The Unexpected Off-Target
Jennifer Doudna thought she had everything figured out when her team designed a CRISPR system to correct a single nucleotide causing blindness in laboratory mice. The guide RNA was perfectly matched. The target sequence appeared nowhere else in the genome according to their computational analysis. Yet when they sequenced the edited cells weeks later, they found cuts in seventeen different locations across four chromosomes. None of these matched their intended target.
This wasn’t a rookie mistake or sloppy technique. This was 2019, seven years after CRISPR-Cas9 had changed molecular biology forever, and one of the technology’s pioneers was facing an uncomfortable truth. Even our most sophisticated gene editing tools remain stubbornly unpredictable, cutting DNA in places we never intended and sometimes missing their targets entirely.
The Prime Editing Promise That Wasn’t
In October 2019, David Liu’s laboratory at the Broad Institute published what seemed like the holy grail of gene editing. Prime editing could theoretically make precise insertions, deletions, and substitutions without creating double-strand breaks. The initial Nature paper showed remarkable precision in cultured human cells, with efficiency rates reaching 50% for certain edits and minimal off-target activity.
But when other laboratories tried to reproduce these results in living organisms, reality proved messier. A 2021 study attempting prime editing in mouse embryos achieved success rates below 5% for most targets. Even more troubling, researchers at Stanford found that prime editing sometimes created large deletions spanning thousands of base pairs, exactly the kind of damage it was designed to avoid. The tool worked, but not reliably enough for most therapeutic applications.
These failures weren’t hidden in supplementary data or buried in obscure journals. They appeared in high-profile publications because the scientific community has learned something important: documenting what doesn’t work is as valuable as celebrating what does. Each failed experiment reveals another layer of complexity in the genome’s defensive mechanisms.
Base Editing’s Bystander Problem
Base editors represent another attempt to make gene editing more surgical. Instead of cutting DNA entirely, they chemically convert one nucleotide to another. Cytosine base editors can change C-G base pairs to T-A pairs, potentially correcting thousands of disease-causing mutations with a single tool. The elegance is undeniable.
Yet base editors create their own category of unintended consequences. In 2021, researchers discovered that cytosine base editors don’t just edit their intended targets. They also modify cytosines in a roughly 150-nucleotide window around the target site, creating what scientists call “bystander editing.” Sometimes these bystander edits are harmless, falling in non-coding regions. But when they occur in coding sequences, they can create new mutations while fixing the original problem.
This discovery emerged from a failed clinical trial preparation. Researchers planning to treat sickle cell disease noticed that their base editor was creating unwanted amino acid changes in the same gene they were trying to fix. The revelation forced a complete reassessment of base editing strategies and highlighted how much we still don’t understand about these molecular scissors.
The Delivery Dilemma
Even perfect gene editors are useless if they can’t reach their targets. Delivery remains CRISPR’s most persistent challenge, and the failures here are particularly telling. Lipid nanoparticles, the gold standard for delivering mRNA vaccines, work poorly for CRISPR components. The particles tend to accumulate in the liver, making it nearly impossible to edit genes in the brain, heart, or muscle tissue where many genetic diseases actually occur.
In 2020, a promising trial for Duchenne muscular dystrophy was halted after preliminary results showed the gene editing machinery reached less than 1% of muscle fibers. The company, Solid Biosciences, had spent five years developing their delivery system, yet the majority of CRISPR never made it past the injection site. Patients showed no clinical improvement, and several developed immune responses against the viral delivery vector.
These delivery failures have sparked entirely new research directions. Scientists are now engineering viruses that target specific cell types, developing protein-based delivery systems, and even exploring physical methods like ultrasound to create temporary pores in cell membranes. Each approach brings its own set of potential failures, but also new possibilities for success.
Embracing the Uncertainty
The accumulating evidence of CRISPR’s limitations hasn’t dampened scientific enthusiasm. Instead, it’s created a more realistic understanding of genome editing. Researchers now routinely test for off-target effects using multiple detection methods. They’re developing new computational tools to predict where unwanted cuts might occur. Most importantly, they’re designing clinical trials with realistic expectations about efficacy and safety.
The latest generation of CRISPR tools reflects this hard-earned wisdom. Cas variants like Cas12f1 are smaller and potentially more specific than the original Cas9. New guide RNA designs incorporate chemical modifications that improve targeting precision. Scientists are even developing “self-destructing” CRISPR systems that stop working after completing their edits, reducing the risk of long-term off-target activity.
What strikes me most about these developments is how they emerged directly from documented failures. The bystander editing problem led to improved base editor designs. Off-target cutting drove the development of high-fidelity Cas variants. Each setback became a stepping stone toward more reliable gene editing tools. In a field moving as rapidly as genome editing, perhaps the most important question isn’t whether our next experiment will succeed, but what we’ll learn when it doesn’t go according to plan.