Stage 4 breast cancer erased by her own immune cells

Researchers at the National Cancer Institute sorted immune cells from inside Judy Perkins's tumor, grew the few that recognized it to billions and returned them; the cancer disappeared.

Immune cells that were already in the tumour

In 1988 Steven Rosenberg and colleagues at the National Cancer Institute reported in the New England Journal of Medicine a preliminary trial in 20 patients with metastatic melanoma. Lymphocytes were extracted from freshly removed tumours, expanded in the laboratory, and returned to the patient with interleukin-2 after a single dose of cyclophosphamide. Objective regression occurred in 9 of 15 patients who had not previously received interleukin-2, and in 2 of 5 in whom it had failed. The regressions lasted from two to more than 13 months. The cells were called tumour-infiltrating lymphocytes, or TILs, and the principle was simple: the immune system had already sent cells into the tumour that recognised it, in numbers too small to win.

For twenty-five years the approach worked in melanoma and in little else. Melanoma carries an unusually high number of mutations, and each mutation can create an abnormal protein fragment, a neoantigen, that T cells can recognise as foreign. Common cancers of the breast, colon and gastrointestinal tract carry far fewer, and immunotherapy of every kind had done little for them.

Finding the cells that matter

The change came from sequencing. In 2014 Eric Tran, Rosenberg and colleagues reported in Science a patient with metastatic cholangiocarcinoma whose tumour was sequenced in full. Among the TILs grown from the patient's tumour was a population of CD4 helper T cells that recognised one specific mutation, in a gene called ERBB2IP. A first infusion in which about 25 percent of the cells were mutation-specific produced a decrease in target lesions and prolonged stabilisation. When the disease progressed, the patient was retreated with a population more than 95 percent pure for the mutation-reactive cells and the tumours regressed again. The method was now to sequence the tumour, list its mutations, test the patient's own TILs against each one, and expand only the cells that reacted.

Patient 4136

The patient at the centre of this story was a 49-year-old woman with oestrogen receptor-positive, HER2-negative metastatic breast cancer that had progressed through multiple lines of chemotherapy and hormonal treatment. She was enrolled in the NCI's phase 2 trial NCT01174121, which tests TILs in metastatic epithelial cancers. Her results were published in Nature Medicine in June 2018 by Nikolaos Zacharakis, Stephanie Goff, Steven Feldman, Rosenberg and colleagues, where she is identified as patient 4136.

Sequencing of her tumour identified 62 somatic mutations. Her TILs were screened against all of them. They recognised four: mutant versions of the proteins SLC3A2, KIAA0368, CADPS2 and CTSB. Cultures from three tumour fragments enriched for those reactivities were expanded over two weeks with interleukin-2, an anti-CD3 antibody and irradiated feeder cells. Deep sequencing of the final product found that eight mutation-reactive T cell receptor clonotypes, seven against SLC3A2 and one against KIAA0368, accounted for 23.1 percent of the sequences. The infused cells were 62.5 percent CD4-positive.

Eighty-two billion cells

Before infusion she received lymphodepleting chemotherapy with cyclophosphamide and fludarabine to make room for the new cells, and a single dose of pembrolizumab, a checkpoint inhibitor, two days before. On day zero she received 8.2 × 10¹⁰ TILs, about 82 billion cells, by intravenous infusion, followed by aldesleukin (interleukin-2) at 720,000 international units per kilogram every eight hours for seven doses. Pembrolizumab was repeated on days 21, 42 and 63. She developed transient grade 3 low phosphate, the expected grade 4 blood count suppression from the chemotherapy, and grade 3 febrile neutropenia during the interleukin-2, which resolved with antibiotics.

At the first assessment six weeks after infusion, her target tumour burden had fallen by 51 percent. All lesions had resolved on imaging at one year. At the most recent evaluation in the paper, 22 months after infusion and 20 months after her last dose of pembrolizumab, she remained in complete response. Biopsies of her chest wall and subcutaneous masses taken before treatment showed no expression of PD-L1, the ligand that pembrolizumab blocks, and the authors wrote that a short course of single-agent pembrolizumab was unlikely to explain the result.

The series behind the case

One patient is a case report. The group's follow-up appeared in the Journal of Clinical Oncology on February 1, 2022. Forty-two women with metastatic breast cancer refractory to prior treatment had a metastatic lesion removed, TILs grown and their tumours sequenced; the median number of nonsynonymous mutations per patient was 112. Twenty-eight of the 42, or 67 percent, had TILs that recognised at least one of their own mutations, with a median of three neoantigens per patient. Thirteen had reactivity strong enough for treatment, eight remained clinically eligible, and six were treated with enriched neoantigen-reactive TILs plus up to four doses of pembrolizumab. Three had objective regressions: the complete response described above, then ongoing beyond 5.5 years, and two partial responses lasting six and ten months. The NIH's account of the study adds that the two partial responders had tumour shrinkage of 52 and 69 percent, that residual disease was later removed surgically, and that both were without evidence of cancer at roughly five and 3.5 years.

The conclusion the authors drew was that most breast cancers are immunogenic after all. The response rate was three of six.

From one lab to an approved product

The approach reached a randomised trial first in melanoma. In 2022 Maartje Rohaan, John Haanen and colleagues reported in the New England Journal of Medicine a phase 3 trial, funded by the Dutch Cancer Society and others, of 168 patients with advanced melanoma, 86 percent of them refractory to anti-PD-1 therapy, assigned to TILs or to ipilimumab. Median progression-free survival was 7.2 months with TILs and 3.1 months with ipilimumab; objective response rates were 49 and 21 percent; median overall survival was 25.8 and 18.9 months. Every TIL patient had a grade 3 or higher adverse event, mostly from the lymphodepleting chemotherapy.

On February 16, 2024, the FDA approved lifileucel, sold as Amtagvi, for adults with unresectable or metastatic melanoma. The agency called it the first FDA-approved tumour-derived T cell immunotherapy. In 73 patients at the recommended dose, the objective response rate was 31.5 percent, with three complete and 20 partial responses; 56.5 percent of responses were maintained at six months, 47.8 percent at nine and 43.5 percent at twelve. The label carries a boxed warning for prolonged severe cytopenia, severe infection, cardiac disorders and fatal treatment-related complications.

What the story does and does not show

The 2018 paper showed that one woman's metastatic, chemotherapy-refractory breast cancer regressed completely after an infusion of 82 billion of her own neoantigen-reactive T cells, and that the regression was still complete at 22 months. The 2022 paper showed three objective responses in six treated patients, with her complete response then beyond 5.5 years. The 2024 approval covers melanoma, not breast cancer, and is based on a 31.5 percent response rate.

The story's card describes a treatment that uses no drug at all. That is not what the record shows. She received cyclophosphamide and fludarabine to deplete her own lymphocytes, seven doses of high-dose interleukin-2 to support the infused cells, and four doses of pembrolizumab; the authors argued that the checkpoint inhibitor was unlikely to be responsible, but the cells were not given alone. The breast cancer evidence remains a case report and a six-patient pilot from a single institution. Of 42 women screened, six were treated and one was cured, and the paper does not establish which features of her tumour or her cells made the difference. The approach requires surgery to obtain tumour, weeks of cell culture, inpatient chemotherapy and interleukin-2, and expertise that exists in few centres.

What the story does show is that cells capable of destroying a breast cancer were present inside it, that whole-exome sequencing made it possible to find the four mutations they recognised, and that grown to 82 billion and returned, they did what no drug had done for her.

The cure was inside the tumor. They found it and multiplied it.