Identification of proliferative and non-proliferative subpopulations of leukemic cells in CLL

Pathogenesis in chronic lymphocytic leukemia (CLL) is strongly linked to the potential for leukemic cells to migrate to and proliferate within lymph-nodes. Previous in vivo studies suggest that all leukemic cells participate in cycles of migration and proliferation. In vitro studies, however, have shown heterogeneous migration patterns. To investigate tumor subpopulation kinetics, we performed in vivo isotope-labeling studies in ten patients with IgVH-mutated CLL (M-CLL). Using deuterium-labeled glucose, we investigated proliferation in sub-populations defined by CXCR4/CD5 and surface (sIgM) expression. Mathematical modeling was performed to test the likelihood that leukemic cells exist as distinct sub-populations or as a single population with the same proliferative capacity. Further labeling studies in two patients with M-CLL commencing idelalisib investigated the effect of B-cell receptor (BCR) antagonists on sub-population kinetics. Modeling revealed that data were more consistent with a model comprising distinct sub-populations (p = 0.008) with contrasting, characteristic kinetics. Following idelalisib therapy, similar labeling suppression across all sub-populations suggested that the most proliferative subset is the most sensitive to treatment. As the quiescent sub-population precedes treatment, selection likely explains the persistence of such residual non-proliferating populations during BCR-antagonist therapy. These findings have clinical implications for discontinuation of long-term BCR-antagonist treatment in selected patients.


Supplementary Methods 1: Immunofluorescence Cell Sorting
Supplementary Table 1  Two mathematical models encapsulating two alternative hypotheses were constructed (Supplementary Figure 1).

Supplementary Figure 1. Parameterised models of CLL proliferation and recirculation
Tissue represents the non-blood sites of cell proliferation at rate p; solid arrows represent transitions between subpopulations denoted by rates, r1 … r5. Cell disappearance (primarily death) is represented by gray arrows with rates of dl, from the non-blood compartment, and db, from the blood compartment. The three blood phenotypes are defined as follows: -For Model B: (2) Where L is the number of CLL cells in the tissue and PBx is the number of CLL cells in the PBx compartment. The CLL cells proliferate with rate p, die with rate dl in the tissue, die with rate db in the blood and move between compartments with rates ri . (In Model B, we experimented with a different death rate db3 in the PB3 compartment but this never resulted in improved fits.) We assume all of these populations are at steady state for the duration of the experiment.
The equations describing the fraction of label in the different compartments are: -Model A: Where FL is the fraction of label in cells in tissue, while Fx is the fraction of label in the PBx compartment; U(t) is the precursor enrichment (modelled as a plateau during deuterium administration with an exponential decay afterwards) and b = 0.73. We have allowed a different exit rate ! ! * for cells that have recently divided in the lymph compartment. During the fitting process we restricted the relative sizes of the blood compartments, as each of them has been represented with a sample that is at least 10% of the total CLL blood compartment.
Furthermore, we restricted the size of the CLL population in the tissue to be at least as big as the size of the CLL population in the blood compartment.
The two models given by (3) and (4)

Reproducibility of GC-MS measurements
Each sample was analysed at least in quadruplicate after having been abundance-matched by dilution or concentration so as to eliminate artefactual effects of sample size on measured isotope ratios. The standard deviations of each set of replicate measurements were collated according to cell sorting strategy (Table S2). N is the number of independent samples analysed in each group. The median SD is expressed in the same units as the enrichment data, i.e. normalised to the fraction of new cells equivalent to a one-day labeling period (F, %/day) by dividing by the area under curve of the measured glucose enrichment data during labeling.

Supplementary
In terms of minimal detectable enrichments. taking a typical SD of 0.03 gives a 90% chance of detecting a difference equivalent to 0.07 %/d between two samples with n=4 measurements for each sample. Since multiple measurements are made along the labeling curve, the minimal detectable proliferation rate in CLL cells is likely to be less than this figure. Typical peak enrichments in this study are of the order of 0.20-0.32 %/day equivalent (1st to 3rd IQR).

Supplementary Figures
Supplementary Figure 2

. Detail of first seven days of deuterium labelling in CD19 + leukemic cells
Data as in Figure 3 but only for days 0-7.