Authors 1 -3: Contributed equally to this work
Authors 8-9: Shared last author
Background
Genetic material from aggressive B-cell lymphoma is typified by recurrent multiple somatic alterations. Given some of the disease's most profound characteristics-large tumor cell turnovers and its evident physical relation to the cardiovascular system-it presents itself as an excellent candidate for cell-free DNA (cfDNA) based research through liquid biopsies. Recent studies on cfDNA mainly focus on targeted sequencing for single nucleotide variation and translocation detection. This technique remains expensive, requires targeted panels and is yet to be clinically standardized. In contrast, the much cheaper shallow whole genome sequencing (sWGS) for copy number variation (CNV) detection, is operative at most hospitals that offer noninvasive prenatal testing (NIPT). Despite these benefits, the potential of this approach for lymphoma remains insufficiently explored.
Methods
Between May 2016 and June 2019, 129 patients (pts; median age of 59; Ann Arbor stage I (4%), II (21%), III (20%) and IV (55%)) with confirmed lymphoma (86 diffuse large B-cell lymphoma (DLBCL), 39 Hodgkin lymphoma (HL) and 4 grey zone lymphoma (GZL)) were recruited. Ultimately, 236 samples were collected, including liquid biopsies at initial diagnosis; follow-up liquid biopsies at relapse or surveillance (25 pts); and paired formalin-fixed paraffin-embedded (FFPE) tissue samples (35 pts). Follow- up samples were collected during the following evaluations: DLBCL pts were evaluated by Lugano 2014 criteria after 4 cycles of R- CHOP and at the end of treatment. HL pts were subject to an early interim evaluation after only 2 cycles of ABVD.
cfDNA was analyzed by sWGS (0.5x coverage), where resulting reads were mapped to both human and viral reference genomes, aiming at identifying nuclear CNVs and viral fragments, respectively. Additionally, the cohort was extended with 60 NIPT (liquid negative control) and 9 benign FFPE (tissue negative control) samples in order to statistically determine cancerous copy number profiles and abnormal viral titers.
Results
At staging, 36/44 (82%) of HL and 67/96 (70%) of DLBCL liquid biopsies had detectable CNVs. The remainder possibly had no structural aberrations, or, more likely, had an insufficient fraction of cell-free tumor DNA. Notably, no CNVs could be detected in 87% of HL FFPE samples, suggesting HL tumor DNA is more abundant in plasma. For DLBCL however, liquid copy number profiles were well-represented by their solid equivalent, indicated by an average Pearson correlation of 0.82.
In order to quantify copy number abnormality, we developed a statistic that increases with rising deviance from the overall healthy diploid state. This score, named the copy number profile abnormality (CPA) score, positively and significantly correlates with Ann Arbor stage (p=0.044), the International Prognostic Index (IPI; p=0.038) and metabolic tumor volume (MTV; p=0.0004), derived from PET/CT scans.
In 15 HL (34%), 18 DLBCL (19%) and 2 GZL (50%) samples, a statistically (q<0.01) augmented amount of plasmatic Epstein-Barr viral fragments was detected at staging. The performance of this method was compared to classical chromogenic in situ hybridization (CISH) as a gold standard and yielded sensitivity and specificity statistics of 0.89 and 0.94, respectively. Of interest, liquid biopsies positive for EBV had higher concentrations of viral fragments in comparison to matched solid tissue.
Detected CNVs at staging were often highly indicative for histological subtype. Leave-one-out cross-validation in combination with established modeling techniques shows that, eg, distinguishing HL from DLBCL, is characterized by an area under the curve of 0.9 during receiver operating characteristic analysis.
Amongst 129 pts, 12 did not go in sustained CR. Sequential samples of these pts were compared to 13 similar cases with surveillance samples in CR. In the latter, all CNVs disappeared, whilst this was not the case for the remaining bad responders (Fig1). In this group, longitudinal analysis revealed that relapse is essentially not associated with a change in CNV pattern.
Conclusion
Both CNVs and viral DNA fragments can be detected frequently and accurately in liquid biopsies of lymphoma pts. In two longitudinal cases, a rise in CNV parameters predicted relapse.
Deeren:Alexion, Amgen, Janssen, Roche, Sunesis, Takeda, Sanofi: Consultancy, Honoraria, Membership on an entity's Board of Directors or advisory committees, Research Funding.
Author notes
Asterisk with author names denotes non-ASH members.