Browse the corpus

Immune checkpoint inhibitors (ICIs) have transformed the treatment for many cancer patients, with a subset achieving durable remissions1,2. The major adverse effects of these agents are autoimmune, termed immune-related adverse events (irAEs), that can affect virtually any organ3. While cell-free DNA (cfDNA) assays are increasingly used to track responses to therapy, the detection of irAEs continues to rely almost entirely on clinical recognition. A laboratory-based test revealing the distribution and extent of immune-mediated tissue injury could prove informative. We here describe an approach to evaluate damage to nine different organs through the analysis of cfDNA in the blood.

, the detection of irAEs continues to rely almost entirely on clinical recognition. A laboratory-based test revealing the distribution and extent of immune-mediated tissue injury could prove informative. We here describe an approach to evaluate damage to nine different organs through the analysis of cfDNA in the blood. We evaluated 14 patients with solid tumors receiving ICIs, including 6 patients with irAEs and 8 patients who did not experience an irAE (Table S1, available with the full text of this letter at NEJM.org). Plasma samples were drawn at baseline before initiation of therapy, and between weeks 4 and 8 of treatment. cfDNA was prepared for methylation analysis and the same template molecules were also used for whole genome sequencing to identify tumor fractions in the plasma4. Tissue-specific methylation markers were then used to deconvolute the origin of cfDNA molecules and estimate the fraction shed from each organ5. Because each tissue in the body has a unique DNA methylation pattern, the tissue of origin of cfDNA fragments can be inferred through genome-wide methylation analysis. Only dead cells release DNA into the circulation, so damage to specific organs can be identified by a significant increase in tissue-specific cfDNA during treatment compared to baseline.

s a unique DNA methylation pattern, the tissue of origin of cfDNA fragments can be inferred through genome-wide methylation analysis. Only dead cells release DNA into the circulation, so damage to specific organs can be identified by a significant increase in tissue-specific cfDNA during treatment compared to baseline. We found that all six patients with clinically apparent irAEs had evidence of tissue damage before or at the time of diagnoses. To our surprise, our findings suggest the tissue damage was not restricted to the organs that were clinically involved, with evidence of multiorgan damage in all six patients, and a median increase of 6.1-fold in tissue-specific cfDNA in each patient (Figure 1 and Table S1). Evidence of tissue damage preceded clinical diagnoses of irAE in three of the six patients by 4, 20, and 236 days (Table S1). In contrast, none of the eight patients without clinically documented irAEs had evidence of multi-organ damage, though damage to a single organ (lung) was found in one patient (8167). Thus, a significantly higher fraction of patients with irAEs had molecular evidence of multi-organ damage compared to the patients who did not experience an irAE (6 of 6 vs. 0 of 8; P=0.0003, Fisher’s Exact Test). Though with such small numbers of patients, the strength of the conclusion may not be robust.

ient (8167). Thus, a significantly higher fraction of patients with irAEs had molecular evidence of multi-organ damage compared to the patients who did not experience an irAE (6 of 6 vs. 0 of 8; P=0.0003, Fisher’s Exact Test). Though with such small numbers of patients, the strength of the conclusion may not be robust. It is conceivable that the multiorgan damage was due to spread of the cancers in the patients who developed irAE. This was unlikely on clinical grounds as an increase of 6.1-fold in disease burden during the first 4–8 weeks of treatment (matching the increase in tissue-specific cfDNA) would be highly unusual. Moreover, according to RECIST criteria, only two of the six patients with irAEs developed progressive disease during the duration of the study (Table S1). Finally, to evaluate systemic disease burden with a molecular assay, we evaluated chromosome copy number changes in the plasma of all patients. Using this widely used measure of disease burden, we found that the burden of neoplastic disease increased only in a minority of patients (3 of 14), including two patients with and one without irAEs (Table S2). An increase in neoplastic disease was not correlated with either irAE or the extent of multi-organ damage (Table S2). Lastly, the organs showing cfDNA increases did not align with the patients’ clinical histories of metastatic disease or the expected patterns of metastatic spread based on the primary tumor sites (Table S2). The only exceptions were lung-specific signals observed in two patients (4648 and 4651) with known lung metastases. However, both patients demonstrated clinical responses to therapy and had low tumor fractions in plasma, making it unlikely that tumor progression accounted for the >10-fold increases in lung-specific cfDNA (Table S2).

tions were lung-specific signals observed in two patients (4648 and 4651) with known lung metastases. However, both patients demonstrated clinical responses to therapy and had low tumor fractions in plasma, making it unlikely that tumor progression accounted for the >10-fold increases in lung-specific cfDNA (Table S2). These studies provide a tool to help assess irAEs using molecular parameters. Our results challenge the traditional view of irAEs as isolated organ-specific toxicities and instead reveal a more diffuse and systemic loss of immune tolerance. The identification of multi-organ damage in all studied patients with irAEs suggests that the clinically documented irAEs represent only the tip of the iceberg with respect to the organ damage caused by treatment with ICIs. Further validation in a larger prospective cohort would be necessary to determine how other potential causes of muti-organ damage, such as sub-clinical infections or as yet unknown effects of immune checkpoint inhibitors, could confound the release of cfDNA in specific clinical scenarios. Additional time points and expanded methylation panels representing more tissue types would also be necessary to establish whether or how this test could be used to inform patient care.