The Brakes Released

To treat autoimmune disease, medicine strengthens the immune system’s brakes. To treat cancer, it releases them. What cancer immunotherapy has achieved, who it helps, what it costs, and why both fields are working on the same brakes.

In the spring of 2012 a six-year-old girl named Emily Whitehead had leukaemia that had come back after treatment. Doctors at the Children’s Hospital of Philadelphia took some of her T cells, gave them an engineered receptor that recognises a protein on leukaemia cells, and put them back about a month later. She was the first child to receive what is now called CAR-T therapy. Ten years later the hospital announced that she was still free of cancer, and the doctor who led the trial, Stephan Grupp, who had “always been cautious about calling this a cure”, said that he believed she was cured.[1] Siddhartha Mukherjee told her story in The New Yorker in 2019.[2]

The Second Lock, the essay this one accompanies, is about the immune system’s brakes: the regulatory T cells that keep it from attacking the body, and what happens when they fail. Cancer immunotherapy is the same story told the other way round. A tumour survives in part because the immune system’s brakes are holding back the cells that could kill it. Release the brakes, and some tumours melt away. Release them too far, and the body’s own tissues are attacked. The same molecules, CTLA-4 and PD-1, appear in both essays; so do the same engineered cells, which went from leukaemia to lupus.

Mukherjee, an oncologist and the author of The Emperor of All Maladies, the Pulitzer Prize-winning biography of cancer, is a good guide to this field, and a candid one about its limits. This essay follows the questions his writing raises: how the idea of using the immune system against cancer went from a hunch to a Nobel Prize, how much it has achieved, for whom, and at what price. It checks the claims against the trials, and it runs the toy model of the first essay again, with a tumour added, to show why the line between a cure and an autoimmune disease can be thin.

71.9months: median survival with two checkpoint drugs in advanced melanoma, after at least ten years of follow-up, against 19.9 with one of them alone

20%of US patients with cancer estimated to respond to a checkpoint drug in 2023, out of 57% eligible for one

12/12rectal cancers with broken DNA repair that disappeared on a PD-1 blocker alone, without surgery or radiotherapy

1.23%of patients on a combination of checkpoint drugs who died of a side effect, in a meta-analysis of 112 trials

A hunch

Coley’s toxins

The idea is old. In 1891 William Coley, a surgeon in New York, “injected streptococcal organisms into a patient with inoperable cancer”. Over the next forty years, as head of the bone tumour service at Memorial Hospital, he injected more than 1,000 patients with bacteria or bacterial products. The method was criticised, radiotherapy and then chemotherapy offered something more predictable, and Coley’s toxins fell out of use; a historian of the subject calls him the “Father of Immunotherapy”.[3]

For most of the twentieth century the hunch did not become a treatment. Tumours grow from the body’s own cells, and the mechanisms described in The Second Lock, deletion in the thymus and regulatory T cells, are designed to stop the immune system from attacking those. What changed the field was the discovery that T cells which could attack a tumour often already exist, and are being held back.

Fig. 1 — From Coley’s toxins to personalised vaccines. The events described in this essay; teal for the early science, pink for the clinical era.

The brakes

A Christmas experiment

Two molecules on the surface of T cells turned out to be brakes. In 1992 Tasuku Honjo’s group in Kyoto found a gene they called PD-1, while studying how cells die, not how the immune system is restrained.[4] Its role as a brake became clear only later. CTLA-4 was already known to function “as a brake on T cells”, in the Nobel Committee’s words, and it is also one of the markers of regulatory T cells. James Allison, then at the University of California, Berkeley, asked what would happen if the brake were released. In the Nobel Committee’s account, “Allison and co-workers performed a first experiment at the end of 1994, and in their excitement it was immediately repeated over the Christmas break. The results were spectacular.”[5] Mice given an antibody against CTLA-4 rejected their tumours, “including preestablished tumors”.[6]

The clinical proof took fifteen years. In 2010 a trial of 676 patients with advanced melanoma that had resisted other treatment found that ipilimumab, an antibody against CTLA-4, extended median survival from 6.4 months, with a tumour vaccine alone, to about 10 months.[7] Antibodies against PD-1 followed, with a key study in 2012, and in 2018 Allison and Honjo shared the Nobel Prize in Physiology or Medicine “for their discovery of cancer therapy by inhibition of negative immune regulation”.[5]

Fig. 2 — One set of brakes, two ways to fail. With too little regulation the immune system attacks the body; with too much, a tumour escapes. Cancer immunotherapy pushes from the right; the treatments for autoimmunity in The Second Lock push from the left. A sketch, not a measurement.

The two Nobel Prizes, 2018 for releasing the brakes and 2025 for discovering the cells that apply them, are two halves of one picture (Figure 2). A tumour that a T cell could recognise survives because the brakes are on: regulatory T cells gather in tumours, and many tumours display PD-L1, the molecule that presses on the PD-1 brake. Checkpoint drugs lift those brakes on every T cell in the body, not only on the ones that recognise the tumour.

The tail

Survival that levels off

What made oncologists take notice was not the median, a gain of a few months, but the shape of the survival curve: with checkpoint drugs a fraction of patients stop dying of the disease. Pooling 1,861 patients from twelve ipilimumab trials, Dirk Schadendorf and colleagues found that 22 per cent were alive at three years and that “the survival curve began to plateau around year 3, with follow-up of up to 10 years”, at about 21 per cent.[8]

Fig. 3 — Median survival in advanced melanoma, in months. The 2010 trial compared ipilimumab with a gp100 tumour vaccine in patients who had already been treated (Hodi et al.); CheckMate 067 compared nivolumab, ipilimumab and the two together in untreated patients, with at least ten years of follow-up (Wolchok et al. 2025). The two trials enrolled different patients and should not be compared bar to bar.

The combination of the two brakes went further. CheckMate 067 assigned 945 previously untreated patients with advanced melanoma to nivolumab (against PD-1), ipilimumab (against CTLA-4) or both.[9] Its final report, with at least ten years of follow-up, gives a median overall survival of 71.9 months with the combination, 36.9 with nivolumab and 19.9 with ipilimumab. Median melanoma-specific survival with the combination was more than ten years, not yet reached. And among patients who were alive and free of progression at three years, 96 per cent of those on the combination were still alive, as far as melanoma was concerned, at ten.[10] For a disease whose median survival with advanced spread used to be measured in months, that is the change. It is also, for most patients with most cancers, not what happens.

Who responds

One patient in five

Vinay Prasad and Alyson Haslam have counted, year by year, how many patients with cancer in the US are eligible for a checkpoint drug, by the indications the FDA has approved, and how many of those would be expected to respond, using the response rates in the trials. In 2011, 1.5 per cent were eligible and 0.1 per cent would respond. By 2018 it was 43.6 and 12.5 per cent.[11] In their update to 2023, with 11 drugs and 20 types of tumour, eligibility had reached 56.6 per cent and response 20.1 per cent: “Only about one-fifth of the patients will respond.”[12] Response here means that the tumour shrinks substantially, not that the patient is cured, and Prasad is a frequent critic of how the benefits of cancer drugs are reported, so these are not a promoter’s numbers. They are also, in a field that began at zero, a large number of people.

Fig. 4 — Eligible and responding. Estimated share of US patients with cancer eligible for a checkpoint drug (black) and expected to respond to one (pink). 2011–2018 from Haslam & Prasad (2019, 6 drugs, 14 indications); 2023 from Haslam, Olivier & Prasad (2025, 11 drugs, 20 tumour types), so the last segment also reflects a wider count.

Who are the responders? One clear answer came from a group of tumours whose cells cannot repair a common kind of DNA copying error. Without mismatch repair, mutations pile up, and each mutation can change a protein into a fragment the immune system has never seen. In 2015 Dung Le and colleagues gave a PD-1 blocker to 41 patients. Colorectal cancers with broken repair, which carried on average 1,782 mutations, responded in 4 of 10 cases; colorectal cancers with working repair, with 73, in none of 18.[13] Their next study, of 86 patients with twelve different cancers that shared the same defect, found responses in 53 per cent and complete responses in 21 per cent.[14] In May 2017 the FDA approved pembrolizumab for any solid tumour with this defect, “the first time that FDA has approved a cancer treatment based solely on the presence of a genetic feature in a tumor”.[15]

The most striking result came from rectal cancer. A standard treatment for locally advanced rectal cancer is chemotherapy and radiotherapy followed by surgery to remove the rectum. In 2022 Andrea Cercek and colleagues gave twelve patients whose tumours had broken mismatch repair six months of dostarlimab, a PD-1 blocker, and nothing else. All twelve tumours disappeared, with no trace on MRI, PET scans, endoscopy, examination or biopsy, and none had returned after 6 to 25 months.[16] In the expanded study published in 2025, all 49 patients with rectal cancer who completed treatment had a complete clinical response, 37 of them sustained at twelve months; across all tumour types, 82 of 103 patients avoided surgery, and 92 per cent were free of recurrence at two years.[17] Rectal cancers with this defect are a minority, but for those patients the standard of care is changing.

Fig. 5 — When the DNA repair system is broken, PD-1 blockade works far more often. Share of patients responding in four studies; numbers of patients on the bars. The rectal cancer studies measured complete clinical response, the others objective response, so the bars are not strictly comparable.

Mukherjee’s image for what is going on is ecological. Speaking to Harvard Medicine magazine in 2026, he said that “immunotherapy in some ways changes the immune soil that cancer grows in, but certain cancers have mechanisms to hide away from that immunological milieu. Now we’ve started finding ways to change that milieu.”[18] Tumours with many mutations, in tissues full of T cells, are the easy soil. Tumours with few mutations, or that keep T cells out, the so-called cold tumours, are the hard one.

Living drugs

Cells as medicine

Emily Whitehead’s treatment was the other branch of the field. Instead of releasing the brakes on whatever T cells a patient has, CAR-T therapy gives the T cells a new target. Mukherjee called them “living drugs,” a sort of chimera of the pharmaceutical and the procedural: they are made from each patient’s own cells, they multiply inside the body, and some of them stay for years.[2] In the trial that included Emily, published in 2014, 27 of 30 children and adults with relapsed leukaemia went into complete remission, although every patient had cytokine release syndrome, severe in 27 per cent.[19] In the multicentre trial that followed, 81 per cent of 75 patients were in remission within three months, and half were free of events a year later.[20] On 30 August 2017 the FDA approved the product, Kymriah, calling it “the first gene therapy available in the United States”.[21]

The longest view comes from two adults with chronic lymphocytic leukaemia treated in 2010. Both went into complete remission, and more than ten years later their CAR-T cells could still be found in the blood, now dominated by an unusual population of CD4 cells that had taken on the job of killing.[22] A drug still working a decade after a single dose is something new in medicine.

So is the price. “Perhaps the most immediate implication of the blurring of lines between procedure and drug is the conundrum of price,” Mukherjee wrote, and gave the list prices at the time: $475,000 for a single dose of Kymriah for children with leukaemia, $373,000 for Yescarta, a CAR-T product for lymphoma, before the cost of the hospital stay.[2] Readers should know that he is not a bystander here: he co-founded Vor Biopharma, which works on cell therapies for blood cancers, is the scientific founder of Myeloid Therapeutics, and co-founded Immuneel, a company in Bengaluru that aims to widen access to CAR-T treatment in India.[23][24][25]

CAR-T cells have been much less successful against solid tumours, which offer few targets found only on cancer cells and which keep T cells out. That is starting to change. In a trial at Stanford led by Michelle Monje, CAR-T cells against GD2 given to patients with diffuse midline gliomas, among the deadliest brain tumours, shrank four tumours by more than half, and one patient had a complete response lasting more than 30 months.[26] In China, the first randomised trial of a CAR-T therapy in any solid tumour, against claudin-18.2 in advanced gastric cancer, extended median progression-free survival from 1.77 to 3.25 months, at the cost of serious side effects in 99 per cent of treated patients.[27] These are beginnings. And the same CD19 CAR-T cells that cured Emily are the ones now being tested in lupus, as The Second Lock describes: a cancer drug turned into an attempt to reset the immune system.

Running it here

The window, in a toy model

The model in The Second Lock had self-reactive T cells, the damage they cause, and a brake that stands for regulatory T cells. Here it gets a tumour, which grows, and a second set of T cells that can kill it, held back by a much stronger brake, since tumours gather regulatory T cells and display PD-L1. A checkpoint drug, in the model, removes the same share of every brake in the body: of the brake on the cells that attack the tumour, and of the brake on the self-reactive cells. As before, the numbers are chosen to show the mechanism, not fitted to any cancer; the method note gives the equations.

Fig. 6 — The window, in the toy model. The tumour left after two years of treatment (black, log scale) for drugs of increasing strength, measured as the share of every brake they remove. Pink dots mark where autoimmunity starts, without any trigger, in people whose own brakes are weak (2), ordinary (4), strong (6) or very strong (8) on the scale of The Second Lock.

Figure 6 shows the result. Without treatment the tumour fills 81% of the space it can occupy after two years. As the drug strengthens, little happens at first, and then, once about 57% of every brake is removed, the cells that attack the tumour escape their brake and the tumour drops more than a thousandfold. Autoimmunity has its own threshold, and it depends on the person. With a weak brake of their own, the self-reactive cells take off when only 28% of the brakes are removed, before the tumour responds at all. With an ordinary brake the threshold is 64%, which leaves a narrow window; with a strong or very strong brake, 76% or 82%, a wide one.

Fig. 7 — Two doses for one person with an ordinary brake, in the toy model. With a drug removing 60% of every brake, the tumour (solid black) falls below a thousandth of the space it can occupy by day 287 and the self-reactive cells (teal) stay near rest. With 75%, the tumour (dashed) gets there by day 78, but the self-reactive cells (pink) climb about eightyfold and stay there.

Figure 7 follows one person with an ordinary brake through two years at two strengths. The stronger drug shrinks the tumour faster, and it starts a self-sustaining autoimmune loop of the kind described in The Second Lock, which goes on for as long as the model runs. The model is crude in every particular: real patients have many tissues, each with its own self-reactive cells and its own threshold, which is why side effects strike the colon in one patient and the heart in another. But it shows the shape of the problem clinicians face. The brakes that hold back the tumour’s attackers and the brakes that protect the body are, to a large extent, the same brakes, and the margin between them differs from person to person.

The price

Autoimmunity on demand

The trials show that margin. In the 2010 ipilimumab trial, serious immune-related side effects of grade 3 or 4 occurred in 10 to 15 per cent of patients, and 14 deaths, 2.1 per cent, were related to the drug, seven of them immune-related.[7] With nivolumab and ipilimumab together in CheckMate 067, treatment-related side effects of grade 3 or 4 occurred in 55 per cent of patients, against 16 per cent on nivolumab alone.[9] A meta-analysis of 112 trials with 19,217 patients put the rate of death from side effects at 0.36 per cent for PD-1 blockers, 0.38 per cent for PD-L1 blockers, 1.08 per cent for ipilimumab and 1.23 per cent for combinations (Figure 8). Most deaths on ipilimumab were from colitis; the deadliest side effect was inflammation of the heart, fatal in 52 of the 131 cases reported to the WHO database.[28]

Fig. 8 — The rarest and worst price. Share of treated patients who died of a side effect, by type of checkpoint drug, in a meta-analysis of 112 trials (Wang et al. 2018). Pink: regimens that block CTLA-4, the brake most closely tied to regulatory T cells.

The pattern fits the model and the scurfy mouse of The Second Lock: blocking CTLA-4, a molecule that regulatory T cells carry, is the more dangerous of the two, and combining the brakes is the most dangerous of all. Releasing the brakes on the tumour releases them on the body, and the body pays in proportion.

What comes next

Vaccines made for one person

If a tumour’s mutations are what make it visible, a vaccine can point the immune system at them. Each patient’s tumour is sequenced, the mutations most likely to produce fragments that T cells recognise are chosen, and a messenger RNA vaccine is made for that patient alone. In a randomised phase 2 trial in 157 patients whose melanoma had been removed surgically, adding such a vaccine, mRNA-4157, to pembrolizumab reduced the risk of recurrence by about 44 per cent, although the result just missed conventional statistical significance (p = 0.053).[29] The phase 3 trial, INTerpath-001, with 1,137 patients, was presented at the ESMO congress in October 2026. Recurrence was reduced by 42 per cent (hazard ratio 0.58), with no difference in the first six months and a large one afterwards, as one would expect from a vaccine that takes time to work; serious treatment-related side effects were about as common in both groups, and overall survival is not yet mature.[30] The companies call it the first positive phase 3 result for an mRNA cancer therapy.

The hardest test is pancreatic cancer. In a small study of patients whose tumours had been removed, half of 16 patients developed T cells against their personalised vaccine, and those who did had much longer survival without recurrence than those who did not.[31] After a median of 3.2 years, the difference held, two of the responders had relapsed, and the vaccine-induced T-cell clones were estimated to last 7.7 years on average.[32] Responders and non-responders may differ in ways other than the vaccine, so this is a lead, not proof.

The last approach closes the circle with the 2025 Nobel Prize. Instead of releasing brakes everywhere, it removes the regulatory T cells inside tumours, which carry a molecule called CCR8.[33] Antibodies against CCR8 and newer antibodies against CTLA-4 designed to deplete regulatory T cells in tumours are in early trials; a 2026 review describes them as showing “selective intratumoral Treg depletion with partially improved tolerability”.[34] If they work, they would widen the window in Figure 6: release the brakes on the tumour without releasing them on the body.

Coda

What is known and what is hoped

Established

Blocking CTLA-4 or PD-1 extends survival in advanced melanoma and other cancers, with a lasting plateau for a minority of patients. CD19 CAR-T cells produce remissions in relapsed leukaemia, some lasting more than a decade. Both carry serious autoimmune and inflammatory risks.

Strong evidence

Tumours with broken mismatch repair respond far more often; in rectal cancer with this defect, a PD-1 blocker alone can replace surgery and radiotherapy. About one patient in five with cancer in the US is estimated to respond to a checkpoint drug.

Promising

Personalised mRNA vaccines after surgery for melanoma (one positive phase 3 trial, survival not yet mature); CAR-T cells in brain and stomach cancers.

Early

Removing regulatory T cells only inside tumours; personalised vaccines in pancreatic cancer; in vivo CAR-T cells.

Price is the part of this story that has changed least since Mukherjee called it a conundrum. What has changed most is the picture of the immune system. A century after Coley, the brakes that The Second Lock described as the body’s protection turn out to be one of a tumour’s best defences, and the same knowledge serves both directions. Treatments for autoimmune disease try to strengthen the brakes without leaving the body open to infection and cancer; treatments for cancer try to release them without turning the immune system on the body. The 2018 and 2025 Nobel Prizes rewarded the two halves of that problem. Neither field has solved it, and each is now borrowing the other’s tools.

Open threads

Where this could go

Predicting the window. The model says that the margin between response and autoimmunity depends on each person’s brakes. A test that measured them before treatment, rather than finding out in the clinic, would let doctors choose between one drug and two.

Cold tumours. Many cancers barely respond to checkpoint drugs. Whether vaccines, CAR-T cells or the removal of regulatory T cells inside tumours can change their “immune soil” is the question that will decide how far the one-in-five figure can rise.

Cost. Prices of several hundred thousand dollars a dose limit cell therapy to rich health systems. Manufacturing closer to the patient, cheaper production in countries such as India, and CAR-T cells made inside the body are attempts to change that; none has yet done so at scale.

On method and tools

This essay was researched and written with Claude Opus 5.5 (Anthropic): human specification, editorial direction and critical review; machine research, drafting, computation and figure generation. The idea came from Luis, who suggested Siddhartha Mukherjee’s writing on cancer immunotherapy as a companion to The Second Lock. Mukherjee’s New Yorker article and his Harvard Medicine interview were read in full, as were the Nobel Committee’s 2018 press release, the FDA and NCI announcements and the hospital’s account of Emily Whitehead; his books were not consulted for this essay, and nothing here is quoted from them. For the trials and studies, figures and quotations come from the abstracts; the INTerpath-001 phase 3 figures come from the company’s summary of the conference presentation, as the full paper has not been published. A research agent (Claude) gathered and checked the sources, and each number in the text was compared with what it had read.

The model extends the one in scripts/tolerance.py (see The Second Lock). In scripts/brakes.py a tumour N grows as dN/dt = 0.05N(1 − N) − 0.005TN and is killed by T cells T, which follow dT/dt = 0.1 − 0.1T + T·H(N)/(1 + 20(1 − u))·(1 − T/100), with H(N) = N²/(10⁻⁸ + N²); the self-reactive cells and self antigen follow the equations of The Second Lock with their brake b multiplied by (1 − u), where u is the strength of the checkpoint drug. Autoimmunity is counted as starting when the brake b(1 − u) falls below the point at which the healthy state disappears (1.44); the tumour is called controlled when it is below a thousandth of the space it can occupy after two years. The equations are stepped every 0.05 days; the numbers are in docs/brakes-results.json. The model has one tissue, one tumour and no resistance, and its units are arbitrary; it is there to show a mechanism. The other figures are drawn by scripts/brakes_data.py from the published numbers named in their captions. The cover is drawn from the model.

Authored by: Luis Matos Ferreira — Physicist, Developer, Writer

Related essays on this blog

1. The Second Lock — the companion: regulatory T cells, autoimmune disease, and treatments that reset the immune system.
2. Cancer — an earlier survey here of new cancer treatments, from immunotherapy to gene editing (2024).
3. The Second Reader — what AI is and is not doing in medicine.

Sources

1. Children’s Hospital of Philadelphia, news release on Emily Whitehead, the first child to receive CAR T-cell therapy, ten years after her treatment, 10 May 2022, chop.edu.
2. S. Mukherjee, “The promise and price of cellular therapies”, The New Yorker, 15 July 2019 (print edition 22 July 2019, “New Blood”), newyorker.com.
3. E. F. McCarthy, “The toxins of William B. Coley and the treatment of bone and soft-tissue sarcomas”, Iowa Orthopaedic Journal 26, 154–158 (2006), pmc.ncbi.nlm.nih.gov.
4. Y. Ishida, Y. Agata, K. Shibahara & T. Honjo, “Induced expression of PD-1, a novel member of the immunoglobulin gene superfamily, upon programmed cell death”, EMBO Journal 11, 3887–3895 (1992), doi.org.
5. The Nobel Assembly at Karolinska Institutet, press release, 1 October 2018, nobelprize.org.
6. D. R. Leach, M. F. Krummel & J. P. Allison, “Enhancement of antitumor immunity by CTLA-4 blockade”, Science 271, 1734–1736 (1996), doi.org.
7. F. S. Hodi, S. J. O’Day, D. F. McDermott et al., “Improved survival with ipilimumab in patients with metastatic melanoma”, New England Journal of Medicine 363, 711–723 (2010), doi.org.
8. D. Schadendorf, F. S. Hodi, C. Robert et al., “Pooled analysis of long-term survival data from phase II and phase III trials of ipilimumab in unresectable or metastatic melanoma”, Journal of Clinical Oncology 33, 1889–1894 (2015), doi.org.
9. J. Larkin, V. Chiarion-Sileni, R. Gonzalez et al., “Combined nivolumab and ipilimumab or monotherapy in untreated melanoma”, New England Journal of Medicine 373, 23–34 (2015), doi.org.
10. J. D. Wolchok, V. Chiarion-Sileni, P. Rutkowski et al., “Final, 10-year outcomes with nivolumab plus ipilimumab in advanced melanoma”, New England Journal of Medicine 392, 11–22 (2025), doi.org.
11. A. Haslam & V. Prasad, “Estimation of the percentage of US patients with cancer who are eligible for and respond to checkpoint inhibitor immunotherapy drugs”, JAMA Network Open 2, e192535 (2019), doi.org.
12. A. Haslam, T. Olivier & V. Prasad, “How many people in the US are eligible for and respond to checkpoint inhibitors: an empirical analysis”, International Journal of Cancer 156, 2352–2359 (2025), doi.org.
13. D. T. Le, J. N. Uram, H. Wang et al., “PD-1 blockade in tumors with mismatch-repair deficiency”, New England Journal of Medicine 372, 2509–2520 (2015), doi.org.
14. D. T. Le, J. N. Durham, K. N. Smith et al., “Mismatch repair deficiency predicts response of solid tumors to PD-1 blockade”, Science 357, 409–413 (2017), doi.org.
15. National Cancer Institute, “Pembrolizumab approved for tumors with specific genetic features”, Cancer Currents blog, 2017, on the FDA approval of 23 May 2017, cancer.gov.
16. A. Cercek, M. Lumish, J. Sinopoli et al., “PD-1 blockade in mismatch repair-deficient, locally advanced rectal cancer”, New England Journal of Medicine 386, 2363–2376 (2022), doi.org.
17. A. Cercek, M. B. Foote, B. Rousseau et al., “Nonoperative management of mismatch repair-deficient tumors”, New England Journal of Medicine 392, 2297–2308 (2025), doi.org.
18. M. McDonough, “Siddhartha Mukherjee’s evolving view of cancer”, Harvard Medicine magazine, winter 2026, magazine.hms.harvard.edu.
19. S. L. Maude, N. Frey, P. A. Shaw et al., “Chimeric antigen receptor T cells for sustained remissions in leukemia”, New England Journal of Medicine 371, 1507–1517 (2014), doi.org.
20. S. L. Maude, T. W. Laetsch, J. Buechner et al., “Tisagenlecleucel in children and young adults with B-cell lymphoblastic leukemia”, New England Journal of Medicine 378, 439–448 (2018), doi.org.
21. US Food and Drug Administration, “FDA approval brings first gene therapy to the United States”, press release, 30 August 2017 (archived copy), web.archive.org.
22. J. J. Melenhorst, G. M. Chen, M. Wang et al., “Decade-long leukaemia remissions with persistence of CD4+ CAR T cells”, Nature 602, 503–509 (2022), doi.org.
23. The Boston Globe, “Cambridge biotech seeking to treat blood cancers closes up 108 percent in stock market debut”, 5 February 2021, on Vor Biopharma, bostonglobe.com.
24. Myeloid Therapeutics, “Myeloid Therapeutics launches with over $50 million in financing and two clinical trials”, press release, January 2021, biospace.com.
25. The Daily Pennsylvanian, “Five facts about Siddhartha Mukherjee”, May 2024, for Immuneel Therapeutics and his other companies, thedp.com.
26. M. Monje, J. Mahdi, R. Majzner et al., “Intravenous and intracranial GD2-CAR T cells for H3K27M+ diffuse midline gliomas”, Nature 637, 708–715 (2025), doi.org.
27. C. Qi, C. Liu, Z. Peng et al., “Claudin-18 isoform 2-specific CAR T-cell therapy (satri-cel) versus treatment of physician’s choice for previously treated advanced gastric or gastro-oesophageal junction cancer (CT041-ST-01): a randomised, open-label, phase 2 trial”, The Lancet 405, 2049–2060 (2025), doi.org.
28. D. Y. Wang, J.-E. Salem, J. V. Cohen et al., “Fatal toxic effects associated with immune checkpoint inhibitors: a systematic review and meta-analysis”, JAMA Oncology 4, 1721–1728 (2018), doi.org.
29. J. S. Weber, M. S. Carlino, A. Khattak et al., “Individualised neoantigen therapy mRNA-4157 (V940) plus pembrolizumab versus pembrolizumab monotherapy in resected melanoma (KEYNOTE-942): a randomised, phase 2b study”, The Lancet 403, 632–644 (2024), doi.org.
30. Moderna, “Recapping the phase 3 INTerpath-001 results”, 24 October 2026, on the ESMO 2026 presentation (LBA1); a company summary, with the full paper not yet published, modernatx.com.
31. L. A. Rojas, Z. Sethna, K. C. Soares et al., “Personalized RNA neoantigen vaccines stimulate T cells in pancreatic cancer”, Nature 618, 144–150 (2023), doi.org.
32. Z. Sethna, P. Guasp, C. Reiche et al., “RNA neoantigen vaccines prime long-lived CD8+ T cells in pancreatic cancer”, Nature 639, 1042–1051 (2025), doi.org.
33. The Nobel Committee for Physiology or Medicine, “Scientific background to the Nobel Prize in Physiology or Medicine 2025”, on CCR8 on tumour-infiltrating regulatory T cells, nobelprize.org.
34. M. Navarrete, C. Costoya, F. Galvez-Cancino et al., “Regulatory T cell depletion in cancer: challenges, opportunities, and future directions for antibody development”, Annual Review of Medicine 77, 239–251 (2026), doi.org.