CRISPR is the most talked-about technology in biology. In December 2023, the FDA approved Casgevy, the first gene therapy built on CRISPR technology, for sickle cell disease. In mid-2025, scientists created a personalized CRISPR treatment for an infant in just six months. These are real, historic achievements. But the distance between “CRISPR cured a blood disease” and “CRISPR will cure everything” is vast, and most coverage glosses over it.
This article maps what CRISPR gene editing can actually do today in living organisms, what it cannot do yet, and what it may never be able to do. No breathless futurism. Just the mechanism, the results, and the limits.
How CRISPR Works (The Short Version)
CRISPR is a molecular tool borrowed from bacteria. It uses a protein called Cas9, guided by a short piece of RNA, to find a specific sequence in DNA and cut it. The cell then repairs the break, and scientists exploit that repair process to delete, replace, or insert genetic material.
Think of it like a find-and-replace function for DNA. Except DNA is three billion letters long, the “replace” function is unreliable, and the cell often fixes the cut in ways nobody intended.
What CRISPR Can Actually Do Right Now
Cure single-gene blood diseases
This is CRISPR’s clearest success. Casgevy works by editing a patient’s own blood stem cells outside the body. The edit reactivates fetal hemoglobin, which prevents red blood cells from sickling. In clinical trials, 29 out of 31 evaluable patients (93.5%) were free from severe pain crises for at least 12 consecutive months. The therapy has since been approved in the US, UK, EU, and several other countries.
The key details: the editing happens outside the body (ex vivo), in blood stem cells that are then transplanted back. The patient must first undergo harsh chemotherapy to clear their bone marrow. It works because sickle cell disease is caused by a single, well-understood genetic mutation, and the target cells can be removed, edited, and returned.
Silence disease-causing genes in the liver
The most promising in vivo approach (editing inside a living body) targets the liver. Intellia Therapeutics’ treatment for hereditary transthyretin amyloidosis (hATTR) uses lipid nanoparticlesMicroscopic fat particles used to package and deliver genetic material into cells. The main vehicle for in vivo CRISPR delivery, with a natural affinity for liver cells., tiny fat droplets, to deliver CRISPR components through an IV infusion. The system knocks out the gene producing a toxic protein. Results published from their Phase I trial showed an average 90% reduction in the disease-causing protein, sustained for over two years with no sign of the effect fading.
This works because lipid nanoparticles naturally accumulate in the liver, and the therapeutic goal is simple: break a gene so it stops making a harmful protein. No precise repair needed.
Create personalized one-off therapies
In a landmark case reported in 2025, a team including researchers from the Innovative Genomics Institute developed a bespoke CRISPR therapy for an infant with a rare metabolic disorder (CPS1 deficiency) in just six months. The baby, KJ, received three doses delivered by lipid nanoparticles, with each dose further reducing symptoms. This proved that redosing is possible with LNP delivery, unlike viral vectors, which trigger immune responses on repeat use.
What CRISPR Cannot Do Yet
Edit most organs in the body
The liver gets all the attention because current delivery technology naturally goes there. Lipid nanoparticles accumulate in liver cells after intravenous injection. For diseases of the brain, heart, muscles, kidneys, or lungs, researchers are working on LNP versions that target different organs, but none have entered clinical trials. As the field puts it: the three biggest challenges in CRISPR medicine are delivery, delivery, and delivery.
Make precise corrections reliably
Most CRISPR successes so far involve breaking genes, not fixing them. Cutting DNA is relatively straightforward. Performing a precise repair, swapping one letter for another, inserting a corrected sequence, requires the cell to use a specific repair pathway (homology-directed repair) that is inherently inefficient in human cells. The dominant repair mechanism, non-homologous end joining, just glues the broken ends back together, often introducing small errors.
Newer approaches like base editing and prime editing address this. The first prime editing therapy (PM359) showed positive results in a patient with chronic granulomatous disease in May 2025, restoring immune function in 66% of the patient’s neutrophils. And MIT researchers have reduced prime editing error rates by up to 60-fold compared to earlier versions. But these are early days. Prime editing is slower, harder to deliver, and still being optimized.
Treat complex, multi-gene diseases
Heart disease, diabetes, most cancers, depression, schizophrenia: these conditions involve dozens to thousands of genetic variants, each contributing a tiny fraction of risk, interacting with environment and lifestyle. CRISPR edits one site at a time. Editing multiple sites simultaneously remains experimental and increases the risk of unintended damage.
A 2024 analysis in Nature modeled what polygenicDescribes a trait or disease influenced by many genes, each contributing a small effect. Most common diseases like diabetes and heart disease are polygenic. editing could theoretically achieve: editing just 40 genetic variants could reduce an individual’s lifetime risk of Alzheimer’s, diabetes, and heart disease to under 0.2%. But the same paper estimates this is roughly 30 years away from technical feasibility, and “highly uncertain” even then.
Guarantee safety from off-target effectsUnintended DNA edits that occur when a gene editing tool cuts at sites other than the intended target. A key safety concern in CRISPR therapy.
When CRISPR cuts DNA, it sometimes cuts in the wrong place. These off-target edits are a known risk. But a 2025 review in Nature Communications highlighted a less-discussed problem: large structural variations, including chromosomal translocations and deletions spanning millions of base pairs, that occur even at the intended target site. Standard sequencing methods can miss these entirely. The same review found that drugs used to improve editing precision can cause a thousand-fold increase in structural variation frequency.
Even newer, gentler approaches like base editing and prime editing do not fully eliminate these structural alterations. For approved therapies like Casgevy, the edited cells are screened before being returned to the patient, which provides a safety net. For in vivo editing, where you cannot inspect every edited cell, the stakes are higher.
What CRISPR Probably Cannot Do (Ever)
Create “designer babies”
The idea of editing embryos to enhance intelligence, athleticism, or appearance is science fiction dressed as science. These traits are polygenic (influenced by thousands of genetic variants), deeply entangled with environment, and poorly understood at the molecular level. The 2018 He Jiankui scandal, in which a Chinese researcher edited human embryos and produced three babies, demonstrated the recklessness of this approach: none of the babies received the intended edit correctly, and the novel mutations introduced have never been shown to provide the intended HIV resistance. He Jiankui was sentenced to three years in prison.
Heritable human germline editing remains banned or under moratorium in virtually every jurisdiction. The scientific consensus is clear: we are nowhere near being able to safely or meaningfully enhance complex traits through gene editing.
Replace conventional medicine for common diseases
For the foreseeable future, CRISPR therapies will target rare, severe, single-gene diseases where no good alternatives exist. The economics alone are prohibitive for widespread use. Casgevy’s list price is $2.2 million per patient, requires extended hospitalization and myeloablativeDescribes high-dose chemotherapy that destroys bone marrow cells to make room for a stem cell transplant. Required before CRISPR therapies like Casgevy. chemotherapy, and demands highly specialized medical infrastructure.
The Access Problem
Even where CRISPR works, the question of who gets it remains unanswered. Sickle cell disease disproportionately affects Black and Hispanic communities. Many patients are covered by Medicaid. Very few people with sickle cell disease have received gene therapy, largely because of cost and the complexity of treatment.
The Centers for Medicare and Medicaid Services (CMS) created an outcomes-based payment model in 2024 to try to bridge this gap, with over 30 states participating. But a $2.2 million therapy that requires chemotherapy and months of follow-up is not going to reach most of the estimated 100,000 Americans with sickle cell disease anytime soon, let alone the millions of patients in sub-Saharan Africa where the disease burden is highest.
Meanwhile, US science funding cuts in 2025 brought research funding to its lowest level in decades, with NSF biology funding halved and proposed 40% cuts to the NIH budget. The pipeline of new CRISPR therapies depends directly on this funding.
The Honest Scorecard
CRISPR is a genuinely transformative tool. It has produced the first functional cures for sickle cell disease and beta thalassemia. It is enabling liver-targeted treatments with single-dose, long-lasting effects. The first personalized therapy was designed and delivered in months, not decades.
But it cannot yet reach most organs. It cannot reliably make precise corrections. It cannot address complex diseases caused by many genes. It has known safety risks that are still being characterized. It costs millions per patient. And the gap between a successful clinical trial and equitable global access remains enormous.
The technology is real. The cures are real. The hype about what comes next is where the trouble starts.
In December 2023, Vertex Pharmaceuticals’ Casgevy (exagamglogene autotemcel) became the first CRISPR/Cas9-based therapeutic approved by the FDA, indicated for sickle cell disease (SCD) and transfusion-dependent beta thalassemia (TDT) in patients aged 12 and older. By mid-2025, a team at the Innovative Genomics Institute had designed, received FDA clearance for, and administered a bespoke in vivo CRISPR therapy for an infant with CPS1 deficiency, all within six months. These milestones are genuine. But the mechanistic constraints, delivery bottlenecks, and genotoxic risks that define the field’s actual boundaries receive far less attention than the breakthroughs.
This article examines those boundaries in detail: what CRISPR-based genome editing can mechanistically accomplish in living organisms today, where the technical barriers lie, and what the current evidence says about safety.
Mechanism and Repair Pathway Constraints
CRISPR/Cas9 introduces a double-strand break (DSB) at a guide RNA-specified locus. The therapeutic outcome depends entirely on which DNA repair pathway the cell activates. Non-homologous end joining (NHEJ), the dominant pathway in human cells, is error-prone: it ligates the broken ends directly, frequently introducing small insertions or deletions (indels). This is useful for gene knockout but unsuitable for precise correction.
Homology-directed repair (HDR), the pathway needed for precise sequence replacement, is inherently less efficient than NHEJ and occurs only during the late S and G2 phases of the cell cycle. This severely limits HDR-mediated editing in post-mitotic or slowly dividing cells, including neurons and cardiomyocytes, which are primary therapeutic targets for neurodegenerative and cardiac diseases.
This asymmetry explains a pattern in clinical results: nearly all successful CRISPR therapies to date work by breaking genes, not repairing them. Casgevy disrupts a BCL11A erythroid enhancer to derepress fetal hemoglobin. Intellia’s hATTR treatment (nexiguran ziclumeran) knocks out the TTR gene in hepatocytes. The goal is loss of function, not gain of function or correction.
Delivery: The Liver Monopoly
For in vivo delivery, the field has converged on lipid nanoparticlesMicroscopic fat particles used to package and deliver genetic material into cells. The main vehicle for in vivo CRISPR delivery, with a natural affinity for liver cells. (LNPs) as the leading non-viral vector. LNPs encapsulate CRISPR ribonucleoprotein or mRNA/sgRNA cargo in lipid shells that are taken up by cells via endocytosis. The problem: LNPs have a natural affinity for the liver, accumulating in hepatocytes after systemic (IV) administration. This makes them excellent for liver targets but useless for most other organs.
Intellia’s Phase I data for hATTR illustrates the strength of this approach within its narrow domain: participants showed an average ~90% reduction in serum TTR protein, sustained through 24+ months of follow-up, with all 27 participants reaching the two-year mark maintaining the response. Three participants received a second dose at a higher level, marking the first-ever redosing of an in vivo CRISPR therapy, possible precisely because LNPs, unlike AAV viral vectors, do not trigger neutralizing antibody responses that preclude re-administration.
Extrahepatic LNP delivery is an active research area. Lung-tropic formulations have shown efficient editing in endothelial and epithelial cells in preclinical models. But no extrahepatic LNP-delivered CRISPR therapy has entered clinical trials. For AAV-based delivery, the packaging capacity (~4.7 kb) is too small for standard SpCas9 (~4.2 kb with guide RNA), necessitating split-intein dual-vector strategies or smaller Cas orthologs. And AAV carries its own risks: the first death in a CRISPR clinical trial was a Duchenne muscular dystrophy patient who developed acute respiratory distress syndrome from an immune response to the AAV6 delivery vector.
Beyond Indels: The Structural Variation Problem
Off-target mutagenesis, where Cas9 cuts at unintended genomic sites with sequence similarity to the target, is the most discussed safety concern. But a 2025 Nature Communications review argues that on-target genomic aberrations deserve equal attention. These include:
- Kilobase- to megabase-scale deletions at the cut site
- Chromosomal arm losses and truncations
- Translocations between the target chromosome and off-target sites
- Chromothripsis (catastrophic chromosomal shattering and reassembly)
Critically, standard short-read amplicon sequencing cannot detect these large structural variations because they delete the primer-binding sites, rendering them invisible to the analysis. This means that reported editing efficiencies may systematically overestimate HDR rates and underestimate the true frequency of harmful genomic rearrangements.
The problem is compounded by attempts to improve editing precision. DNA-PKcs inhibitors, widely used to suppress NHEJ and promote HDR, were found to increase structural variation frequency by up to a thousand-fold, including chromosomal translocations at off-target sites. Even high-fidelity Cas9 variants and paired nickase strategies, while reducing off-target indels, still introduce substantial on-target structural aberrations.
For ex vivo therapies like Casgevy, post-editing quality control provides a safety buffer: cells can be characterized before transplantation. For in vivo therapies delivered systemically, no such checkpoint exists. This asymmetry in safety assurance is a fundamental constraint on the expansion of in vivo CRISPR therapeutics.
Base Editing and Prime Editing: Better, Not Solved
Base editors (CBEs and ABEs) chemically convert single nucleotides without introducing DSBs, instead using a Cas9 nickase fused to a deaminase. Prime editors use a nickase-reverse transcriptase fusion guided by a pegRNA to write new sequences at the nick site. Both avoid the double-strand break that triggers NHEJ-mediated indels and large structural variations.
Clinical validation is beginning. Prime Medicine’s PM359, the first prime-editing therapy to enter the clinic, corrected the delGT mutation in NCF1 in a patient with chronic granulomatous disease (CGD). NADPH oxidase activity was restored in 66% of neutrophils by Day 30, well above the 20% threshold considered clinically meaningful. Engraftment was roughly twice as fast as reported for approved CRISPR-Cas9 therapies.
On the precision engineering front, MIT researchers published results in Nature (October 2025) showing that engineered Cas9 mutations in their prime editor (vPE) reduced unintended byproducts by up to 60-fold compared to standard prime editing. In specific editing modes, error rates dropped from one per seven edits to one per 101, and from one per 122 to one per 543.
However, these approaches do not fully solve the safety problem. Nick-based platforms, including base editors and prime editors, may lower but do not eliminate structural variations. The cargo size of base editors and prime editors also exceeds AAV packaging capacity, creating the same delivery constraints that limit standard CRISPR-Cas9.
The PolygenicDescribes a trait or disease influenced by many genes, each contributing a small effect. Most common diseases like diabetes and heart disease are polygenic. Wall
The diseases that cause the most human suffering, including heart disease, diabetes, cancer, and psychiatric disorders, are polygenic: influenced by hundreds or thousands of genetic variants, each with a tiny effect sizeA standardized measure of the magnitude of difference between groups in a study, independent of sample size., interacting with environmental and lifestyle factors.
A 2024 Nature analysis modeled the theoretical consequences of heritable polygenic editing (HPE). The results were striking: editing just 10 variants associated with coronary artery disease was predicted to reduce lifetime prevalence from 6% to 0.1% among edited genomes. Editing 40 variants could reduce lifetime risk of Alzheimer’s disease, schizophrenia, type 2 diabetes, and CAD to under 0.2%.
But the same paper is explicit about the constraints. It is not currently possible to target hundreds or thousands of polymorphisms simultaneously. Very few causal variants for common diseases are known with certainty, because GWAS hits are associations, not mechanisms. Pleiotropy means that variants protective against one disease may increase risk for another. And the authors estimate that multiplex editing of polygenic traits is roughly one human generation (about 30 years) away from technical feasibility, with its desirability “highly uncertain.”
For the foreseeable future, CRISPR therapeutics will remain confined to monogenic or oligogenic targets where the genetic architecture is well-characterized and a single edit produces a large phenotypic effect.
Economics and Access
Casgevy’s list price is $2.2 million. The competing gene therapy for SCD, Lyfgenia (lentiviral, not CRISPR-based), costs $3.1 million. Both require myeloablativeDescribes high-dose chemotherapy that destroys bone marrow cells to make room for a stem cell transplant. Required before CRISPR therapies like Casgevy. conditioning, extended hospitalization, and specialized medical infrastructure that exists at only a handful of centers.
Very few SCD patients have received these therapies since approval. Between 50% and 60% of Americans with SCD are enrolled in Medicaid. CMS launched an outcomes-based payment model in 2024 with over 30 participating states, tying manufacturer rebates to treatment effectiveness. But the structural barriers, cost, complexity, geographic concentration of treatment centers, remain formidable.
The research pipeline faces its own economic headwinds. Reduced venture capital investment has led to significant layoffs in CRISPR-focused companies, with firms narrowing pipelines to prioritize near-term commercialization over broader therapeutic development. And 2025 US science funding cuts brought NSF biology research funding to half its prior level, with proposed 40% cuts to the NIH budget threatening the basic research that feeds clinical translation.
Where the Field Actually Stands
The honest assessment: CRISPR-Cas9 is a validated therapeutic platform for ex vivo editing of hematopoietic stem cells (sickle cell, beta thalassemia) and in vivo knockout of hepatocyte-expressed genes (hATTR, HAE). Base editing and prime editing are entering early clinical trials with promising but preliminary results. Delivery beyond the liver remains preclinical. Precise in vivo correction of most disease-causing mutations is unsolved. Polygenic editing is theoretical. And the structural variation risks of DSB-inducing editors are more complex than previously appreciated.
The technology is powerful, specific, and improving. It is not general-purpose, not risk-free, and not accessible to most of the patients who need it. The gap between what the tool can do and what the headlines promise is where scientific literacy matters most.
