Genetic modification works by making targeted changes to an organism’s DNA, either by cutting, replacing, or inserting specific sequences. The exact method depends on whether scientists are editing a human cell, a crop plant, or a microorganism, but the core logic is the same: get a molecular tool to a precise location in the genome, make a change, and let the cell’s own repair machinery finish the job. Several techniques exist, ranging from older protein-based tools to the now-dominant CRISPR system, and each has distinct strengths.
How CRISPR Editing Works
CRISPR-Cas9 is the most widely used gene-editing tool today, and its mechanism breaks down into three steps: recognition, cleavage, and repair. Scientists design a short piece of synthetic RNA, called a guide RNA, that matches the DNA sequence they want to edit. This guide RNA pairs up with one strand of the target DNA through standard base pairing, the same A-T and C-G matching that holds the two strands of DNA together naturally.
The Cas9 protein, which acts as molecular scissors, stays inactive until it’s bound to the guide RNA. Once the guide RNA locks onto its target, Cas9 cuts through both strands of the DNA at a precise point, three base pairs away from a short recognition tag called a PAM sequence. One part of the Cas9 protein cuts one strand, and a different part cuts the other, producing a clean break in the double helix. This double-strand break is what triggers the cell to respond.
What Happens After the Cut
Once the DNA is broken, the cell activates its own repair systems. There are two main pathways, and which one kicks in determines the outcome of the edit.
The faster, messier option is called non-homologous end joining. The cell essentially glues the broken ends back together, but it often adds or deletes a few letters of DNA in the process. These small errors are enough to disable a gene entirely, which is useful when the goal is to knock out a gene that causes disease or produces an unwanted trait.
The more precise option is homology-directed repair. Here, scientists supply a DNA template alongside the CRISPR components. The cell uses this template as a blueprint to rebuild the broken section, incorporating whatever new sequence the researchers designed. This is how specific corrections or insertions are made. The challenge is that cells strongly prefer the quick-and-dirty repair over the precise version. In practice, many edited cells end up with one copy of the gene correctly repaired and the other disrupted by the error-prone pathway, which means researchers often need to screen many cells to find ones with the exact edit they want.
Editing Without Breaking Both Strands
Because double-strand breaks can be imprecise, newer tools avoid them altogether. Base editors, first developed in 2016, fuse a modified Cas9 protein (one that can find and bind DNA but not fully cut it) to a chemical enzyme that directly converts one DNA letter into another. Cytosine base editors convert a C-G pair into a T-A pair by chemically transforming cytosine into uracil, which the cell then reads as thymine during replication. Adenine base editors do the reverse trick, converting an A-T pair into a G-C pair by changing adenine into inosine, which the cell reads as guanine.
Prime editing goes further. It uses a Cas9 that nicks only one strand of DNA, fused to a reverse transcriptase enzyme that can write new DNA from an RNA template. The guide RNA in this system carries not just the targeting sequence but also the desired edit spelled out in RNA code. After nicking one strand, the reverse transcriptase uses the guide RNA as instructions to write the corrected sequence directly into the genome. Prime editing can make all twelve possible single-letter swaps, plus small insertions and deletions, without ever creating a full double-strand break.
Older Editing Tools: ZFNs and TALENs
Before CRISPR, scientists used two protein-based systems. Zinc finger nucleases pair a DNA-cutting enzyme with small protein modules called zinc fingers, each of which recognizes three to four letters of DNA. Stringing several zinc fingers together creates a protein that binds a unique sequence, typically 9 to 18 letters long, and cuts at that spot.
TALENs work similarly but use a different type of DNA-binding module. Each module is a 33- to 35-amino-acid repeat that recognizes a single DNA letter, creating a simple one-to-one code between protein and target. Both ZFNs and TALENs require engineering an entirely new protein for every new target site, which means assembling and testing large DNA constructs of 500 to 1,500 base pairs each time. CRISPR made this obsolete: targeting a new gene requires only swapping out a 20-letter RNA sequence, while the Cas9 protein stays the same.
Getting the Tools Into Cells
Designing the edit is only half the problem. The editing machinery also has to physically enter the cell and reach the nucleus. Several delivery methods exist, each suited to different situations.
Viral vectors are the most established delivery system for human gene therapy. Adeno-associated viruses (AAVs) have become the preferred choice in clinical trials and approved therapies because they are non-pathogenic, rarely integrate into the host genome, and can sustain long-term gene expression. Their main limitation is a small cargo capacity, which restricts the size of genetic material they can carry. Other viral options, including lentiviruses and adenoviruses, can carry larger payloads but come with trade-offs in safety or durability.
Lipid nanoparticles offer a non-viral alternative that gained widespread visibility through mRNA vaccines. These are tiny fat-based spheres built from four components: ionizable lipids that bind and protect the genetic cargo, helper lipids that stabilize the particle, cholesterol for structural integrity, and a coating of PEG molecules that helps the particles evade the immune system and circulate longer. Once inside a cell, the ionizable lipids respond to the acidic environment of the cell’s recycling compartments, destabilizing the membrane and releasing the cargo into the cell’s interior. Lipid nanoparticles are particularly well suited for delivering mRNA or guide RNA because these molecules only need to reach the cytoplasm, not integrate into the genome.
Electroporation is a simpler physical method: brief electrical pulses create temporary pores in the cell membrane, allowing editing components to slip through. This works well for cells that can be edited outside the body, like blood cells, then returned to the patient.
How Plants Are Genetically Modified
Plant genetic modification uses a different set of delivery strategies. The most common leverages a natural process: the soil bacterium Agrobacterium tumefaciens has evolved to inject its own DNA into plant cells. Scientists replace the bacterium’s natural DNA payload with whatever gene they want to introduce. When the bacterium contacts a plant cell, it processes this DNA into a single strand, attaches a pilot protein to the leading end, and sends it through a specialized molecular syringe (a type IV secretion system) into the plant cell. Once inside, companion proteins coat the DNA strand, protecting it and helping guide it into the plant cell’s nucleus, where it integrates into the genome.
For plant species that don’t respond well to Agrobacterium, scientists use a gene gun. This device coats microscopic gold particles, typically around 0.6 micrometers in diameter, with the desired DNA. A burst of helium gas at roughly 650 psi accelerates these particles through a vacuum chamber and into plant tissue. The gold is biologically inert, so it sits harmlessly inside the cell while the DNA dissolves off and finds its way into the nucleus. This biolistic method works on a wider range of species but is less precise about where the new DNA lands in the genome.
Accuracy and Off-Target Effects
No editing tool is perfectly precise. CRISPR-Cas9 can sometimes cut at sites that closely resemble the intended target, creating unintended mutations elsewhere in the genome. The severity of this problem depends on the guide RNA design and the specific target. In controlled experiments, off-target mutation rates with standard Cas9 have been measured as high as 57% at sites with just a single mismatch from the intended sequence. Even with three mismatches, off-target editing rates of 7 to 12% have been documented.
This has driven significant engineering effort. Modified versions of Cas9, chemical modifications to the guide RNA, and conjugation with protective polymers have all reduced off-target rates. Some optimized systems cut off-target editing to below 4% at the vast majority of tested sites. Base editors and prime editors also tend to produce fewer off-target effects since they don’t create full double-strand breaks, though they have their own forms of imprecision.
From Lab to Approved Therapy
These techniques have moved well beyond the research bench. The FDA has approved more than 35 cellular and gene therapy products. These include treatments for inherited blood disorders like sickle cell disease, inherited blindness, spinal muscular atrophy in infants, and several types of cancer. One of the most notable recent approvals, Casgevy, is the first therapy based on CRISPR editing, used to treat sickle cell disease by editing a patient’s own blood stem cells outside the body before returning them. In agriculture, gene-edited crops with improved disease resistance, longer shelf life, and enhanced nutritional profiles are already on the market in several countries, often developed using CRISPR applied through Agrobacterium or gene gun delivery.