CRISPR diagnostics

CRISPR became famous as genome editing, but diagnosis borrows the machinery without borrowing its purpose: nothing here touches a cell. The trick that turned defensive bacterial immunity into clinical testing is a strange enzymatic habit called collateral cleavage — once certain Cas proteins find their intended target, they begin cutting everything nearby of the right chemical class, guilty or innocent. A diagnostic simply arranges for “everything nearby” to be countable evidence.

One recognition event, millions of casualties

Cas12 recognises DNA targets; Cas13 does the same for RNA. Each arrives guided by a small RNA whose twenty-ish positions pair programmatically against the sequence of interest — this pairing supplies all the identity decision, so rewriting the test means ordering a different short RNA rather than developing new antibodies. In isolation that recognition would be just another very good affinity interaction. But binding flips the enzyme into an active state that shreds arbitrary single-stranded nucleic acid in solution while ignoring anything double-stranded. Fill the tube beforehand with short synthetic reporters — a fluorescent dye held dark by a quencher until cut, or a hapten-tagged strand sized to travel intact down a paper strip — and one activated enzyme demolishes reporter after reporter by the thousand per minute. Signal rises as a direct census of how much target triggered the system: a molecular relay race where the leg run by the guide decides who, and the leg run by cleavage decides how much.

The quiet partnership upstream

Collateral cleavage accounts for spectacular published limits — individual copies detected within half an hour — but those figures always assume help feeding the front end. Enzyme activation requires finding the target amid enormous dilution, and any real specimen holds far less than one copy per reaction volume at clinically relevant thresholds, so an isothermal amplification stage runs first: recombinase-polymerase or loop-mediated copying inflates the target population until collisions between enzyme and quarry become likely. Identification remains the CRISPR layer’s job, performed on the amplified products; abundance was created upstream. This division also frames the honest reading of sensitivity claims — they belong to the pipeline, never to the nuclease step alone.

Where it strains

Three constraints temper the elegance. Multiplexing several targets in one tube requires orthogonal enzymes or distinguishable reporters, since every activated nuclease attacks the whole shared pool — clever combinations of Cas12 and Cas13 answer only modest panel sizes. Quantification lags the binary answer, because indiscriminate shredding saturates quickly; kinetics help but demand instrumented fluorescence rather than paper strips. And guide design inherits the usual specificity subtleties: mismatches near the seed region poison recognition while distant ones pass unnoticed, tolerances that differ across the Cas family and must be validated per assay rather than assumed. None of these points subtracts from the genuine achievement — genome-searchable specificity delivered to an unpowered readout — but they mark precisely where current work concentrates.