177 - The development of cancer immunotherapy and its promise for treating advanced cancers

Rosenberg's group developed high-dose IL-2 in the early 1980s after the discovery of IL-2 by Morgan, Ruscetti, and Gallo in 1976. The first patient treated had melanoma; the third and fourth had renal cancer. Response rates in both diseases turned out to be 15–20% with about a third of complete responders having durable regressions — approximately 5–7% of all treated patients experienced a complete, lasting cure. Before this, choriocarcinoma and germ cell tumors were the only systemic cancers curable by drug therapy.

The TIL approach arose from the observation that tumors are already infiltrated by reactive lymphocytes — Rosenberg's intuition was to look within the cancer itself for cells already doing battle against it. Growing TILs from peripheral blood gave 15% response rates; growing from within the tumor gave 30–35% in melanoma. The evolution was to select TILs specifically reactive against identified neoantigens, building on the neoantigen identification assay.

Rosenberg's lab partnered with Zelig Eshhar to build the first anti-CD19 CAR-T system. Johnson and Johnson evaluated and declined. Kite ultimately commercialized it into a standard-of-care treatment. The conceptual barrier for solid tumors: CAR-T recognizes surface molecules, but most solid-tumor antigens are intracellular proteins that must be processed into peptides and presented on HLA molecules. You need a T-cell receptor, not a CAR, to recognize them.

Rosenberg frames the neoantigen landscape data as both exciting and sobering. The good news is that virtually every cancer is immunologically visible — the tumor's somatic mutations create neoantigens, and his lab has documented this in 195 cancers across all major solid tumor histologies. The bad news is that 392 of 393 neoantigens identified were patient-specific (only KRAS was shared). The path forward requires solving three problems: faster cheaper tumor sequencing; reliable prediction of which mutations produce immunogenic peptides on a given patient's HLA type; and efficient methods to expand the relevant T-cell clones.

MSI-high (dMMR) tumors fail to repair replication errors, accumulating thousands of insertion-deletion mutations that produce frameshift neoantigens — highly immunogenic because they are completely novel amino acid sequences never seen by thymic tolerance selection. PD-1 and CTLA-4 blockade then allows T-cells recognizing these neoantigens to expand and kill. The converse is also important: in microsatellite-stable colon cancer or pancreatic cancer, there are far fewer mutations, so even fully activated T-cells find little to attack.

Rosenberg describes these patients as people 'who would probably be expected to live no more than 6 months without a miracle.' They come to NCI for 'Hail Marys.' If 20% are saved, that is remarkable against a baseline of essentially zero for metastatic disease. Published proof-of-principle cases now include liver, bile duct, breast, colon, cervical, and ovarian cancers.

Rosenberg explains that the US sees roughly 600,000 cancer deaths per year, of which about 550,000 are from solid epithelial cancers — colon, pancreas, lung, breast, ovary, prostate and similar. Surgery cures more than half of those who are operable. But the approximately half who are inoperable or who develop systemic spread essentially all die. Standard chemotherapy regimens in metastatic colorectal cancer raised median survival from 8–10 months to about 2.5 years — 'an example where life has been extended by years' — but the cancer always wins. Attia describes it as 'the dirty little secret of oncology.'

Rosenberg's lab at the NCI used IL-2 first alone and then in combination with tumor-infiltrating lymphocytes (TILs) grown in IL-2. Response rates in melanoma were about 15–20% with IL-2 alone; a third of those complete responders had durable regressions. When TILs were added (30–35% response rate in melanoma), the responses were real but not more durable. Crucially, every cancer type other than melanoma and renal cell was resistant — 'the first patient that we treated with this revised regimen happened to have a melanoma; the third and fourth had renal cancer' — and those two histologies remained the only solid-tumor exceptions to the 100% death rate until the advent of checkpoint inhibitors.

Rosenberg explains the fundamental biology: lymphocytes circulate constantly — the heart pumps them past every tissue every 14–15 seconds. When they encounter a mutant peptide on an MHC molecule that matches their TCR, they can kill the cell — but only if their brakes are off. Checkpoint inhibitors remove those brakes. In MSI-high tumors, mismatch-repair deficiency generates a very high density of neoantigenic mutations, providing abundant targets for the reactivated T-cells. In microsatellite-stable colon cancer or pancreatic cancer, there are far fewer mutations, so even fully activated T-cells find little to attack.

Rosenberg traces the conceptual lineage: gene-therapy research in his lab starting in 1989 (first human gene therapy, published in NEJM) → inserting the IL-2 gene into TILs → chimeric antigen receptors for CD19. The key insight was that CD19 is expressed on malignant B-cells but also on normal B-cells, and you can accept losing the normal B-cells (manageable with IVIG) because they are not essential. For most solid tumors, this trade-off does not work — you cannot accept destroying normal pancreatic, liver, or lung tissue even to kill the cancer.

The good news: the Achilles heel of every cancer is the very mutations that caused it. In 195 cancers regardless of histology, Rosenberg's lab found immunogenic neoantigens. The bad news: 392 of 393 antigens were unique — not shared by any other patient in the cohort (with the exception of KRAS). This means every treatment requires sequencing the patient's tumor, identifying the mutation-derived peptides, confirming T-cell reactivity, expanding those specific T-cells, and infusing them — a highly personalized, expensive, time-intensive process.

Rosenberg describes growing TILs as 'intuitively' looking within the tumor itself for cells already doing battle against the cancer. The early TIL work in the 1980s produced real but short-lived responses. The evolution was to select TILs specifically reactive against identified neoantigens, expand those selectively, and then genetically engineer them if natural reactivity is insufficient. The resulting published responses in liver, bile duct, colon, and breast cancers represent the first glimpse of a path to treating the common epithelial cancers.

Rosenberg describes the process: RAC initially voted 13-4 against, then reversed to unanimous. NIH director James Wyngaarden insisted on unanimous consent. Biotechnology activists sued to stop it. The final marker-gene insertion was the scientific precursor to everything that followed: inserting the IL-2 gene into TILs, then chimeric antigen receptors leading to CAR-T, then TCR gene transfers targeting neoantigens.

Rosenberg's naive follow-up experiment — transfusing blood from the spontaneously regressed patient into another patient with gastric cancer — failed, but planted the concept: something in the immune system caused tumor rejection. The kidney-transplant case was more direct — cancer from an immunosuppressed donor grew under immunosuppression and vanished when it was stopped. Both cases led him to NCI in 1974 with a specific research agenda.