Plasma nucleosome profiling reports on tumor burden and molecular subtypes in small cell lung cancer
Gavriel Fialkoff, Nobuyuki Takahashi, Israa Sharkia, Jenia Gutin, Nadav Hermoni, Michael Nirula, Rajesh Kumar, Lorinc Pongor, Samantha Nichols, Linda Sciuto, Kanak Parmar, Parth Desai, Priya Suresh, Melissa Abel, Rajaa El Meskini, Myriam Maoz, Yakir Rottenberg, Shoshan Nevo, Hovav Nechushtan, Tamar Peretz, Diana Roame, Ayala Hubert, Jonathan E. Cohen, Azzam Salah, Mark Temper, Albert Grinshpun, Zoe Weaver-Ohler, Arun Rajan, William Douglas Figg Sr., Aviad Zick, Ronen Sadeh, Nir Friedman, Anish Thomas
Gavriel Fialkoff, Nobuyuki Takahashi, Israa Sharkia, Jenia Gutin, Nadav Hermoni, Michael Nirula, Rajesh Kumar, Lorinc Pongor, Samantha Nichols, Linda Sciuto, Kanak Parmar, Parth Desai, Priya Suresh, Melissa Abel, Rajaa El Meskini, Myriam Maoz, Yakir Rottenberg, Shoshan Nevo, Hovav Nechushtan, Tamar Peretz, Diana Roame, Ayala Hubert, Jonathan E. Cohen, Azzam Salah, Mark Temper, Albert Grinshpun, Zoe Weaver-Ohler, Arun Rajan, William Douglas Figg Sr., Aviad Zick, Ronen Sadeh, Nir Friedman, Anish Thomas
View: Text | PDF
Clinical Research and Public Health Oncology Pulmonology

Plasma nucleosome profiling reports on tumor burden and molecular subtypes in small cell lung cancer

Abstract

BACKGROUND Small-cell lung cancer (SCLC) is an aggressive malignancy with a poor prognosis and marked transcriptional heterogeneity that may drive distinct therapeutic vulnerabilities. Clinical translation of molecular subtyping has been limited by restricted access to tumor biopsies, particularly at relapse.METHODS We applied chromatin immunoprecipitation of cell-free nucleosomes carrying active histone modifications followed by sequencing (cfChIP-seq) to 441 plasma samples from individuals with advanced SCLC, other neuroendocrine carcinomas, or non-SCLC cancers, as well as from healthy controls. Plasma cfChIP-seq profiles were integrated with matched tumor transcriptomes from 73 samples, including 41 time-matched pairs.RESULTS cfChIP-seq captured the epigenetic and transcriptional landscape of tumor-derived cell-free DNA (cfDNA), including SCLC tissue- and cell-of-origin signatures. A quantitative cfChIP-seq–derived SCLC score tracked radiographic tumor burden and was associated with prognosis. Signals at promoters of lineage-defining transcription factor genes, including ASCL1, NEUROD1, POU2F3, and ATOH1, correlated strongly with matched tumor RNA expression and supported noninvasive inference of SCLC transcriptional subtypes directly from plasma.CONCLUSION Plasma cfChIP-seq provides a practical liquid biopsy platform for real-time assessment of tumor burden, tumor state, and molecular subtype in SCLC. These findings support further development of cfChIP-seq for precision monitoring and subtype-informed therapeutic stratification in SCLC.TRIAL REGISTRATION ClinicalTrials.gov NCT02484404, NCT02487095, NCT02769962, NCT03554473, NCT03896503, and NCT02146170.FUNDING Center for Cancer Research; Intramural Program of the NCI (ZIA BC 011793); European Research Council (ERC) (Adg no. 101019560 “cfChIP”).

Authors

Gavriel Fialkoff, Nobuyuki Takahashi, Israa Sharkia, Jenia Gutin, Nadav Hermoni, Michael Nirula, Rajesh Kumar, Lorinc Pongor, Samantha Nichols, Linda Sciuto, Kanak Parmar, Parth Desai, Priya Suresh, Melissa Abel, Rajaa El Meskini, Myriam Maoz, Yakir Rottenberg, Shoshan Nevo, Hovav Nechushtan, Tamar Peretz, Diana Roame, Ayala Hubert, Jonathan E. Cohen, Azzam Salah, Mark Temper, Albert Grinshpun, Zoe Weaver-Ohler, Arun Rajan, William Douglas Figg Sr., Aviad Zick, Ronen Sadeh, Nir Friedman, Anish Thomas

×

Figure 2

SCLC plasma samples exhibit distinct cfChIP-seq signals that correlate with tumor burden and survival.

Options: View larger image (or click on image) Download as PowerPoint
SCLC plasma samples exhibit distinct cfChIP-seq signals that correlate w...
(A) Study outline. Plasma and tumor samples were collected from a cohort of healthy individuals and patients with SCLC at various time points during treatment. cfChIP-seq was performed on plasma samples and RNA-seq on tumor biopsy samples. Plasma cfChIP-seq–based gene expression profiles were benchmarked against tumor RNA-seq, and key SCLC transcription regulators were examined. (B) Genes with significantly high coverage in selected SCLC (n = 20) samples (see Methods) compared with healthy baseline values. For each gene, we compared mean normalized coverage in the SCLC samples (y axis) with a healthy cohort reference (x axis). A total of 3,197 genes with significantly higher coverage (q < 0.001) in at least 6 SCLC samples are marked (see Supplemental Table 4). Significance was determined as previously described (29). (C) Median SCLC score for different cohorts calculated by linear regression (see Methods). Each dot is a sample, and box plots summarize each group’s SCLC score distribution. The horizontal line represents the median, the box bounds represent the 25th and 75th percentiles, and the whiskers extend to the largest and smallest values within 1.5 times of the IQR. Pre/post refers to the timing of the collection relative to treatment. ****P < 0.0001, by Kruskal-Wallis test. (D) Principal components 1 and 2 of SCLC (orange) and healthy (green) samples. Transparency indicates SCLC score as in C. PCA was done using all Refseq genes (~25,000). This representative plot represents a single analysis of the cohort. (E) Correlation of the SCLC score with other plasma and imaging-based measures of tumor burden. Pearson correlation coefficients and P values were determined using the ggpubr stat_cor function. DELFI, DNA evaluation of fragments for early interception. (F) Changes in the SCLC score and radiological tumor burden in a patient with SCLC over the treatment time course (SCLC0191). Tr. 1, durvalumab and olaparib; Tr. 2, topotecan and M6620; RT, radiation therapy; Tr. 3, investigational therapy; CR, complete response; SD, stable disease; PD, progressive disease. (G) Kaplan-Meier survival plot for patients with an SCLC score above (red; median survival: 3 months) or below (blue; median survival: 8.5 months) the median. The survival curve was generated using pretreatment samples. P value was calculated by log-rank test. This survival analysis was performed once on the patient cohort.