Carmen 2003

[Romilda Tiger]

CarmenLynch has two comedy specials: one in English titled Queef Week and one in Spanish called Carmen Lynch en Espaol (both streaming on YouTube.) Carmen also has a multitude of late night TV appearances including The Tonight Show Starring Jimmy Fallon, The Late Show with Stephen Colbert, The Late Show with David Letterman, Conan and The Late Late Show with Craig Ferguson. Carmen has also appeared on Inside Amy Schumer, That Damn Michael Che, Life & Beth and the documentary Hysterical (produced by Jessica Kirson.) Carmen can be found on all social media @carmencomedian.

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Infectious diseases are some of the greatest threats to human health and global security, yet there is no broadly available molecular test for the vast majority of disease-causing microbes, limiting their diagnosis and surveillance. Of the many viral species capable of infecting humans (576 of which had been sequenced and 169 of which had at least 10 published genomes12 by October 2018), only 39 had diagnostics approved by the FDA (US Food and Drug Administration; ). While laboratory developed tests have been developed for clinical testing of diverse pathogens at specific facilities, these tests can have long turnaround times and are rarely multiplexed. Routine comprehensive diagnostic testing would provide a previously unavailable data stream to inform patients, healthcare workers and policy makers to suppress and mitigate outbreaks. However, these tools are not widely available owing to the lack of a scalable and multiplexed technology to quickly and inexpensively identify any circulating pathogen (Fig. 1a). Comprehensive disease detection by sequencing or microarray hybridization provides detailed information about pathogen genotypes and evolution, but is difficult to implement on a wide scale owing to the cost and logistical demands of sample preparation4,5,6,13. Rapid, low-cost detection methods, such as CRISPR-based approaches, antigen-based tests, PCR or loop-mediated isothermal amplification (LAMP), detect only one or a small number of pathogens in a given reaction1,2,3,7,14,15,16. Combining the strengths of these approaches, an ideal diagnostic and surveillance technology would be highly multiplexed and easily scale across hundreds of samples.

Miniaturized and self-organizing microfluidic technology enables massive multiplexing of biochemical and cellular assays17,18,19,20,21. We recently developed a microwell-array system that harnesses miniaturization and self-organization to perform comprehensive combinatorial experiments. In this system, the user prepares a collection of inputs as droplet emulsions, and the input droplets organize themselves in the wells of the array, creating all possible pairwise combinations in replicate without additional user effort or active instrumentation8. We envisioned that CRISPR-based nucleic acid detection could be integrated with the microwell-array system to test many amplified samples for many analytes in parallel.

Accurate testing of multiple samples for hundreds of microbial pathogens requires higher throughput than is offered by existing multiplexed detection systems2,24,25. To enable highly multiplexed detection with high sample throughput, we developed a set of 1,050 solution-based colour codes using ratios of 4 commercially available, small-molecule fluorophores. Using the 1,050 colour codes, 99.5% of droplets can be correctly classified after permissive filtering that retains 94% of droplets (Extended Data Fig. 4, Supplementary Discussion 5). To match the throughput enabled by our 1,050 colour codes, we designed a massive-capacity chip (mChip) that allows more than 4,500 statistically replicated tests per chip (Extended Data Fig. 5). mChip reduces the reagent cost per test more than 300-fold relative to standard multiwell-plate SHERLOCK tests, while reducing pipetting steps and turnaround time (Extended Data Table 1, Supplementary Discussions 6, 7, 8).

As an outbreak of COVID-19 emerged during the manuscript review process, we rapidly incorporated a new test27 for the novel coronavirus SARS-CoV-2 into a coronavirus panel taken from the HV panel, demonstrating the power of this modular master set to be adapted to real-world challenges (Fig. 2d). Using a single mChip, more than 400 samples can be tested in parallel against our coronavirus panel.

Human samples from patients with dengue, HCV, HIV and Zika were obtained commercially from Boca Biolistics under their ethical approvals. Influenza samples were obtained from the Centers for Disease Control and Prevention under their ethical approvals. All protocols subsequently performed by the researchers were approved as a Not Human Subjects Research determination no. NHSR-4318 issued by the Broad Institute of MIT and Harvard.

Synthetic DNA targets were ordered from Integrated DNA Technologies and resuspended in nuclease-free water. Resuspended DNA was serially diluted to 104 copies per μl and used as inputs to PCR or RPA reactions.

For droplet pooling, a total droplet pool volume of 150 l of droplets was used to load each standard chip; a total of 800 l of droplets was used to load each mChip. To maximize the probability of forming productive droplet pairings (amplified sample droplet + detection reagent droplet), half the total droplet pool volume was devoted to target droplets and half to detection reagent droplets. For pooling, individual droplet mixes were arrayed in 96W plates. A multichannel pipette was used to transfer the requisite volumes of each droplet type into a single row of eight droplet pools, which were further combined to make a single droplet pool. The final droplet pool was pipetted up and down gently to fully randomize the arrangement of the droplets in the pool. The pooling step is rapid (

Imaging data were analysed with custom Python scripts. Analysis consisted of three parts: (1) pre-merge image analysis to determine the identity of the contents of each droplet based on droplet colour codes; (2) post-merge image analysis to determine the fluorescence output of each droplet pair and map those fluorescence values back to the contents of the microwell; (3) statistical analysis of the data obtained in parts 1 and 2.

The contents of each droplet were determined from images taken before droplet merging: a background image was subtracted from each droplet image, and fluorescence channel intensities were scaled so the intensity range of each channel was approximately the same. Droplets were identified using a Hough transform, and the fluorescence intensity of each channel at each droplet position was determined from a locally convolved image. Compensation for cross-channel optical bleed was applied, and all fluorescence intensities were normalized to the sum of the compensated 647 nm, 594 nm and 555 nm channels. For 4-channel datasets, analysis of 3-colour space was performed directly on normalized intensities. For 5-channel datasets, droplets were divided into UV intensity bins for downstream analysis (Extended Data Fig. 4). The 3-colour space within each UV bin was analysed separately. The 3-colour intensity vectors for each droplet were projected onto the unit 2-simplex, and density-based spatial clustering of applications with noise (DBSCAN) was used to assign labels to each colour code cluster. Manual clustering adjustments were made when necessary. For 5-channel datasets, UV intensity bins were recombined after assignments to create the full dataset.

The SNP index was calculated for each sample and each mutation by taking the ratio of the derived-allele-targeting crRNA and the ancestral-allele-targeting crRNA. In the heat maps, SNP indexes were normalized by row (in Fig. 4b, d).

For Zika detection experiments (Fig. 1c), detection mixes were supplemented with MgCl2 at a final concentration of 6 mM prior to droplet merging. For comparison between CARMEN and SHERLOCK (Extended Data Fig. 3b, c), a Biotek Cytation 5 plate reader was used for measuring fluorescence of the detection reaction. Fluorescence kinetics were monitored using a monochromator with excitation at 485 nm and emission at 520 nm with a reading every 5 min for up to 3 h.

For each crRNA, a threshold for detection was set at 3 s.d. above the background fluorescence. Cross-reactivity was defined as off-target reactivity above threshold. Low-reactivity was defined as no reactivity above threshold. Selective was defined as on-target reactivity above threshold and no cross-reactivity.

To determine whether any crRNA in an experiment was uninterpretable due to signal above background in healthy control samples, the median signal across all crRNAs was calculated for each control sample. (Reactivity of the control samples across the 169-plex panel is expected to be very sparse, so the median value is a reliable measure of background signal.) Next, for each crRNA, a ratio was calculated of (numerator) the signal from the control sample with that crRNA and (denominator) the median for that control sample across all crRNAs. If any crRNA showed reactivity with a control sample that was >6x the median signal for that control sample, the crRNA was considered to be uninterpretable for that experiment. For each interpretable crRNA, the signal from each sample was divided by the median signal from the healthy control samples for that crRNA. Signal that was 6 above the median background signal was considered a positive result.

The threshold for each crRNA may be set individually, as the reactivity of a crRNA is sequence-specific. For H-subtyping crRNA, the signal from each sample was divided by the median signal from the healthy control samples for that crRNA. Signal that was 6 above the median background signal was considered a positive result. The N-subtyping crRNAs are less reactive, so a more sensitive threshold is necessary to accurately differentiate signal from background. For each N-subtyping crRNA, the median and standard deviation of the control samples was calculated, and a threshold of 7 s.d. above the median was used to determine signal above background.

3a8082e126