Longevity & AgingResearch PaperOpen Access

Blood histidine levels track with immunotherapy success across 16 cancer cohorts

Profiling 4,336 plasma samples from 1,714 patients, researchers linked higher histidine to better immunotherapy outcomes. Succinate and long-chain fatty acids pointed the other way.

Thursday, November 12, 2026 0 views
Published in Nat Med
Histidine molecule glowing above a blood plasma sample, with T cells attacking a tumor and protein-rich foods in the background

Summary

Immune-checkpoint inhibitors help only some patients with advanced cancer. Researchers profiled 154 plasma metabolites in 4,336 samples from 1,714 patients across 16 cohorts and five tumor types. A machine-learning model combined metabolites with clinical data to predict 12-month progression-free survival. Histidine, an essential amino acid from dietary protein, was the strongest favorable marker. Succinate and long-chain fatty acids were linked to worse outcomes. Higher histidine also tracked with longer survival in several external cohorts and with lower cancer risk in smokers with cardiovascular disease. In mice, oral histidine tended to enhance checkpoint-inhibitor effects and slowed tumor establishment. Per the abstract, histidine-rich diets improved survival in patients without microbiome signatures of histidine breakdown. The findings are associative and need prospective testing before anyone changes diet or treatment.

0:00--:--

Detailed Summary

Why it matters: Immune-checkpoint inhibitors (ICIs) transformed cancer care, but most patients still do not benefit durably. Metabolism shapes how T cells and dendritic cells function and how tumors compete with them for nutrients. Little is known about which circulating metabolites predict who responds, or whether diet or the gut microbiome could be adjusted to improve outcomes.

What was studied: The team ran targeted mass-spectrometry metabolomics on 4,336 on-treatment plasma samples from 1,714 patients in 16 cohorts across Europe and North America. The cohorts covered lung, kidney, bladder, colon and skin cancers and five ICI-based modalities, including fecal microbiota transplantation. They trained a dynamic random-forest survival model (DynForest) on 105 patients from the IML1 and SABR cohorts, using 154 metabolites measured over time plus clinical variables. Variables were stepwise pruned to a compact signature. The model was validated in 30 patients (PANDORE, neoadjuvant bladder cancer). Histidine was then tested in seven external cohorts, in smokers at high cancer risk, in immune-cell profiling, and in mouse tumor models.

Key results: The final signature combined age, body mass index (BMI), renal function and a handful of metabolites. Histidine was the dominant predictor, while succinate and long-chain fatty acids were negatively associated with outcome. Reported AUCs were about 0.88 in training and 0.73 in validation, though the 95% confidence interval was wide (51–95%). Higher histidine was associated with longer survival in IML1–SABR, AtezoTRIBE, and the dendritic-cell vaccine arm of MIND-DC. It also tracked with better progression-free and overall survival in PRIMM and in the 743-patient CheckMate-025 trial, where it interacted significantly with treatment arm. High histidine was linked to lower risk of developing or dying of tobacco-related cancer in two smoker cohorts, with a non-significant trend in a third. Histidine rose over time in responders, and in the PROMIT trial a larger chemotherapy-induced rise preceded anti-PD-1 response. Histidine correlated with activated CD4+ T cells, TH1 and TFH markers and TCR-engaged double-positive T cells. In mice, oral L-histidine tended to speed ICI-driven rejection of fibrosarcoma and melanoma and delayed tumor establishment when given prophylactically. It raised plasma and tumor histidine and shifted tumor adenosine monophosphate and tryptophan.

Implications: Histidine may serve as a dynamic marker of favorable immune fitness and as a candidate target for nutritional intervention. The abstract reports that histidine-rich diets improved progression-free survival in patients lacking dysbiotic microbiome signatures linked to histidine catabolism, which suggests that gut microbes influence whether dietary histidine helps.

Caveats: The training and validation cohorts were small and heterogeneous. A clinical-only model performed comparably in PANDORE (AUC 80.8%), so the added value of metabolites is unproven. Most human data are observational, and some mouse effects were trends rather than firm differences. The supplied text was truncated before the microbiome and mechanistic details, so those are summarized from the abstract only.

Key Findings

  • Machine learning on 154 plasma metabolites plus clinical data predicted 12-month progression-free survival (AUC ~0.88 training, 0.73 validation).
  • Histidine was the dominant favorable predictor; succinate and long-chain fatty acids were negatively associated with immunotherapy outcome.
  • Higher histidine tracked with longer survival in external cohorts including the 743-patient CheckMate-025 trial, AtezoTRIBE and PRIMM.
  • High histidine was associated with lower tobacco-related cancer risk in smokers with cardiovascular disease (PREVALUNG, ROBINSCA).
  • Oral histidine tended to boost anti-PD-1 and anti-CTLA-4 efficacy and delayed tumor establishment in mouse models.

Methodology

Multicohort study of 4,336 plasma samples from 1,714 patients in 16 cohorts, using targeted mass spectrometry and a DynForest joint longitudinal-survival model with variable-importance-based feature pruning. Training used IML1–SABR (n=105), validation used PANDORE (n=30), and histidine was checked in seven external cohorts with Cox and Kaplan–Meier analyses. Findings were followed up with immunophenotyping, fecal metagenomics and metabolomics, and histidine-supplementation experiments in tumor-bearing mice.

Study Limitations

The training set (n=105) and validation set (n=30) were small, and the validation AUC had a wide confidence interval. A clinical-only model performed similarly in validation, and the human associations are observational and may reflect nutrition or general health. Mouse effects were partly trends, and the full microbiome and dietary analyses were not available in the supplied text.

Enjoyed this summary?

Get the latest longevity research delivered to your inbox every week.

Enter your email to subscribe: