Cystic fibrosis: one broken channel, and the drugs that learned to fold it

Start here: what cystic fibrosis is

Cystic fibrosis is caused by mutations in a single gene on chromosome 7, CFTR, which encodes the cystic fibrosis transmembrane conductance regulator — a protein-kinase-A-activated epithelial anion channel that moves chloride across the surface of secretory epithelia, and with it salt and water, in the lung, pancreas and other organs ^{[1] [2]}. It is autosomal recessive: two faulty copies are needed, one from each parent.

When the channel does not work, the thin watery layer that normally sits on the airway surface is not maintained. Mucus becomes thick and sticky, cilia cannot clear it, and the lungs are colonised by bacteria that are then almost impossible to eradicate. The same failure blocks the pancreatic ducts, so digestive enzymes never reach the gut — exocrine pancreatic insufficiency — which is why untreated infants fail to grow. Sweat glands, which normally reabsorb chloride on its way to the skin, cannot; the sweat stays salty, and that is the basis of the diagnostic test.

The mechanism is not one defect but several. Most CF mutations either reduce the number of CFTR channels reaching the cell surface — synthesis or processing mutations — or impair the function of the channels that do get there — gating or conductance mutations — or both ^[1]. More than 2,000 CFTR mutations have been described, conferring a range of molecular and functional phenotypes ^[3]. The most common by far is F508del (also written ΔF508 or Phe508del), a deletion that causes the protein to misfold and be destroyed in the endoplasmic reticulum, so almost none reaches the membrane ^[2]. A worldwide analysis of 72,431 CF chromosomes across more than 80 regions confirmed wide mutational heterogeneity but found a significant positive correlation between ΔF508 frequency and CF incidence in regional populations ^[4].

Who gets it. CF is the most common lethal autosomal recessive disorder among people of European ancestry ^[5], though it occurs worldwide and the mutation spectrum differs by ancestry ^[4].

Why it is in the news, and why this review exists. For most of its history CF was a paediatric disease — the literature reaches back to its description in 1938 ^[6]. Care improved steadily and undramatically for decades: airway clearance, antibiotics, enzymes, nutrition. Then something different happened. Drugs were developed that act on the defective protein itself, and the disease's trajectory changed. Median survival is now estimated at approximately 50 years, and CF is no longer an exclusively paediatric disease ^[7].

That is the rarest kind of story in medicine: a single-gene disease where knowing the gene actually led to a drug that fixes what the gene got wrong. This review is about how that happened, what it did and did not solve, and the equity problem it left behind.

Three pillars follow — measurements, medicines, and progress — with a model between the first two that explains the whole thing in one number.

Pillar 1: measurements and diagnosis

Newborn screening

Most children with CF in high-income countries are now identified before they are ill. Screening begins with immunoreactive trypsinogen (IRT) measured on the newborn blood spot — a pancreatic enzyme precursor that leaks into the blood when the pancreatic ducts are obstructed — followed by DNA analysis in a two-tier design, with recommendations that multiple CFTR alleles be included to increase sensitivity across populations ^{[4] [8]}.

Screening is not merely earlier labelling. Neonatal screening prevents severe malnutrition and improves long-term growth, because pancreatic insufficiency can be treated from the start rather than after months of failure to thrive ^[9].

Screening also produced a new category the field had to name. Newborns with a high IRT but inconclusive functional and genetic testing are designated CFTR-related metabolic syndrome / CF screen positive, inconclusive diagnosis — terms now merged and treated as equivalent ^[8]. These are children who may never develop CF, and telling them apart is an ongoing problem.

The sweat chloride test

This is the quantitative heart of CF diagnosis, and it deserves its status. Because CFTR normally reabsorbs chloride in the sweat duct, a failing channel leaves chloride in the sweat — so sweat chloride concentration is an inverse readout of CFTR function in a living person. The CF Foundation consensus is explicit that diagnoses associated with CFTR mutations, in everyone from newborns to adults, should be established by evaluation of CFTR function with a sweat chloride test ^[8].

The thresholds are conventional and quantitative. Classic or typical CF is one or more phenotypic characteristics plus a sweat chloride greater than 60 mmol/L; a normal result is below 30 mmol/L and a borderline result is 30–60 mmol/L. Patients with non-classic or atypical CF have a CF phenotype in at least one organ system with a normal or borderline sweat chloride, and in them the diagnosis requires a disease-causing mutation on each CFTR gene or direct quantification of CFTR dysfunction by nasal potential difference. Most of those patients have exocrine pancreatic sufficiency and milder lung disease ^[10].

Note what that last sentence does: it ties a number to a phenotype. That link is the centrepiece below. Sweat measurement itself is being modernised, with wearable devices for autonomous sweat extraction and analysis ^[11].

Genotyping, and why mutation class matters

Sequencing the two CFTR alleles does two things. It confirms the diagnosis, using the mutation annotations curated by the Clinical and Functional Translation of CFTR project ^[8]. And — since 2012 — it determines which drugs a patient is eligible for, because the modulators act on specific molecular defects.

The traditional class I–VI scheme sorts mutations by whether they impair transcription, translation, folding, trafficking, gating or conductance. The field has since argued that this is too clean: mutations may impose combinatorial defects in CFTR channel biology, and it is precisely that observation that led to the conclusion that combining pharmacotherapies against single defects would beat the available low-efficacy monotherapies ^[3]. That prediction turned out to be right, and Pillar 2 is its consequence.

Genotype predicts some things well and others badly, and the classic study says so plainly. Comparing 399 compound heterozygotes with matched ΔF508 homozygotes across 14 countries, the R117H/ΔF508 genotype clearly differed — more often pancreatic sufficient (87% vs 4%), older at diagnosis (10.2 vs 2.5 years), lower sweat chloride (80 ± 18 vs 108 ± 14 mmol/L). But for none of the genotypes studied could predictions be made about common complications or the severity or course of pulmonary disease ^[5]. Genotype sets the floor; it does not tell you how the lungs will go. CFTR mutations also turn up outside classic CF — for instance in idiopathic pancreatitis ^[12].

Tracking the lungs: FEV₁

Once someone has CF, the number that follows them is percent-predicted FEV₁ — forced expiratory volume in one second, as a percentage of what a person of their age, sex and height should manage. It is the primary endpoint in essentially every CF trial ^{[13] [14] [15]}, the variable registries use to track disease modification ^[16], and the one whose long-run decline defines the disease's course ^[17].

Centerpiece: a simple simulatable model of residual CFTR function

Everything above converges on one question. Sweat chloride measures how much CFTR function a person has. How much does that function matter?

Two published measurements answer it, and because there are exactly two of them they determine a two-parameter curve with no free parameters left over. Model the probability of exocrine pancreatic sufficiency as a logistic function of sweat chloride:

P(sufficiency) = 1 / (1 + exp((SC − SC₅₀) / s))

Solving from the genotype-matched pair — 80 mmol/L → 87% sufficient, and 108 mmol/L → 4% sufficient ^[5] — gives SC₅₀ = 90.5 mmol/L and s = 5.5. Both were derived, neither was chosen.

What the model explains. Four things.

First, why a little function goes a long way. The curve is steep exactly where patients sit. Twenty-eight millimoles per litre of sweat chloride — a difference invisible to a patient, undetectable without the test — is the difference between 4% and 87% pancreatic sufficiency, and between diagnosis at two and a half years and diagnosis at ten ^[5]. The disease is not linear in the protein.

Second, why correcting a small fraction of the protein is worth doing. This is the argument that justified the whole modulator programme, and it was made in vitro before it was made in patients. VX-809 improved F508del-CFTR processing and raised chloride secretion in patient-derived bronchial epithelia to about 14% of non-CF cells — and the paper names that level as "associated with mild CF in patients with less disruptive CFTR mutations" ^[2]. Fourteen percent of normal is not a cure. On this curve it does not have to be.

Third, why the drugs' sweat-chloride numbers are the headline they are. Ivacaftor moved G551D patients 48.1 mmol/L ^[13] — 1.7 times the entire distance separating the two genotype phenotypes on the left panel. That is not a marginal biochemical effect; on this axis it is a change of disease category.

Fourth, why sweat chloride is a good surrogate at all. The same trial that reported the 48.1 mmol/L fall also reported a 10.6-point FEV₁ gain, 55% fewer exacerbations and 2.7 kg more weight, in the same patients over the same 48 weeks ^[13]. The functional readout and the clinical outcomes moved together, which is what licenses using the cheap measurement as a stand-in for the expensive one.

What the model deliberately does not do. It predicts pancreatic sufficiency, not lung disease — and the source is emphatic that genotype does not predict the pulmonary course ^[5]. The logistic shape is an assumption; two points cannot establish a functional form. Sweat chloride reflects CFTR in the sweat duct, not the airway. And it says nothing about time: CF is progressive, and a curve of phenotype-versus-function has no axis for the years of infection and inflammation that actually destroy lungs ^[18].

Pillar 2: medicines — two eras

Era 1: managing the consequences

For fifty years, everything in CF care treated downstream consequences, and it worked better than people expected.

Airway clearance and infection control. The CF airway is colonised early and progressively, above all by Pseudomonas aeruginosa; longitudinal study of young children shows the pattern of acquisition and persistence that defines the disease's course ^[19]. Once established, Pseudomonas forms biofilms that resist both antibiotics and the immune system ^[20], which is why eradication has to be attempted early and why chronic suppressive therapy became standard. A European consensus set out antibiotic strategy against Pseudomonas specifically for CF ^[21], within a broader account of the pathophysiology and management of pulmonary infection in the disease ^[22]. Non-tuberculous mycobacteria are an increasingly recognised second problem ^[23], and the airway-clearance and bronchiectasis principles overlap with non-CF bronchiectasis care ^[24].

Pancreatic enzymes and nutrition. Replacing the enzymes the blocked pancreas cannot deliver, and treating CF as a nutritional disease as much as a pulmonary one, is the other half of era-1 care — and the reason newborn screening matters ^[9].

Transplantation remains the end-stage option, and registry data track it as an outcome ^[16].

None of this touched the cause. The lung disease still progressed ^{[18] [17]}, and the emerging challenges of the era were catalogued as such ^[25].

Era 2: fixing the protein

The insight that changed things was that different mutation classes need different pharmacology ^[3], and that the two main defects call for two different kinds of molecule.

Potentiators open channels that are present but shut. Ivacaftor (VX-770) increases CFTR channel open probability. In cultured human bronchial epithelia carrying G551D on one allele and F508del on the other, it increased chloride secretion roughly ten-fold, to about 50% of that seen in non-CF epithelia, reduced excessive sodium and fluid absorption so the airway surface stayed hydrated, and increased cilia beating ^{[1] [26]}. It works across multiple gating mutations, not just G551D ^[27].

The clinical result was the proof of concept for the entire field. In patients aged 12 and over with at least one G551D allele, ivacaftor against placebo produced a 10.6 percentage-point greater gain in percent-predicted FEV₁, effects visible by two weeks and sustained through 48; 55% lower odds of a pulmonary exacerbation; 8.6 points better on the CFQ-R respiratory domain; 2.7 kg more weight; and the 48.1 mmol/L fall in sweat chloride that anchors the figure above ^[13]. Registry follow-up in the US and UK subsequently found ivacaftor-treated patients had significantly lower risks of death (0.6% vs 1.6%), transplantation (0.2% vs 1.1%), hospitalisation (27.5% vs 43.1%) and pulmonary exacerbation (27.8% vs 43.3%) than matched comparators, with better preserved lung function — observational, but consistent with disease modification rather than symptom relief ^[16].

Correctors help the protein fold and reach the membrane. That is the F508del problem, and it needed a different molecule. VX-809 (lumacaftor) improved F508del-CFTR processing in the endoplasmic reticulum with an EC₅₀ of 81 ± 19 nM, producing corrected protein with normal-like proteolytic susceptibility, membrane residence time and single-channel open probability ^[2]. It reached phase IIa in F508del homozygotes ^[28], and was combined with ivacaftor on exactly the combinatorial logic above ^{[29] [3]}.

Triple therapy. The step that transformed the field was adding a second corrector. Elexacaftor–tezacaftor–ivacaftor achieves allosteric folding correction of F508del and of rare CFTR mutants ^[30]; the phase 3 programme established efficacy in people homozygous for F508del ^[31], in those with one F508del allele plus a minimal-function allele via the VX-659 and VX-445 route ^[32], in children aged 6 to 11 ^[33], and — the trial that most cleanly isolates the added benefit — in people whose second allele was a gating or residual-function mutation already being treated with ivacaftor or tezacaftor–ivacaftor. Even in that heavily pre-treated group, triple therapy added 3.5 percentage points of FEV₁ (95% CI 2.2–4.7) and lowered sweat chloride by a further 23.1 mmol/L (95% CI 20.1–26.1) against active control, with CFQ-R respiratory scores rising 10.3 points against 1.6 ^[14].

Because F508del is on at least one allele in most patients worldwide ^[4], triple therapy made the majority of the CF population modulator-eligible at once. The precision-medicine framing followed ^[34].

The gap this leaves

Modulators require a protein to modulate. They do nothing for mutations that produce no CFTR at all — nonsense, frameshift and severe splicing alleles that stop the protein being made. Those patients, disproportionately from populations whose mutation spectra were less studied and less represented in trials ^[4], are left with era-1 care in an era-2 world. That is an equity problem, not merely a scientific one, and it is the main thing the next section is trying to solve ^[35].

Pillar 3: progress

Reaching the non-responders

If the problem is that no protein is made, the answer has to supply the gene or the message rather than chaperone the product.

Gene therapy has been attempted longest and delivered least, though not nothing. A randomised, double-blind, placebo-controlled phase 2b trial nebulised a plasmid-DNA/cationic-liposome complex monthly for a year in 140 patients with any combination of CFTR mutations, and found a significant but modest treatment effect of 3.7% in relative percent-predicted FEV₁ (95% CI 0.1–7.3, p = 0.046), reflecting stabilisation in the treated group against decline in the placebo group ^[15]. That is a real result and an honest measure of how hard airway gene delivery is. Viral vector platforms have advanced considerably since, with AAV the most developed ^{[36] [37]}, and the therapeutic-target landscape has been reviewed with these approaches in view ^[35].

Better models to test in

Two model systems changed what can be tested before a patient is dosed. Human airway organoids can now be expanded long-term for disease modelling ^[38], drawing on the airway basal stem cells that regenerate the epithelium ^[39]; the general organoid toolkit has matured alongside ^{[40] [41] [42] [43] [44]}, including pancreatic organoids relevant to the other major affected organ ^[45]. Because CFTR function can be measured directly in organoids derived from an individual patient, these systems make it possible to ask whether a specific rare mutation will respond to a modulator — which is exactly the question the trials cannot answer for alleles too rare to enrol.

The disease that changed shape

The cumulative effect is visible in population data rather than in any single trial. Incidence is falling in most countries — a consequence of carrier screening and prenatal diagnosis ^[7]. Survival has improved substantially, to a median around 50 years ^[7]. Registry evidence supports genuine disease modification rather than symptomatic benefit ^[16]. And the clinical problem has migrated from paediatrics to adult medicine, which brings its own set of complications the field is still learning to manage ^{[25] [17]}.

CFTR biology itself continues to generate spin-offs in the other direction: CFTR inhibitors, identified by high-throughput screening, block cholera-toxin-induced intestinal fluid secretion ^[46] — the same channel, the opposite problem.