Spina, Bruscaggin, Cuccaro, et al. (2018) — Circulating Tumor DNA Reveals Genetics, Clonal Evolution, and Residual Disease in Classical Hodgkin Lymphoma

Bibliographic Reference

Spina, V., Bruscaggin, A., Cuccaro, A., Martini, M., Di Trani, M., Forestieri, G., Manzoni, M., Condoluci, A., Arribas, A., Terzi-Di-Bergamo, L., Locatelli, S. L., Cupelli, E., Ceriani, L., Moccia, A. A., Stathis, A., Nassi, L., Deambrogi, C., Diop, F., Guidetti, F., … Rossi, D. (2018). Circulating tumor DNA reveals genetics, clonal evolution, and residual disease in classical Hodgkin lymphoma. Blood, 131(22), 2413—2425. https://doi.org/10.1182/blood-2017-11-812073

Core Argument

Classical Hodgkin lymphoma (cHL) has lagged behind non-Hodgkin lymphomas in genetic characterization due to two major technical obstacles: the rarity of neoplastic Hodgkin and Reed-Sternberg (HRS) cells in tumor biopsies (often <5% of the cellularity) and the routine formalin fixation that degrades DNA quality. The authors demonstrate that circulating tumor DNA (ctDNA) from plasma, profiled by a highly sensitive and robust deep next-generation sequencing approach (CAPP-seq), overcomes both hurdles. ctDNA mirrors HRS cell genetics with 87.5% sensitivity for biopsy-confirmed mutations, enabling noninvasive genotyping, longitudinal tracking of clonal evolution under different treatment modalities, and monitoring of residual disease during multiagent chemotherapy. The study identifies STAT6 as the most frequently mutated gene in cHL (~40% of cases) — a finding missed by prior exome sequencing of microdissected HRS cells — and proposes ctDNA quantification as a radiation-free complement to interim PET/CT for early identification of chemorefractory patients. Collectively, the results position ctDNA as a precision medicine biomarker in a disease where limited tissue access has historically constrained molecular characterization.

Methods

Study design and patients. Retrospective observational design. 80 newly diagnosed and 32 relapsed/refractory cHL patients were included. All relapsed/refractory patients had failed both first-line therapy and autologous transplant salvage. 349 blood and tissue samples were analyzed in total.

Biological material. Cell-free DNA (cfDNA) was isolated from plasma collected at diagnosis (pre-treatment, n=80), during ABVD chemotherapy (course 2 day 1 interim and end-of-treatment; n=24), at refractory progression (n=32), before and after autotransplant salvage (n=6), before and after brentuximab vedotin (n=5), and before and during nivolumab (n=5). Tumor gDNA from microdissected HRS cells and normal germline gDNA from peripheral blood granulocytes were analyzed for comparison.

CAPP-seq. A targeted resequencing panel of 77 genes (191,271 bp target region) recurrently mutated in mature B-cell tumors was designed and applied to cfDNA using the CAncer Personalized Profiling by deep Sequencing (CAPP-seq) approach — a validated ultra-deep NGS method with sensitivity of ~10^-3. Multiplexed libraries were sequenced on Illumina platforms (NextSeq/MiSeq) with 150- or 300-bp paired-end reads, achieving a mean coverage of 49,413 for ctDNA.

Validation. In a discovery set of 15 patients, plasma ctDNA genotyping was compared paired tumor gDNA from microdissected HRS cells (mean coverage: 17,333) blinded to the ctDNA results. Areas devoid of HRS cells were also sequenced to confirm mutation origin. Hotspot STAT6 mutations were independently validated by allele-specific PCR. Results from independent duplicate experiments showed high concordance (R^2 = 0.978).

Imaging. PET/CT scans were performed at baseline, after 2 ABVD courses (interim), and at end of treatment, or during nivolumab therapy. Central blind analysis by trained nuclear physicians used Deauville criteria for chemotherapy and LYRIC criteria for nivolumab.

Statistical analysis. Progression-free survival and overall survival by Kaplan-Meier. Unsupervised hierarchical clustering (Euclidean distance, complete method). Categorical variables by chi-square/Fisher’s exact test; continuous by Mann-Whitney U test. Two-sided P < 0.05. Multiple comparison correction by false discovery rate.

Key Findings

• ctDNA mirrors HRS cell genetics with 87.5% sensitivity (84/96 biopsy-confirmed mutations detected; 95% CI: 79.2%–92.8%). Variants detected in ctDNA but absent in microdissected HRS cells were bona fide neoplastic per their molecular profile and disappearance on disease remission, and were attributed to subclonal or anatomical heterogeneity of the tumor. No mutations were detected in biopsy areas devoid of HRS cells, confirming mutation origin from neoplastic cells.
• STAT6 is the most frequently mutated gene in cHL, affecting 37.5% of newly diagnosed patients. Recurrently mutated genes at >=20% frequency also included TNFAIP3 (35.0%) and ITPKB (27.5%). Mutational profiles converged on six pathways: NF-kB (46.2%), PI3K/AKT (46.2%), cytokine signaling/STAT6 (37.5%), epigenetic regulators (35.0% cumulative), immune surveillance/immune evasion (27.5%), and NOTCH signaling (20.0%).
• Clonal evolution under treatment is the rule: all 13 patients with longitudinal diagnosis/relapse samples showed clonal shifts. Under chemotherapy and brentuximab vedotin, diagnosis/relapse tumor pairs branched through acquisition of phase-specific mutations from an ancestral clone that persisted throughout the disease course — STAT6, GNA13, ITPKB, and TNFAIP3 mutations were preferentially inferred in the ancestral clones, indicating they are early events. Under nivolumab, ancestral clones were cyclically suppressed and replaced by novel clones harboring new mutations, interpreted as drug-promoted aggression against neoantigens and tumor evasion by generating new mutations.
• A 2-log (100-fold) drop in ctDNA after 2 ABVD courses was the optimal cutoff for predicting complete response and cure; patients achieving <2-log drop had inferior progression-free survival (all events were progressions). ctDNA quantification resolved inconsistencies in interim PET/CT: cured patients who were interim PET/CT-positive had >2-log ctDNA drops, while relapsing patients who were interim PET/CT-negative had <2-log drops. The lack of correlation between log-fold change of maximum SUV and log-fold change of ctDNA (baseline to interim) is consistent with SUV reflecting metabolic activity of the inflammatory microenvironment whereas ctDNA reflects tumor load.
• ctDNA enabled noninvasive genotyping of 81.2% of newly diagnosed cHL patients (average 5 nonsynonymous mutations/case; mean allele frequency 5.5%, range 0.29%–74.0%). Pretreatment ctDNA concentration correlated with Ann Arbor stage and prognostic group, indicating ctDNA as a surrogate marker of tumor load. Histologic subtypes showed different genetic profiles: STAT6 and TNFAIP3 mutations were enriched in nodular sclerosis cHL and in EBER-negative cases.

Concepts Introduced or Used

• ctDNA genotyping — Use of circulating tumor DNA from plasma as a noninvasive surrogate for tumor biopsy DNA, applied here to classical Hodgkin lymphoma for the first time at scale with a validated ultra-deep sequencing approach. Relevant to clonal-evolution monitoring and intratumor-heterogeneity assessment.
• Clonal evolution under therapy — The demonstration that chemotherapy and brentuximab vedotin leave ancestral clones intact while immunotherapy (nivolumab) suppresses ancestral clones and drives rapid emergence of novel clones. This differential pattern across treatment modalities is directly relevant to therapy-resistance and population-bottleneck dynamics.
• Molecular residual disease — ctDNA quantification as a complement to CT imaging for early identification of chemorefractory patients, addressing the ~20-30% discordance between interim PET/CT results and final outcome.
• Ancestral reservoir — The persistence of an ancestral clone through chemotherapy and brentuximab vedotin, serving as a reservoir that propagates successive disease relapses, a concept formalized in the clonal evolution literature on therapy-resistance.
• Clonal divergence — The fraction of mutations unique to baseline vs. relapse time points (Figure 6D) defining evolutionary distance between sequential tumor pairs, relevant to branching-evolution and intratumor heterogeneity.

Entities Referenced

• STAT6 — Signal transducer and activator of transcription 6; identified as the most frequently mutated gene in cHL (~38% of cases), with hotspot mutations c.1249A>T and c.1255G>A validated by allele-specific PCR. pSTAT6 nuclear expression in HRS cells confirmed functional consequences.
• TNFAIP3 (A20) — Tumor suppressor gene in the NF-kB pathway; mutated in 35.0% of cHL cases.
• ITPKB — Inositol-trisphosphate 3-kinase B; mutated in 27.5% of cHL cases.
• B2M — Beta-2-microglobulin; component of MHC class I; B2M-mutated cases showed absent MHC-I expression on HRS cells, confirming the functional impact of mutations discovered in ctDNA.
• GNA13 — G protein subunit alpha 13; in the list of genes preferentially inferred in ancestral clones.
• ABVD — Chemotherapy regimen (adriamycin, bleomycin, vinblastine, dacarbazine); standard first-line treatment for advanced cHL studied here.
• Brentuximab vedotin — Anti-CD30 antibody-drug conjugate; salvage therapy for relapsed/refractory cHL.
• Nivolumab — Anti-PD-1 immune checkpoint inhibitor; salvage immunotherapy for relapsed/refractory cHL.
• CAPP-seq — CAncer Personalized Profiling by deep Sequencing; the ultra-deep NGS method used for ctDNA analysis.
• Hodgkin and Reed-Sternberg (HRS) cells — The neoplastic cells in cHL, typically <5% of biopsy cellularity, whose rarity historically limited genetic characterization.
• PET/CT — Positron emission tomography/computed tomography; standard imaging modality for cHL response assessment, with interim PET/CT having ~20-30% discordance with final outcome.

Limitations (as stated by authors)

• Targeted resequencing is underpowered compared with exome or whole-genome sequencing for discovering new candidate genes.
• The target region did not cover the SOCS1 gene, which may contribute to cytokine signaling pathway mutations not captured by the assay.
• Alternative genetic mechanisms not covered by the panel — such as B2M gene locus deletion or mutation/deletion of HLA genes — may contribute to MHC-I loss in B2M wild-type cases.
• The PI3K/AKT pathway was affected in 46.2% of patients, consistent with preclinical evidence of addiction, but the study did not include the full spectrum of genetic lesions affecting this pathway.
• The mutation profiles of newly diagnosed and refractory cHL were largely overlapping, suggesting that gene mutations not covered by the target region or molecular mechanisms not captured by CAPP-seq may contribute to the chemorefractory phenotype.
• Potential contribution of clonal hematopoiesis (selected by chemotherapy) to ctDNA mutations was ruled out by analysis of paired granulocytes in this study, but the authors flag this as a general concern for ctDNA studies (citing Busque et al., 2012).
• The study is retrospective and observational; the ctDNA cutoff for predicting outcome (2-log drop after 2 ABVD courses) needs prospective validation.
• The incorporation of both PET/CT and ctDNA monitoring into clinical trials was noted as necessary to precisely define their cumulative sensitivity and specificity.

Relevance to Clonal Evolution

This paper provides one of the clearest demonstrations of treatment-dependent clonal evolution patterns in a human lymphoma, distinguished by the use of ctDNA rather than tissue biopsies to track subclonal dynamics longitudinally.

Two evolution modes, two bottleneck severities. The central evolutionary finding is that chemotherapy/brentuximab vedotin and immunotherapy (nivolumab) produce qualitatively different clonal trajectories. Under chemotherapy, ancestral clones persist and acquire new mutations at relapse — a shallow bottleneck that preserves the tumor’s phylogenetic trunk (shared STAT6, GNA13, ITPKB, TNFAIP3 mutations). The inference that these truncal mutations are early events and survive chemotherapy places cHL in the same framework as other therapy-resistant lymphomas (Okosun et al., 2014, cited in the paper): a reservoir ancestral population resists chemotherapy and propagates successive relapses. This is a direct instance of therapy-resistance operating through pre-existing clonal persistence — the ancestral clone carries inherent or acquired fitness that allows it to survive the selective pressure of genotoxic therapy.

Under nivolumab, by contrast, ancestral clones are “cyclically suppressed and replaced by novel clones” — a deep bottleneck that eliminates the phylogenetic trunk and forces the tumor to re-diversify. The authors interpret this as “drug-promoted aggression against cancer neoantigens stemming from mutations and the attempt made by the tumor to evade treatment by generating new mutations.” This pattern maps onto the population-bottleneck framework: a deep bottleneck destroys the existing compression (the ancestral clone’s genomic program was successfully recognized by the immune checkpoint blockade), forcing survivors to re-enter exploration phase and generate new compressions (novel clones with new mutations). The rapidity of this re-diversification — occurring within months under nivolumab — was described as “unexpected in cHL” and mirrors the branching relapse architecture documented after deep bottlenecks in myeloma (Miething, 2019).

ctDNA as an evolutionary monitoring tool. The paper establishes ctDNA as a noninvasive window into these dynamics, enabling longitudinal sampling at multiple time points that tissue biopsies cannot feasibly provide. This technical contribution is significant for clonal evolution research: it opens the possibility of tracking evolutionary trajectories in real time during treatment, rather than inferring them from static diagnosis/relapse pairs. The demonstration that ctDNA allele frequencies track clonal shifts and can distinguish ancestral from newly acquired mutations is a methodological proof-of-concept for the broader field of evolutionary tracking in cancer.

Complement to imaging. The finding that ctDNA quantification resolves inconsistent interim PET/CT results — where metabolic activity of the inflammatory microenvironment can mask or mimic tumor burden — addresses a recognized clinical gap in cHL management. From an evolutionary perspective, this matters because ctDNA measures the evolving tumor population directly (tumor load), while PET/CT measures a proxy (metabolic activity of the inflammatory microenvironment that surrounds HRS cells). The dissociation between these two signals (Figure 7A: no correlation between log-fold change of maximum SUV and log-fold change of ctDNA; supplemental Figure 11) means that evolutionary dynamics may be happening beneath the resolution of imaging — a concern that extends beyond cHL to any cancer where the tumor microenvironment dominates the imaging signal.