Specificity & Precision of Minimal Residual Disease Monitoring in DLBCL Using Ig-HTS

Background: Dynamic minimal residual disease (MRD) monitoring in the diffuse large B-cell lymphoma (DLBCL) setting has been shown to predict treatment failure at early timepoints and relapse at later timepoints. Precise and reliable approaches are critically important for accurate detection of residual disease. We sought to evaluate the precision of heavy-chain (IGH) and light-chain (IGK/L) MRD measurement within a standard clinical workflow.

Methods: We developed a database of clinical clonoSEQ (Adaptive Biotechnologies) and clonoSIGHT reports (Sequenta, Inc.) identifying the dominant malignant IGH or IGK/L clonotypes from the tumor biopsy samples of 285 patients collected as a part of routine clinical care for MRD monitoring. Included patients had been diagnosed with DLBCL at either Stanford University or the National Cancer Institute, 130 of whom had samples sequenced with the clonoSEQ assay and 155 with the clonoSIGHT assay. A control database was generated from all publicly available immunoSEQ (Adaptive Biotechnologies) B-cell repertoires with IGH or IGK/L sequencing, spanning both malignant and healthy patient contexts. The patients within the clinical DLBCL and control databases were distinct without any overlap. We assessed whether each malignant IGH or IGK/L clonotype from the clinical DLBCL database shared overlap with clonotypes present in the control IGH or IGK/L repertoires. A match was considered positive only if the sequence context of the malignant clonotype was an exact match with the sequence of a control clonotype; exact substrings were allowed given the differences in potential primer sites between assays.

Results: The clinical database included 341 dominant IGK/L sequences from 187 patients and 271 dominant IGH sequences from 195 patients, where dominance was determined by the clinical assay. We generated a control IGK/L database of 618,775 sequences from 71 subjects and a control IGH database of 65,017,932 sequences from 1698 subjects. IGH demonstrated high precision (99.3%) with only two clinically dominant sequences demonstrating overlap with the control repertoire. Despite the smaller control database, IGK/L demonstrated significantly reduced precision of 92% (Fisher’s exact p-value < 0.001) with 29 sequences demonstrating overlap with the control repertoire, all of which were limited to the subset of patients who had been sequenced by clonoSEQ. At the patient level, 19 of the 105 patients with clonoSEQ IGK/L sequencing had clinically trackable (Fig. 1A), putatively malignant sequences that were shared with the control database (precision 84.7%). The frequency of these clinically dominant sequences in the control repertoires ranged from 8.27x10^-8 to 0.575 with a mean frequency of 0.000598. Most non-unique clonotypes were shared across multiple control subjects (range 1-37, mean 13, Fig. 1B).

Conclusions: The high rate of overlap seen between putatively malignant, dominant IGK/L clonotypes and a publicly available control repertoire suggests substantially reduced validity of approaches relying on IGK/L tracking for minimal residual disease monitoring. The overlapping light chain clonotypes were not only present at relatively high frequencies in both the clinical samples and control repertoire, but were also seen across multiple control subjects, suggesting MRD assays that track only IGK/L may be unable to reliably differentiate malignant clonotypes from background light chain repertoires. Caution should be used when MRD is detected using IGK/IGL given reduced specificity and precision of these sequences.