Sickle cell disease: one letter of DNA, and the first CRISPR cure

Start here: what sickle cell disease is

Red blood cells are essentially bags of one protein. Haemoglobin fills them, binds oxygen in the lungs, and releases it in the tissues. Adult haemoglobin is built from four subunits — two α-globins and two β-globins — and the cell's job is to squeeze through capillaries narrower than itself, deliver oxygen, and repeat, about a hundred thousand times over a four-month life.

Sickle cell disease (SCD) starts with a single amino-acid substitution in the β-globin chain. That one change does something no other haemoglobin variant does at scale: when the mutant haemoglobin S (HbS) gives up its oxygen, the molecules stick to one another and assemble into long stiff fibres inside the cell. The cell, previously a flexible disc, becomes rigid and distorted — the "sickle" shape. From that single event the whole disease unfolds: polymerized cells jam in small vessels, and they rupture. The modern synthesis describes a vicious cycle of four processes: HbS polymerization, impaired rheology and adhesion-mediated vaso-occlusion, haemolysis-mediated endothelial dysfunction, and the resulting inflammation and redox instability, driving progressive small- and large-vessel vasculopathy ^{[1] [2] [3]}.

Clinically that means two things at once. Haemolytic anaemia — cells destroyed faster than they can be replaced. And vaso-occlusive crises — episodes of sudden, severe pain as blocked vessels starve tissue of oxygen, along with end-organ ischaemia-reperfusion injury and infarction ^{[1] [4]}.

Who it affects. The sickle mutation is not rare and not randomly distributed. Geostatistical mapping of haemoglobin surveys shows high allele frequencies across most of sub-Saharan Africa and the Middle East ^[5]. There are roughly 300,000 births a year with sickle cell anaemia, over 75% of them in Africa ^[6], and the Global Burden of Disease programme has assessed prevalence and mortality for 204 countries while noting that previous analyses, hampered by underdiagnosis and single-cause death attribution, gave "only a small insight into the suspected high population health effect" of the disease ^[7]. It remains, in the framing of its own literature, a neglected chronic disease of rising global importance ^{[8] [9] [10]}.

Why the mutation is common — the malaria story. A mutation this damaging should have been eliminated by natural selection. It was not, because carriers are protected against malaria. Someone with one sickle allele and one normal allele — genotype HbAS, "sickle cell trait" — does not have sickle cell disease but does resist Plasmodium falciparum. This is the textbook case of balancing selection, known as the malaria hypothesis, and mapping HbS allele frequencies against pre-intervention malaria endemicity confirmed the geographical relationship at global scale, strongly in Africa ^[11]. It is not history: in Gabon, a 10% increase in P. falciparum prevalence is associated with a 4.3% increase in trait carriers — "malaria remains a selective factor in current human populations" — and, through migration, sickle cell trait is now the most important genetic disorder in France, affecting one birth in every 2,400 ^[12]. Quantitatively, HbAS protection against clinical malaria in Kenyan children rose "from only 20% in the first 2 y of life to a maximum of 56% by the age of 10 y, returning thereafter to 30%" ^[13]. Sickle cell disease is, in the most literal sense, the cost of a defence.

Why it is in the news. In 2023–24 the first gene-therapy cures for a common genetic disease reached approval, one of them the first licensed CRISPR therapy in medicine. That is the subject of the third pillar.

Three pillars follow — measurements, medicines, and progress — with a simple simulatable population-genetics model between the first two that explains why this allele is still here.

Pillar 1: measurements and diagnosis

Finding it: newborn screening and haemoglobin separation

SCD is diagnosed by looking directly at which haemoglobins a person makes. Because HbS carries a different surface charge from normal HbA, the variants separate under an electric field or on a chromatography column — haemoglobin electrophoresis and HPLC are the standard methods, and HPLC is what large field studies use to genotype thousands of samples ^{[12] [14]}.

The distinctions that matter clinically:

• HbAA — two normal β-globin genes. Unaffected.
• HbAS — sickle cell trait: one sickle allele. Usually no disease, and malaria-protective ^[13].
• HbSS — sickle cell anaemia: two sickle alleles. The classic, most severe form.
• HbSC and HbS/β-thalassaemia — compound heterozygotes with a second, different β-globin variant. Generally milder than HbSS but genuinely affected — and the milder forms have distinctly better survival ^[15].

The reason to do this at birth, before any symptom, is that early identification enables the interventions that changed the disease's natural history. The United States and United Kingdom cut sickle cell anaemia mortality from 3 to 0.13 deaths per 100 person-years through newborn screening, prevention of infection, and comprehensive care ^[6]. Population newborn screening programmes for haemoglobinopathies are well described ^[14]. The consequence is visible in cohort data: in the Dallas Newborn Cohort, 93.9% of children with sickle cell anaemia and 98.4% of those with milder forms now live to become adults, sepsis is no longer the leading cause of death, and all recent deaths occurred at 18 years or older — most shortly after the transition to adult care ^[15]. The mortality gradient between settings is stark: a Tanzanian cohort of 1,725 patients found a mortality rate of 1.9 per 100 person-years, highest under age five ^[6].

Measuring what the disease is doing

Haemolysis markers. Because red cells are being destroyed, the laboratory shows anaemia plus the debris of lysis. Lactate dehydrogenase (LDH), released from ruptured cells, is more than a damage marker: it functions as a biomarker of a specific haemolysis-associated clinical subphenotype — nitric-oxide resistance with priapism, leg ulceration and pulmonary hypertension ^[16]. The mechanism is that free haemoglobin released into plasma scavenges nitric oxide, the body's principal vasodilator, so intravascular haemolysis produces endothelial dysfunction as a direct chemical consequence ^{[17] [18] [19] [20]}. That is why pulmonary hypertension is an independent risk factor for death in SCD ^[21].

The vaso-occlusive crisis is the defining clinical event and it is diagnosed clinically, not by a test. Pain in SCD has been studied as its own subject for decades ^{[4] [22]}, and the biology is now understood as more than mechanical plugging: P-selectin on activated endothelium and platelets mediates adhesion of sickle erythrocytes to the vessel wall, making the crisis an adhesion event as much as a rheological one ^{[23] [24]}. Its most dangerous form is acute chest syndrome — new pulmonary infiltrate with fever and respiratory symptoms — a leading cause of death whose causes and outcomes were catalogued by the Cooperative Study of Sickle Cell Disease ^{[25] [26] [27] [28]}.

One measurement problem is not biological. Pain in SCD is subjective, chronic, recurrent, and disproportionately affects a racialized population, and the literature documents an unequal burden of pain and disparities in its treatment ^[29]. Under-treatment of crisis pain is a recognized failure mode of care.

Transcranial Doppler: predicting stroke before it happens

The most successful screening test in SCD deserves its own paragraph, because it turned a catastrophic complication into a preventable one.

Stroke is a major cause of morbidity and mortality in children with SCD, and it happens because the large intracranial arteries narrow. Narrowing accelerates blood flow, and flow velocity can be measured non-invasively with transcranial Doppler (TCD) ultrasonography. Adams and colleagues showed prospectively that velocity in the middle cerebral artery predicted first cerebral infarction, with abnormal defined as ≥170 cm/s ^[30].

The STOP trial then closed the loop. Children with sickle cell anaemia and no stroke history whose time-averaged mean velocity in the internal carotid or middle cerebral artery was ≥200 cm/s were randomized to standard care or to chronic transfusion aimed at reducing HbS to less than 30% of total haemoglobin ^[31]. Screen, identify, transfuse, prevent. Follow-up work showed that stopping prophylactic transfusion reverses the benefit ^[32], and later trials tested hydroxycarbamide as a transfusion-sparing alternative for maintaining TCD velocities ^{[33] [34]}.

Centerpiece: a simple simulatable model of why the allele persists

A mutation that kills homozygotes should disappear. The standard one-locus model shows exactly why this one does not.

Assign relative fitnesses to the three genotypes in a malarious environment:

• w(AA) = 1 − s — normal homozygote, susceptible to malaria; s is the fitness cost malaria imposes
• w(AS) = 1 — the heterozygote, malaria-protected: the fittest genotype
• w(SS) = 1 − t — sickle cell disease; t is its fitness cost

Because the heterozygote is fittest, selection cannot drive either allele to fixation: eliminating the sickle allele destroys the protected genotype, and eliminating the normal allele produces only affected homozygotes. The population settles at an interior stable polymorphism, and the equilibrium sickle-allele frequency has a strikingly simple closed form:

q* = s / (s + t)

The equilibrium depends only on the ratio of the two costs. From it, Hardy–Weinberg gives the birth frequencies directly: q\*² of births have sickle cell disease, and 2q\(1 − q\) carry the trait.

Grounding. The model form is the textbook heterozygote-advantage case, and this locus is its canonical example — named as such in the substrate: "This traditional example of balancing selection is known as the 'malaria hypothesis'," with the global geography of HbS allele frequency confirmed against pre-intervention malaria endemicity ^[11]. That the heterozygote genuinely carries the advantage the model requires is grounded independently: HbAS protection against clinical malaria of 20% under age 2, peaking at 56% by age 10 ^[13], with selection demonstrably ongoing today ^{[12] [35] [36]}.

Parameters are illustrative and flagged. Translating a published "56% protection against clinical malaria" into a lifetime relative fitness deficit requires assumptions the source does not make. So the values of s swept in the figure are a plausible range rather than an estimate, and the three values of t are scenarios rather than measurements. What is not illustrative is the mathematics: the script verifies that the closed form is the fixed point a forward simulation of the selection recursion actually converges to, that it is stable against perturbation in both directions, that it depends only on s/t, and that it collapses to zero when the heterozygote advantage does.

What the model explains. Three things.

First, the persistence itself. With s = 0.15 and t = 1 — a modest carrier advantage against a homozygote who does not reproduce — the equilibrium allele frequency is 0.130, giving 22.7% trait carriers and 1.7% of births affected. Carrier frequencies of that order are exactly what is observed across high-malaria Africa ^{[5] [11]}. A small heterozygote advantage sustains a lot of disease.

Second, why removing malaria does not quickly remove the allele. q* is set by the ratio s/t; as malaria control drives s toward zero the equilibrium falls, but allele frequencies move on generational timescales while public health moves on annual ones. The allele is a record of past selection, which is why it travels with populations to places that never had malaria ^[12].

Third — and this is the counter-intuitive one — improving sickle cell care raises the equilibrium frequency. Lower t means a smaller fitness penalty on SS, so q rises: at s = 0.15, going from t = 1.0 to t = 0.4 moves q from 0.130 to 0.273. This is not a paradox to be explained away; it is the population-genetic mechanism behind the projection that "the global burden of sickle cell anaemia is set to rise as a consequence of improved survival in high-prevalence low- and middle-income countries and population migration to higher-income countries" ^[37]. Success at keeping patients alive increases the number of patients. That is an argument for planning capacity, not for withholding care.

Limits, honestly. This is a one-locus, constant-fitness, infinite-population, random-mating model, and every one of those assumptions is wrong in detail. Real malaria resistance is polygenic — α-thalassaemia, G6PD deficiency, blood-group variants, ATP2B4 and immune loci all contribute, and together the known variants explain only part of the heritable component ^{[35] [36]}. Fitnesses vary with malaria transmission intensity, age, and healthcare access rather than being constants. And the model says nothing about why HbAS protects — a question still under debate, with evidence that protection involves accelerated acquisition of malaria-specific immunity and not only innate mechanisms ^[13].

Pillar 2: medicines

Therapy in SCD has escalated through three tiers, and the logic of all of them traces back to one quantity.

The kinetic target: delay time

HbS polymerization does not begin the instant oxygen leaves. There is a latency — the delay time — followed by abrupt fibre formation, and the delay is "exceedingly concentration dependent," well described by a double-nucleation mechanism ^{[38] [39]}. That observation produced the kinetic hypothesis: what matters is the race between the delay time and the capillary transit time. If a red cell gets through the capillary and back to the lungs before polymer forms, nothing happens. In small volumes such as a single erythrocyte the delay acquires a stochastic component from random waiting times for the first nucleus, which lengthens the average delay and "adds further protection from vaso-occlusion" ^[38].

Every disease-modifying therapy in this section works by lengthening that delay or preventing the downstream consequences when it is too short ^[40].

Fetal haemoglobin is the body's own answer. HbF (α₂γ₂) does not participate well in the HbS polymer. Classic biophysics showed that adding HbF prolongs the delay time far more strongly than adding normal HbA does, and that below about 20% HbF content "almost no HbF was incorporated into the gel phase" — it dilutes the polymerizable pool and interrupts fibre assembly ^[41]. Raise HbF and you lengthen the delay. That single sentence explains hydroxyurea, and it explains the gene therapies.

Tier 1: hydroxyurea

Hydroxyurea (hydroxycarbamide) is a ribonucleotide reductase inhibitor ^[42] repurposed as an HbF inducer; induction appears to run partly through nitric-oxide-dependent activation of soluble guanylyl cyclase ^{[43] [44]}, with response varying between individuals ^[45].

The Multicenter Study of Hydroxyurea established the clinical effect in adults with three or more crises a year: median crises fell from 4.5 to 2.5 per year, time to first crisis lengthened from 1.5 to 3.0 months, and acute chest syndrome occurred in 25 versus 51 patients ^[46]. Longer follow-up showed effects on mortality and morbidity ^{[47] [48] [49]}, the BABY HUG trial extended use to very young children ^[50], paediatric safety was established ^[51], and a 17.5-year follow-up characterized the long-term risk-benefit balance ^{[52] [53]}. Hydroxyurea is cheap, oral, and remains the backbone of care worldwide — which matters most where the disease is most common ^{[9] [54]}.

Tier 2: three targeted agents

All three were approved on the strength of trials in the substrate, and all three are modest rather than transformative.

• Voxelotor binds haemoglobin and increases its oxygen affinity, so less HbS is in the deoxygenated, polymerization-competent state — a direct anti-polymerization strategy. In a phase 3 trial, 51% of the 1500 mg group achieved a haemoglobin rise of more than 1.0 g/dL at week 24 versus 7% on placebo ^[55]. It raises haemoglobin; whether that translates into fewer crises is the question the trial design leaves open.
• Crizanlizumab is a monoclonal antibody against P-selectin, blocking the adhesion step of vaso-occlusion rather than the polymer ^[23]. In the phase 2 SUSTAIN trial, high-dose crizanlizumab reduced the annual rate of sickle-cell-related pain crises versus placebo, in patients with and without concomitant hydroxyurea ^{[56] [57]}.
• L-glutamine targets oxidative stress, increasing the reduced form of nicotinamide adenine dinucleotides in sickle erythrocytes. In phase 3, median pain crises were 3.0 versus 4.0 on placebo ^[58].

Taken together these represent real but incremental gains on three different points of the cascade ^[59].

Tier 3: cure — transplantation, and then gene therapy

Allogeneic haematopoietic stem-cell transplantation has been curative for decades, and its limit has always been donors and toxicity. In 87 consecutive patients transplanted in France from matched siblings after myeloablative conditioning — cerebral vasculopathy the principal indication — overall survival was 93.1% and event-free survival 86.1% at a median 6 years, with graft-versus-host disease the main cause of transplant-related mortality ^[60]. International survey data and guidelines confirm the picture ^{[61] [62] [63] [64]}, and reduced-intensity conditioning extended eligibility ^[65]. But most patients have no matched sibling.

Gene therapy removes the donor problem by editing the patient's own cells. Two routes reached patients.

The lentiviral route adds a working, anti-sickling β-globin gene to autologous stem cells using a viral vector. The first patient treated for SCD this way was reported in 2017 ^[66], following the same platform's use in transfusion-dependent β-thalassaemia ^[67], and it became lovotibeglogene autotemcel (lovo-cel) ^{[68] [69]}.

The CRISPR route does something more elegant: it does not fix the mutation at all. Instead it reactivates the fetal haemoglobin the patient already has the genes for. BCL11A is the transcription factor that switches γ-globin off after birth, acting through long-range interactions with SOX6 and corepressor complexes ^{[70] [71]}. Knock it out everywhere and you damage stem cells — ubiquitous knockdown "profoundly impaired long-term engraftment" — but knock it down only in the erythroid lineage, by disrupting its erythroid-specific enhancer, and you get stable engraftment plus a large HbF increase, reversing the sickle phenotype ^{[72] [73] [74]}.

That is precisely what exagamglogene autotemcel (exa-cel) does. In the first report, CD34+ cells were electroporated with CRISPR-Cas9 targeting the BCL11A erythroid-specific enhancer; roughly 80% of alleles were modified with no evidence of off-target editing, and after myeloablation the two treated patients — one with β-thalassaemia, one with SCD — showed high levels of editing more than a year later, pancellular increases in fetal haemoglobin, transfusion independence, and, in the SCD patient, elimination of vaso-occlusive episodes ^[75]. Pancellular matters: HbF spread evenly across all red cells protects every cell, whereas the same average concentrated in a few cells leaves the rest to sickle.

The phase 3 trial enrolled patients aged 12 to 35 with at least two severe vaso-occlusive crises in each of the two prior years; the primary endpoint was freedom from severe vaso-occlusive crises for at least 12 consecutive months ^{[76] [77] [78]}, with the parallel programme in transfusion-dependent β-thalassaemia ^{[79] [80]} and reported gains in health-related quality of life ^{[81] [82]}.

The honest caveats belong here rather than in a footnote. The procedure still requires myeloablative conditioning with busulfan ^[76] — the same chemotherapy that makes transplant hazardous, with its own infertility and malignancy risks. And CRISPR editing is not consequence-free at the genome level: large on-target rearrangements including megabase-scale chromosomal truncations have been documented, which is why high-fidelity Cas9 delivered as ribonucleoprotein and careful off-target assessment are part of the platform ^{[83] [84] [85] [86] [87]}.

Pillar 3: progress — and the access problem

The scientific milestone is real. SCD was the first disease for which the causal molecular lesion was identified, and it has now become the first common genetic disease with an approved CRISPR-based cure ^{[75] [76]}. The strategy — do not repair the broken gene, reactivate the healthy paralogue the body switched off — is a template other genetic diseases will borrow ^{[87] [73]}.

The access problem is equally real, and it is the dominant open question. Consider the geography. Over 75% of affected births are in Africa ^[6]; the therapy requires apheresis, a GMP editing facility, myeloablative conditioning, and weeks of inpatient transplant-grade care ^[76]. Cost-effectiveness analyses of these therapies have been framed explicitly in distributional and equity-enhancing terms, precisely because the population affected has historically been under-served ^{[88] [89]}, and readiness for gene therapy is being studied as a psychosocial as well as a medical question ^[90]. The Lancet Haematology Commission on defining global strategies to improve outcomes in SCD is the field's attempt to hold both facts at once ^{[10] [91] [9]}.

The practical consequence is that the cheap interventions matter more, not less. Newborn screening, penicillin prophylaxis, TCD screening, and hydroxyurea are what moved mortality from 3 to 0.13 per 100 person-years in high-income settings ^{[6] [15]}, and they remain unevenly implemented where the disease is concentrated ^{[9] [54] [8]}. A cure available to a few thousand people does not change the global burden; hydroxyurea and screening available to millions would.

And the burden is projected to grow — for the reason the model above makes explicit. Improved survival in high-prevalence countries plus migration to higher-income ones means more people living with SCD, not fewer ^{[37] [5] [7]}.