Trying to Cure My Cystic Fibrosis

An old European folk saying goes: “Woe to the child who tastes salty from a kiss on the brow, for he is cursed and soon will die.” The observation that salty sweat meant death in infancy more or less held true for much of human history.

The reason was not understood until 1989, when researchers identified the cystic fibrosis transmembrane conductance regulator (CFTR) gene as the cause of cystic fibrosis (CF). This single gene encodes the CFTR protein, an anion channel that lets negatively charged ions, including chloride, in and out of cells. When mutations destroy this channeling function, ion levels throughout the body become unbalanced, and both sodium and chloride stay in the sweat, hence the folk saying. That particular feature is less harmful than it is odd: after a long run on a hot day, I am left covered in salt crystals as my sweat evaporates. I didn’t realize how unusual this was until my wife pointed it out.

More damaging is the lack of CFTR in pancreatic ducts, the little tubes that carry enzymes from the pancreas into the intestines to help digest food. For people with CF, the pancreatic duct fluid becomes too thick and sticky for these enzymes, and infants with the disease struggle to absorb the nutrients from food. Fatal if untreated, this can be managed over one’s lifetime by taking pills containing purified pancreatic enzymes.

Lung mucus proves even more of a problem.

When we breathe, there is a decent chance of bacteria, viruses and pollutants entering the lung. Normal defense is mucociliary clearance: designated lung cells secrete proteins called mucins that form a layer of mucus, and anything breathed in usually gets trapped there. Meanwhile, ciliated lung cells constantly beat their cilia in synchrony to move the mucus layer up toward the throat, where it can be swallowed. The result is that inhaled bacteria are cleared from the lung before doing much damage.

But mucus must be precisely calibrated: too liquid and it won’t trap particles, too solid and it becomes hard for the cilia to clear. Its viscosity is tuned by water content, itself controlled via osmosis by the amount of salt secreted into the mucus. And salt secretion into mucus is dependent on CFTR.

And so, the absence of CFTR causes a dangerous cascade. The mucus becomes thick and can’t be cleared; the bacteria sticks around in the mucus and turns into chronic infection, which can be difficult or impossible to eradicate with antibiotics. This can lead to permanent lung damage, followed by total lung failure and death.

Antibiotics help fight this cascade, as does inhaling hypertonic saline to replace the missing salt, and inhaling a DNA-degrading enzyme;¹ all of these are aided by “high-frequency chest wall oscillations,” wherein a special vest is worn that vibrates, mechanically loosening mucus and allowing it to be cleared. These treatments helped the life expectancy for CF rise from 10 years to 40. They are why I am alive—born in 1992 and diagnosed with CF shortly thereafter, I would not be writing this article without them.

But still, these therapies only delay the inevitably fatal loss of lung function. Much better would be restoring CFTR function itself, which would stop disease progression at its root.

Modulating CFTR

Vertex Pharmaceuticals was founded in 1989 on a radical idea: what if we could develop drugs on purpose? At the time, drug discovery often leaned on serendipity, opportunism, and brute-force screening of chemical compounds; penicillin was discovered by accident (ish),² many of the first cancer drugs were poisons that killed cancer cells only marginally faster than non-cancer cells, and the diabetes drug metformin was inspired by plants used in traditional or “folk” medicine. Vertex’s founders wanted to make drugs in a rational way, so they built the company around X-ray crystallography, which measures the three-dimensional structure of proteins.

Protein activity is based on chemical composition and a protein’s shape, as many proteins fold to create a small pocket called an “active site,” where biochemical action occurs. Vertex leadership believed that knowing a protein’s structure would enable them to design molecules that could block its active site, thus preventing that particular protein from working.

This mostly didn’t go as planned. While Vertex had a few early successes with antivirals, those drugs were quickly overtaken by competitors. The company lost money on net for its first decade because it turned out that having a known protein structure is only a small part of making a good drug; the drug’s toxicity and stability in the body are just as important.

Then in 2001, after 12 years of rejecting structure-blind drug screens, Vertex bought Aurora Biosciences, a company building them.

When the companies ran a screen for the CFTR protein, they found molecules that could bind to the protein and modulate its activity. Most people with CF make plenty of CFTR protein, but these copies are broken. The molecules that Vertex discovered could tweak these broken proteins just enough to render them functional.

Vertex turned a hit from that screen into the drug ivacaftor. This drug worked for only one particular variant of CFTR—a mutant called G551D³ found in about 5% of CF patients—but it was a groundbreaking proof of concept. It was also the first time CF could be treated at its root cause.

Vertex didn’t stop there, either. It next created Orkambi, which stacked ivacaftor on top of a second drug, lumacaftor, that could treat CF patients with copies of the most common mutation, ΔF508.⁴ Then, the company won approval in 2019 for a mixture of three drugs, which they called Trikafta. (The three drugs are elexacaftor and tezacaftor, which help the CFTR protein fold and reach the cell’s surface, and ivacaftor, which helps the protein admit anions.)

It’s impossible to overstate how great Trikafta is. People taking it often describe something called “The Purge”: hours of coughing up all the built-up mucus in their lungs. People born with CF today who are treated with Trikafta have a near-normal lifespan and lack the ancient “curse of a salty brow.”⁵

Expected lifespan with cystic fibrosis has more than doubled

Based on forty years of data from the Canadian Cystic Fibrosis Registry, which tracks the health of people with the disease.

Median predicted survival 95% confidence interval

I have mixed feelings, though, about the CFTR modulator era. Vertex’s achievements are amazing, but 5–10% of CF sufferers have essentially no CFTR protein because their form of the disease is caused by “nonsense” or “frameshift” mutations in the CFTR gene, meaning not even broken copies of the protein get made. And without a broken protein, the modulator drugs have nothing to act upon.⁶

Belonging to this latter group, perhaps I can be forgiven for my frustration at the general triumphalist narrative about CF. And this is why, in my own research, I focus on gene therapy.

Viruses for Good

Immediately after the CFTR discovery, and before anyone had heard of modulator drugs, the main therapeutic idea was gene therapy, which would put a correct version of the CFTR gene into lung cells. The concept seemed so simple that I was informed in elementary school that CF would be cured within five years. (I remember being annoyed that I was going to have to continue breathing treatments until high school, which seemed like forever.) The first trials began in 1993, but even after decades of work, there still is no viable gene therapy cure.

Most gene therapy efforts to date have tried engineering viruses to carry a “healthy” copy of the CFTR gene and transport them into lung cells. But mucociliary clearance is specifically meant to protect against inhaling viruses, and the CF mucus layer is so thick and sticky that penetration is difficult. Many of these engineered viruses are likely to get trapped in mucus without even touching a lung cell.⁷

Another problem is that the CFTR protein isn’t spread evenly in the airways. Ciliated cells make little CFTR, secretory cells make more, and a rare type of cell—called a pulmonary ionocyte—makes massive amounts. Delivering a healthy copy of CFTR to abundant and accessible ciliated cells, then, is probably an inefficient way to fix the lungs.⁸ An alternative is gene editing, where the native CFTR genes in the patient’s own DNA are directly fixed, preserving the native ionocyte- and secretory cell-centric CFTR expression pattern.

Whether gene therapy or gene editing, any viral vector-based strategy must deal with the body’s strong immune response to viruses, which makes dosing multiple times fraught—essentially, the first dose vaccinates the patient against future doses. This means that therapies need to last for as long as possible, to minimize the number of doses needed over a lifetime. Standard gene therapy using healthy, non-integrating CFTR delivery has an intrinsically limited duration, because the CFTR genes are diluted away as the lung refreshes itself over a months-long timescale.

Gene editing is stable as cells divide, so in principle it could last longer, but only by solving an even more difficult delivery problem than the one I just described: the editing would have to target basal cells. These are stem cells that churn out new cells to restore the airway surface over time, but they are buried beneath an outer cell layer, which a delivery vehicle must cross to reach—difficult because the outer cell layer has “tight junctions” that restrict even the passage of water.⁹

Nor is it clear that editing the CFTR gene in standard basal cells would be sufficient to cure the disease. The airway has “hillock” structures that contain a special class of damage-resistant basal cells. These hillocks can withstand acid, cigarette smoke and influenza, and they repopulate the rest of the airway after lung damage. So it’s well within the realm of plausibility that a gene editing therapy does its job everywhere but in hillocks, the patient goes through The Purge and seems to have a new lease on life, but then gets a severe case of the flu, after which the lung’s airway regrows from unedited, diseased hillocks. Then we’re back to square one.

Given these challenges, it’s no surprise we are without gene therapy for CF. Adenoviruses were tried for most of the 1990s before being abandoned. Inhaled lentivirus was in development for decades and entered the clinic in 2025, but Boehringer Ingelheim, the company behind it, cancelled its development in February 2026. Adeno-associated viruses (AAVs), already used for nine FDA-approved therapies, were tried in the 2000s before being abandoned.¹⁰ However, sometimes such therapies are picked up again.

AAVs reappeared in 2023, when 4D Molecular Therapeutics (4DMT) used a technology called “in vivo directed evolution” to identify AAVs that could overcome barriers to inhaled delivery. It has reported somewhat ambiguous clinical data, wherein the company is claiming robust CFTR delivery to lung cells in an initial trial, but has only reported a modest effect on lung function. (4DMT did not respond to requests for comment.)

While, as far as I can tell, the 4DMT treatment offers nothing like the Trikafta Purge, “ambiguous” is a win compared to what came before. 4DMT is gearing up to dose patients a second time, which will be a critical step for this potential therapy.

Lipids for Lungs

Lipid nanoparticles, or LNPs, offer a way around some of the problems with viral gene therapy. They lack the viral proteins that drive immune responses, have more flexible size limits, and are easier to manufacture. But they still have to get through the lung’s defenses.

Already used in Onpattro, a siRNA medication for a genetic form of amyloidosis, and the mRNA vaccines for COVID-19, LNPs are balls of lipids packed with RNA. The lipids protect the RNA from degradation and help it enter cells, where it becomes trapped inside compartments called endosomes. Once the LNP escapes, the RNA can drive expression of CFTR protein directly, or trigger cells to make gene editors that fix the broken CFTR gene.

To work, the LNP must disrupt endosomes but not the protective cell membrane. This is tricky to do, though, because endosomal membranes and cell membranes are similar. To achieve this, LNPs (and viruses) take advantage of an endosome’s acidity; the fluid outside cells is relatively pH-neutral, so the lower pH inside an endosome acts as a chemical cue to activate membrane disruption after the LNP has been taken up by a cell. While viruses have evolved protein machinery that uses this acidification to trigger endosomal escape, LNPs use a special kind of lipid—called an “ionizable lipid”—embedded in their shell to do the same.

These ionizable lipids have been the subject of most LNP research and were used in the Moderna and Pfizer/BioNTech’s COVID vaccines, as well as Onpattro. Still, success in the liver or in vaccines does not guarantee success in the lung. Inhaled LNPs still face mucus, limited uptake by airway cells, and the tight junctions that restrict access to basal cells. The LNPs used in Moderna’s COVID vaccine are nowhere near potent enough for inhaled gene therapy. Instead, we need lipids suited to the lung.

When I joined Daniel Anderson’s lab as a postdoctoral researcher at MIT, I wanted to make exactly this type of LNP. I used air-liquid interface (ALI) cultures derived from human cells to do so; this is a type of cell culture that exposes airway cells to air, just as they are when in a lung, helping them develop lung-like anatomy.

I worked with Idris Raji, an organic chemist who made dozens of new ionizable lipids, and screened them for RNA delivery to ALI cultures. The best LNPs worked in both the cell cultures and mice, particularly when nebulized using stabilization strategies developed by my collaborator Allen Jiang. But I did not believe they were potent enough for clinical use.

To narrow down which structures to try next, I trained an AI model on published data linking lipid structure to LNP activity. Idris synthesized its predicted winners, and I tested them. The resulting lipids were decent, but when I added a branched lipid tail containing an ester bond to them, which rational design principles suggested would improve tolerability and activity, the lipids became much better at delivering RNA to lung cells. Interestingly, the AI model hated this decision, but I ignored it, recognizing that because its training data had few direct comparisons that could isolate the effect of ester bonds, its predictions would be brittle—whether it loved or hated esters, the preference was meaningless.

The resulting lipids, FO-32 and FO-35, performed as well as or better than every control we tested in mice; in particular, FO-32 was two- or three-fold better. This experience of ignoring AI’s recommendation is the source of my personal motto, which I remind myself of whenever I’m tempted to blindly trust AI output: there’s no substitute for thinking!

Mouse lungs differ considerably from ours, though, so I next worked with John Engelhardt, an expert in ferret lung biology, to test the lipids in ferrets, whose airways more closely resemble our own.

Happily, FO-32 and FO-35 were the first LNPs ever published to work well in ferret conducting airways, with performance comparable to that of at least one clinical-stage inhaled LNP. I am working on improving them in my research group, now based at Case Western Reserve University in Ohio.¹¹ But getting RNA into an airway cell is only part of making a therapy that human patients can live with.

Into the Clinic

In 2018, Translate Bio began testing an inhaled gene therapy for CF in patients, using LNPs to carry CFTR mRNA into airway cells so they could make working CFTR protein. Arcturus Therapeutics, a Vertex/Moderna collaboration, and ReCode Therapeutics followed with their own inhaled LNP therapies. None of these trials have succeeded to date, but the reasons for failure are, as far as I can tell, potentially enlightening.

The Vertex/Moderna trial was the most interesting: it was paused because many patients had nasty side effects, then restarted with a modified protocol, and then cancelled again for the same side effects. My guess is that the company was probably seeing signs of efficacy, but also lung inflammation. Therefore, they paused the trial to figure out whether they could alleviate the inflammation (possibly with a lower dose or more anti-inflammatory medicines), decided maybe they could, restarted the trial, got the same results, and then admitted defeat.

What’s interesting is that inflammation is generally table stakes for a gene therapy; what’s a little inflammation next to having your fatal disease fixed? The problem is that bad side effects are one thing for a one-time dose, and quite another when a treatment has to be given repeatedly. These trials delivered CFTR mRNA, which lasts only briefly inside cells, so patients need repeated doses of LNPs to keep making the protein. One way to mitigate those effects, then, could be to instead deliver DNA, which is highly stable and can last for months or years inside of cells.

Nonviral DNA delivery adds yet another layer of difficulty: whereas mRNA can start working the moment it escapes from the endosome, DNA must get into the nucleus. Viruses have evolved many mechanisms for arriving there, but LNPs have no such mechanism, so DNA delivery tends to be thousands of times less potent than RNA delivery.

Fortunately, there are promising signs that a DNA-based gene therapy for cystic fibrosis may be on the horizon. Poseida Therapeutics and Engage Bio, recently acquired by Roche and Eli Lilly, have each reported impressive DNA delivery capabilities with proprietary techniques. Poseida is not tackling CF to my knowledge, but the CF Foundation was an investor in Engage Bio. A publicly reported “core-then-shell” LNP variant from Jacob Brenner at UPenn also shows major progress.¹²

Another possibility to reduce the total burden of inflammation is to use LNPs to carry gene editing tools, encoded as RNA, for a one-time delivery. By correcting the genomes of stem cells, for example, the dosing frequency could be drastically lowered. Prime Medicine is working on that, and the CF Foundation also announced in October that they are supporting a small, investigator-initiated trial into base editing for a particular CF-causing mutation (called R553X, found in about 2% of U.S. patients). Between gene editing and DNA delivery, I’m optimistic that we may see a cure for my form of CF in the next decade.

Illusions of Control

Things were going great for me as summer began in 2023. My research was humming, my second kid was on the way, and I had blasted an opposite-field home run in the season’s first game of intramural softball. The next day, it appeared I’d strained a chest muscle doing that, but whatever.

Growing up, CF seemed more annoying than dangerous. I played baseball and soccer, ran cross country (very badly, though I don’t think CF was to blame), and made the varsity volleyball team. The breathing treatments took an hour a day, but I used that time to watch sports, read or work. Once I went to college, I had to carefully plan my treatments before going to parties. CF’s practical effect on me was far milder than the term “fatal genetic disease” might imply.

If anything, CF gave me a gift of a capital-P Purpose in life. It allowed me never to doubt I was doing the right thing by pursuing a cure. To be less generous to myself, it held me back in some ways by being a ready-made excuse—anything I wasn’t good at, or not good enough at, I could tell myself was because I had it harder than everyone around me.

CF was also helpful personal motivation. Exercise is basically the best thing you can do when you have CF; it helps directly with mucociliary clearance by mechanically disrupting the mucus layer, and also rehydrates the lung mucus.¹³ So I exercised obsessively, never letting myself fall out of shape, and was convinced that was why CF didn’t affect me.

Meanwhile, that strained chest muscle of mine didn’t heal like I expected. In fact, over the next few weeks it got worse until it was excruciating to take deep breaths. Any MD reading this knows what comes next, because “CF patient with extreme chest pain when taking deep breaths” really only means one thing, and that thing is not a pulled muscle.

Indeed, when I finally got myself checked, it turned out that the chest pain wasn’t muscular but rather pleuritic: I had pneumonia with an empyema (infected fluid buildup in my chest) and ended up in the hospital for several days, followed by a couple of weeks of intravenous antibiotics.

My life acquired a clear “before/after the pneumonia” distinction. I never fully recovered, with a permanent loss of about 15% of my lung function, and radiologist comments on my X-rays went from “unremarkable” to “diffuse bronchiectasis characteristic of a history of CF.” Bronchiectasis is the permanent lung damage that sets in over time with CF, and which I so assiduously managed to avoid until I didn’t. Exercise has lost some of its joy, less “fun to do” and more “stressful not to do.” I’ve become obsessive about healthy eating, and normal colds send me into a spiral of panic.

My health isn’t such a source of pride anymore, and it’s unclear it ever should have been. The course of CF is strongly modulated by genetic variation beyond the CFTR gene, and I almost certainly have good genetics for lung function in all ways except the big one. But whatever the underlying facts, my carefully cultivated internal locus of control, in which I believed that if I did everything right I could stop CF from manifesting, has curdled into self-recrimination that I didn’t go to the doctor sooner, before the permanent lung damage set in.

But despite all that, the effect of my pneumonia shouldn’t be overstated. Given that the Vertex modulators don’t work for me, my life expectancy is theoretically something like 40 (I’m 33), and by all rights I should be in the hospital multiple times a year. But that has not been my experience. The lung function decline isn’t noticeable in daily life: I can run 10 miles, and keep up with everyone else at a pickup basketball game; I’ve only had one more concerning lung episode since the pneumonia. If I could go under a Rawlsian veil of ignorance and come out the other side into a random person’s life, there’s no way in a million years I’d accept the offer.

The very first thing I ever remember wanting to be was the Yankees center fielder. Then I turned 10, realized that goal was unrealistic, and decided I wanted to cure CF instead.

But if there is one thing I have learned throughout my career, it’s that the Yankees center fielder position may be marginally more achievable than my being the one to cure CF. It’s not that professional baseball is a particularly advisable career path, but becoming the Yankees center fielder is a thing that happens; Bernie Williams does, in fact, exist.

Now, I believe in my bones that I will live to see a cure for CF, and that the odds of my being instrumental in this cure are considerable. But being “the one” to cure CF, or really any disease, is a contradiction in terms. Causation is poorly defined in a complex system, but thousands of people spent their lives trying to make sure that I don’t suffocate to death in my own mucus. If all my research goes perfectly—an LNP I invent delivers DNA in the manner I develop—this will be possible only because of the decades of research into LNPs themselves; the discovery of CFTR; and a hundred other advances big and small. The field has also been advanced by the CF Foundation, a well-funded National Institutes of Health, and of course that bedrock of American research: the voters and taxpayers who make it all go. So to those of you reading this, thank you.

---

Jacob Witten is an Assistant Professor in Genetics and Genome Sciences at Case Western Reserve University.

Header image by Ella Watkins-Dulaney.

Cite: Witten, J. “Trying to Cure My Cystic Fibrosis.” Asimov Press (2026). DOI: 10.62211/28wh-42jy

1. Bacterial lung infections (and other causes too numerous to mention) can cause chronic lung inflammation, in which immune cells called neutrophils flail ineffectually at the bacteria. Stressed neutrophils will sometimes explode and spew the DNA from their nuclei everywhere in the hope that it will form a net-like structure that traps bacteria—appropriately called a neutrophil extracellular trap, or NET. There is so much neutrophil inflammation in CF, and so many resulting NETs, that a significant fraction of the mucus layer by weight is DNA. This DNA is a major contributor to the stickiness of the mucus in CF. The inhaled DNA-degrading enzyme breaks down the DNA, thus thinning the mucus layer. ↩
2. The famous story that Alexander Fleming accidentally exposed bacteria to the Penicillium mold that produced penicillin is almost certainly false, but the most plausible accounts still involve a significant amount of luck. Certainly, Fleming had no idea that something like penicillin was possible when he ran the experiments to discover it. ↩
3. G551D replaces glycine (G) with aspartate (D) at position 551. CFTR proteins with this swap can reach the cell surface, but don’t open normally. ↩
4. ΔF508 deletes phenylalanine (F) at position 508. These proteins misfold, so most copies are destroyed before reaching the cell surface. ↩
5. A recently approved successor drug, Alyftrek, is even better. ↩
6. To be fair, there are ways to address nonsense mutations using drugs that inhibit a process called “nonsense-mediated decay” or, separately, promote readthrough of these mutations. However, the effects of these drugs in cell culture are usually smaller than those of Trikafta, and they interfere with important cell processes in a way that rings alarm bells for me (nonsense-mediated decay is present and functionally important in every cell). Personally, I wouldn’t put money on nonsense mutation-specific drugs, though I’m glad somebody is working on them, and I really hope I’m wrong for reasons the next paragraph makes clear. ↩
7. “But Jacob,” I hear you object, “who cares? In CF, mucociliary clearance is blocked, so even if the virus moves very slowly, it will have plenty of extra time to get through the mucus layer!” That is sensible, and I’ve wondered the same. I’m not sure, but I think the issue is that (1) viruses penetrate mucus so slowly that they fall apart before reaching cells; (2) clearance is slow but not nonexistent, so inhaled viral vectors do eventually get cleared; (3) viral vectors become essentially irreversibly trapped, so even three days or three years is not enough for them to reach cells; (4) plugs of mucus block the deposition of inhaled viruses into the deeper airways; or (5) the mucus layer’s importance is overrated, and other factors determine gene therapy failure. ↩
8. To effectively hydrate mucus, CFTR protein needs to work together with another protein called NKCC1, but ciliated cells have a lot less NKCC1 than secretory cells. ↩
9. It is possible to briefly disrupt these junctions, but how safe that might be in humans is unclear. ↩
10. We got an FDA-approved gene therapy for the brain (Kebilidi) before we got anything for the lungs, even though the brain is the most thoroughly protected organ in the body and the lungs are, you know, just sitting there breathing stuff in. I cannot possibly stress enough how hard a problem this is. ↩
11. You can get a sneak preview of LNP improvements and various much wilder developments for the low, low price of funding my lab with a grant or investing in the startup I’m working on. ↩
12. I’m biased—Jake is my cousin’s husband and a mentor—but in my opinion, Jake is by far the most interesting researcher in LNPs today, especially for CF-relevant gene therapy. In addition to the DNA delivery work, he’s also made major progress on reducing DNA-related toxicity, alleviating a major mechanism of inflammation from inhaled LNPs, and identifying and solving blood clotting issues related to intravenous LNP delivery to the lung. Watch that space! ↩
13. Exercise reduces the activity of a sodium channel called ENaC, inhibiting sodium hyperabsorption from the mucus. Keeping sodium ions in the mucus helps keep it hydrated. ↩