BTK inhibition drives mobilization and compartmental shifts of CLL subclones

Abstract

Abstract BACKGROUND Chronic lymphocytic leukemia (CLL) shows compartmentalization, whereby genetically or functionally distinct subclones of CLL cells are unevenly distributed among different anatomical sites, mainly the peripheral blood (PB), bone marrow, and lymph nodes (LN). Previous work demonstrated compartmentalization of genetically distinct subclones (Sun, 2023). BTK inhibitors (BTKi) mobilize CLL cells from LN into the PB, causing the characteristic treatment-induced rise in lymphocytosis. Here we investigated shifts in CLL compartmentalization upon initiation of a BTKi. METHODS Paired PB and LN samples were collected from patients starting treatment with a BTKi on study (NCT01500733, NCT02337829). Whole exome sequencing (WES), and bulk RNA sequencing was performed samples from 65 patients treated with either ibrutinib (n=35; baseline and 24h or 72h after first dose) or acalabrutinib (n=30; baseline and after 96±20h of treatment). Matched PB samples were collected within 3 months pre-treatment and on the same day as the on-treatment LN. Clonal architecture and cancer cell fractions (CCF) were analyzed using SuperFreq, with PyClone-VI used for validation and refinement. To compensate for variable tumor purity of the samples, the CCF of subclones was normalized to the CCF of the founding clone. Subclones with CCF >5% and shifts ≥10% were considered significant. Significance in CCF shifts was set ≥20% when LN impurity could skew results. River plots were generated to allow for visualization of clonal dynamics. Statistical analyses were performed in R (v4.2.1), with significance set at P<.05. RESULTS Pairs of PB and LN samples sequentially obtained at baseline and on BTKi treatment were analyzed for 19 patients, yielding 4 tumor samples per patient. In 13 (68%) patients, the LNs were from ipsilateral sites, and in 6 (32%) from contralateral sites. Common driver mutations included TP53 (n=9), NOTCH1 (n=8), and SF3B1 (n=6), including multi-hits. SuperFreq identified 2 to 6 subclones (median 3) per patient. Pre-treatment, 12 subclones showed PB bias, and 14 subclones had higher CCF in LN than PB. On the BTKi, of 61 total subclones tracked in 19 patients, 24 (39%) demonstrated a compartmental shift. Among 10 subclones without compartment bias at baseline, 8 (80%) shifted into PB and 2 (20%) were more common in the on-treatment LN than PB. Mobilization to PB, defined as appearance of a new subclone or increase of CCF ≥10% in PB, was observed for 5 subclones across 4 patients (all on day 4 of BTKi). In 2 patients the subclones with NOTCH1 mutations increased from 29%→39% and 61%→71%, respectively. Another case with a PB CCF increase of 4.2%→20% was associated with a gain of an ATM mutation in the subclone. Of the 19 patients, 4 had 1-2 subclones detected in the pre-treatment PB that decreased by CCF≥10% or became undetectable in the on-treatment PB. In these cases, the median on-treatment PB CCFs decreased compared to pre-treatment (median -15%, P=.0032), suggesting a brisk treatment effect towards select subclones. Not all subclones that increased in PB on-treatment were detected in the sampled LNs, consistent with putative heterogeneity across LN sites. Infact, in 6 cases (2 contralateral and 4 ipsilateral biopsy pairs), the pre-treatment LNs demonstrated subclonal shifts >|20%| when compared to on-treatment LNs. Ipsilateral sites often contained multiple LNs, indicating that different nodes may have been biopsied. In conclusion, these findings highlight marked heterogeneity between PB and LN sites, and suggest that each LN may function as its own unique compartment, at times harboring distinct CLL subclones that are not universally shared across disease sites. Many subclones shifted compartment bias on treatment, but not all LN resident subclones appeared to be mobilized. A deeper understanding of compartmentalization is critical for tracking clonal evolution and understanding treatment resistance.