum-haplotype containing windows is depicted. Bars indicate mean of three replicates.