To determine the role of CD49d for response to Bruton's tyrosine kinase inhibitors (BTKi) in patients with chronic lymphocytic leukemia (CLL).
In patients treated with acalabrutinib (n = 48), CD49d expression, VLA-4 integrin activation, and tumor transcriptomes of CLL cells were assessed. Clinical responses to BTKis were investigated in acalabrutinib- (n = 48; NCT02337829) and ibrutinib-treated (n = 73; NCT01500733) patients.
In patients treated with acalabrutinib, treatment-induced lymphocytosis was comparable for both subgroups but resolved more rapidly for CD49d+ cases. Acalabrutinib inhibited constitutive VLA-4 activation but was insufficient to block BCR and CXCR4–mediated inside–out activation. Transcriptomes of CD49d+ and CD49d− cases were compared using RNA sequencing at baseline and at 1 and 6 months on treatment. Gene set enrichment analysis revealed increased constitutive NF-κB and JAK–STAT signaling, enhanced survival, adhesion, and migratory capacity in CD49d+ over CD49d− CLL that was maintained during therapy. In the combined cohorts of 121 BTKi-treated patients, 48 (39.7%) progressed on treatment with BTK and/or PLCG2 mutations detected in 87% of CLL progressions. Consistent with a recent report, homogeneous and bimodal CD49d-positive cases (the latter having concurrent CD49d+ and CD49d− CLL subpopulations, irrespective of the traditional 30% cutoff value) had a shorter time to progression of 6.6 years, whereas 90% of cases homogenously CD49d− were estimated progression-free at 8 years (P = 0.0004).
CD49d/VLA-4 emerges as a microenvironmental factor that contributes to BTKi resistance in CLL. The prognostic value of CD49d is improved by considering bimodal CD49d expression.
High CD49d expression in patients with chronic lymphocytic leukemia (CLL) was linked to reduced treatment-induced lymphocytosis and inferior response to the first-in-class Bruton's tyrosine kinase inhibitor (BTKi), ibrutinib. Here, we confirm and extend these findings in patients treated with acalabrutinib, a more selective BTKi. Acalabrutinib inhibited constitutive VLA-4 activation but was insufficient to block microenvironment mediated inside–out activation. The transcriptome of CD49d+ cases revealed increased constitutive pro-survival and activation signaling over CD49d− cases that was maintained during acalabrutinib treatment. For both ibrutinib- and acalabrutinib-treated patients, CD49d was a strong prognostic factor. Notably, cases with any CD49d expression on CLL cells had inferior progression-free survival compared with cases entirely negative for CD49d. At progression, BTK and/or PLCG2 mutations were detected in most patients. In conclusion, CD49d expression increases the risk of developing BTKi resistance, serves as a strong prognostic factor, and is a possible target for combination therapy.
Chronic lymphocytic leukemia (CLL) is a malignancy of mature clonal B cells that accumulate in the blood, bone marrow, and lymphoid tissues. Host factors in the tissue microenvironment trigger B-cell receptor (BCR) activation to support growth and survival of CLL cells. Targeting Bruton's tyrosine kinase (BTK) downstream of the BCR has emerged as a successful treatment strategy for CLL (1). Ibrutinib, the first-in-class BTK inhibitor (BTKi), was found to improve progression-free survival (PFS) in multiple randomized trials and is approved for all lines of therapy in CLL (2). More recently, in head-to-head comparisons, acalabrutinib and zanubrutinib showed comparable or superior clinical efficacy with ibrutinib (3, 4). Most clinical studies of BTKis have used continuous dosing as tolerated or until disease progression. Factors associated with inferior PFS include history of prior therapy for CLL, presence of TP53 alterations, high beta-2-microglobulin (β2M), and elevated lactate dehydrogenase (LDH; ref. 5). In addition, CD49d expression has been associated with shorter PFS in patients treated with ibrutinib (6, 7). In most patients progressing on ibrutinib, mutations altering cysteine 481 (C481) in the active site of BTK and/or mutations in phospholipase C-gamma 2 (PLCG2) are identified (8, 9).
BTK also plays a fundamental role in BCR- and chemokine-controlled homing and adhesion of CLL cells to the microenvironment (1). Treatment with BTKis often results in a transient increase in lymphocytosis following treatment initiation (10). This increase has been attributed to the efflux of CLL cells from lymphoid organs due to impaired interaction with the microenvironment (11, 12). Within the microenvironment, CD49d, the α4-subunit of the integrin very late antigen 4 (VLA-4) on CLL cells, binds to vascular cell adhesion molecule-1 (VCAM-1) and fibronectin and mediates both cell–cell and cell–matrix interactions (13, 14). These interactions play a crucial role in the migration and retention of CLL cells in lymphoid tissues, where cells receive survival-supporting signals that protect them from drug-induced apoptosis (13, 14). VLA-4 can be rapidly activated by BCR signals via inside–out activation, resulting in changes in VLA-4 conformation (affinity) and promoting CLL cells' adhesive capacities. In vitro and in vivo inhibition of BCR signaling by ibrutinib has been shown to inhibit VLA-4 integrin function and impair the adhesive properties of CLL cells (11, 12). Interestingly, in ibrutinib-treated patients, high CD49d expression (more or equal to 30% positive cells cutoff value) was correlated with reduced lymphocytosis and nodal responses (6, 15). Moreover, in cells that highly expressed CD49d, ibrutinib was unable to fully inhibit BCR-mediated VLA-4 activation and cell adhesion (6).
CD49d is expressed in around 40% of CLL cases, with a very heterogeneous expression among patients. CD49d expression in CLL is associated with a poor outcome, and it has been recognized as the strongest flow cytometry–based prognostic marker in CLL (16, 17). However, the underlying biological mechanisms of how CD49d expression affects responses to BTKis and its possible contribution to acquired drug resistance remains elusive. A better understanding of these mechanisms could improve and guide treatment strategies to overcome the negative impact of CD49d on treatment response.
Here, to investigate the role of CD49d in response to BTKi treatments, we first analyzed the changes in CD49d expression and activation in patients with CLL during acalabrutinib treatment. To better understand the biological impact of CD49d expression on CLL cells, we performed a transcriptome analysis comparing gene expression changes between CD49d groups during acalabrutinib therapy.
Patients and Methods
Patients and study design
Patients with relapsed/refractory and high-risk treatment-naïve CLL were enrolled in phase II, single-center study using acalabrutinib (NCT02337829; ref. 18). Written informed consent was obtained in accordance with the Declaration of Helsinki, applicable federal regulations, and requirements from the local Institutional Review Board. Patients (n = 48) were randomized to receive acalabrutinib 200 mg every 24 hours (qd; n = 24) or 100 mg every 12 hours (bid; n = 24). Clinical results and pharmacodynamic measurements have been previously reported (18, 19). Characteristics of 48 patients included in these correlative studies are summarized in (Supplementary Table S1). Peripheral blood (PB) samples were collected at different timepoints prior or during acalabrutinib treatment. Peripheral blood mononuclear cells (PBMC) were isolated using lymphocyte separation medium (ICN Biomedicals) and viably frozen in 90% FBS and 10% dimethyl sulfoxide (Sigma) in liquid nitrogen. CD49d expression levels were reported as a percentage of CD49d-positive cells within CLL cells compared with (fluorescence minus one) FMO and isotype negative controls using flow cytometry. Patients with CLL were divided into CD49d-positive (CD49d+, ≥30% of CLL cells express CD49d) and CD49d-negative (CD49d−, <30% of CLL cells express CD49d) groups based on the established 30% cutoff value, as assessed by flow cytometry (15, 20). Another classification was based on the CD49d expression patterns as previously reported (7): Homogenous-positive CD49d (homCD49d+), ≥30% of CLL cells express CD49d uniformly; homogenous-negative CD49d (homCD49d−), <30% CLL cells express CD49d uniformly; bimodal CD49d (bimCD49d), samples with two distinct cell populations, one completely negative, with a fluorescence signal identical to that of the negative control, and the other positive, characterized by a fluorescence signal completely above the cutoff value.
Absolute lymphocyte counts (ALC) counts were collected before treatment and at different timepoints during acalabrutinib treatment. CD49d expression and outcome data were also available for 73 patients treated with ibrutinib (NCT01500733; ref. 21). Clinical and biological features of this cohort are summarized in (Supplementary Table S2). Representativeness of study participants is summarized in Supplementary Table S3. Time to progression (TTP) was defined as the time from the start of acalabrutinib or ibrutinib until progression. Cancer cell fractions of BTK and PLCG2 mutations at progression were assessed by digital PCR as previously described (Supplementary Table S4; ref. 8). Data on 121 patients combined from the two cohorts were available to compute the four-factor prognostic index (CLL-4), including relapsed/refractory (RR) CLL status, presence of TP53 alteration, β2M >5 mg/L, and LDH >250 U/L (Supplementary Tables S1 and S2). On the basis of the CLL-4 score, patients were divided into 3 groups: Low (0–1), intermediate (2), and high (3–4) risk, and TTP was estimated by the Kaplan–Meier method. Patients with progressive disease were counted as events; study discontinuation for any reason, including death, was censored. The log-rank test was used to compare TTP probabilities between subgroups. Extended multivariate analysis was carried out to assess the association between risk factors (including CD49d) and TTP using Cox proportional hazards regression.
PBMCs from CLL samples were stained with anti-CD19 APC, anti-CD5 PE-Cy7, and anti-CD3 APC-H7 antibodies (BD Biosciences) to identify CLL cells. Cells were pre-incubated with Aqua Blue dead cell exclusion dye (Invitrogen), followed by surface staining using anti-CD49d PE antibody and acquired on LSRFortessa (BD Biosciences) using FMO and isotype controls. Data were analyzed using FlowJo (Version 10; TreeStar), and were reported as a percentage of positive cells within the CLL population.
VLA-4 activation assay
Analysis of VLA-4 activation was evaluated as described previously (6, 22). Briefly, PBMCs (1×105 per experimental condition) were incubated for 30 minutes at 37°C with different concentrations of unlabeled VLA-4-specific ligand (LDV, R&D Systems) in the presence of the conformationally sensitive β1 integrin anti-CD29 PE (clone HUTS-21) commercial antibody (BD Biosciences), that specifically recognizes the hybrid domain that is exposed in the activated and ligand-occupied VLA-4 conformation. Cells were also stained with anti-CD19 APC, anti-CD5 PE-Cy7 commercial antibodies (BD Biosciences), and DAPI (Sigma) during the last 15 minutes of incubation and acquired on LSRFortessa. The mean fluorescence intensity values of labeled HUTS-21 mAbs were plotted versus the concentration of the LDV to generate sigmoidal dose–response curves. The receptor occupancy (RO) values were determined [ranging from 0.0 (no RO) to 1.0 (100% RO)] using a sigmoidal dose–response equation with variable slope. The EC50 values of the HUTS-21–binding curve provide a measure of the VLA-4 ligand-binding affinity. RO was used to measure binding affinity (at 10 nmol/L LDV), higher RO with a larger fraction of ligand-occupied receptors at this concentration indicates higher affinity (22). When indicated, cells were treated with F(ab)2 anti-IgM (10 μg/mL) to stimulate the BCR (Jackson ImmunoResearch Laboratories), CXCL12 (0.2 μg/mL) to stimulate the CXCR4 chemokine receptor (R&D systems), and Manganese chloride (MnCl2; 3 mmol/L) as a technical positive control to induce maximal integrin activation. When indicated, cells were treated, in vitro, with acalabrutinib (1 μmol/L), or duvelisib (1 μmol/L), for 1 hour at 37°C.
Previously reported RNA sequencing data from patients treated with acalabrutinib (Gene Expression Omnibus GSE136634) were used to analyze changes in gene expression between CD49d groups during acalabrutinib therapy (18, 23). Custom gene set were extracted from the Staudt laboratory database (https://lymphochip.nih.gov/signaturedb/ version of 03/09/2020). Differential expression (DE) analysis was conducted in R, version 4.1.1, using the DESeq2 package (24). Adjusted P values for each DE gene were based on Benjamini–Hochberg correction model and this was used to adjust for the number of comparisons. DE gene cutoff value was based on a P value of ≤0.05 and log2FC ≥ |2| for each of the three time points. The model design used for DESeq2 is a generalized linear model where each patient and time were used as covariates. DE analyses were performed on the raw reads count table for each time point between the CD49d+ and CD49d− groups. Pre-ranked gene set enrichment analysis (GSEA; ref. 25) was performed using a curated set of lymphocyte gene-expression signatures (23, 26).
Statistical analysis and mathematical modeling
The Mann–Whitney U test was used to calculate statistical significance between the different patient groups. To compare measurements in individual patients across time or different treatment conditions, the Wilcoxon matched-pairs signed-rank test was used. Categorical data were compared with the Fisher's exact test. All statistical analyses were performed using the statistical package GraphPad Prism version 9 (GraphPad Software, Inc.).
Conduct and primary results of the ibrutinib (21, 27) and acalabrutinib (18) clinical studies have been previously reported. The RNA-seq data analyzed in this study were previously reported and are available in the Gene Expression Omnibus data repository (accession number GSE136634; ref. 18). The data supporting the findings of the current study are available within the article and its Supplementary Data.
Additional methods are provided in the Supplementary Data.
Redistribution lymphocytosis during acalabrutinib treatment in CD49d+ and CD49d− CLL
To evaluate the impact of CD49d expression on acalabrutinib-induced redistribution lymphocytosis, we stratified 48 acalabrutinib-treated patients based on CD49d expression, at baseline. According to the clinically validated 30% cutoff value (15, 20), CLL was classified as CD49d+ in 27 patients (56%) and CD49d− in 21 patients (44%; Fig. 1A; Supplementary Table S1). Before treatment initiation, patients in both subgroups had comparable ALC (Supplementary Fig. S1A, median ALC: CD49d+, 51 × 106/mL; CD49d−, 49.6 × 106/mL, P = 0.9). As previously reported for ibrutinib-treated patients (6, 7), acalabrutinib-induced lymphocytosis differed between the two groups. During the first 4 months of treatment, treatment-induced lymphocytosis was more prominent in CD49d− CLL than in CD49d+ CLL (Fig. 1B).
As we previously reported, the onset of CLL cell release from tissue compartments is virtually immediate after initiation of ibrutinib, and, peak ALC is reached within the first 24 hours in many patients (10). Here, with measurements on days 3, 14, and 28 of cycle 1, we observed a 66% median peak increase in ALC in CD49d+ and 67% in CD49d−, and 48% of patients reached peak ALC on day 3 (Supplementary Fig. S1B). The effect of CD49d on these early changes in treatment-induced lymphocytosis has not been analyzed. Surprisingly, there was no difference in the rise in ALC between CD49d+ and CD49d− groups at early timepoints (Fig. 1C, median change: day 3, CD49d+ 48% vs. CD49d− 42%, P = 0.8; day 14, CD49d+ 28% vs. CD49d− 50%, P = 0.4). However, at one month, the increase in ALC was significantly lower for CD49d+ CLL (median change, 6%) than for CD49d− CLL (median change, 50%; Fig. 1C, P = 0.02). In summary, the degree of the early treatment-induced rise in ALC was comparable between the CD49d+ and CD49d− groups. Resolution of lymphocytosis was more rapid for patients with CD49d+ CLL and after 6 months of acalabrutinib therapy, the median ALC was significantly lower in the CD49d+ group (12 × 106/mL) compared with the CD49d− group (26 × 106/mL; Supplementary Fig. S1C, P = 0.03). In contrast with reports in patients treated with ibrutinib (6), the reduction in lymphadenopathy and spleen size was not significantly different between the two patient groups, after 6, 12 or 24 months of therapy (Supplementary Fig. S1D and S1E).
Effect of in vivo acalabrutinib treatment on CD49d expression
Next, we evaluated the changes in expression of CD49d on CLL cells. After 6 months of treatment, the median frequency of CD49d-positive cells across all patients was 33% compared with 42% at baseline (Fig. 2A, P = 0.1). The frequency of CD49d-positive CLL cells remained unchanged in samples with very high or low CD49d expression. In contrast, some samples with intermediate CD49d expression at baseline seemed to have changed over time (Fig. 2A). These samples exhibit a bimodal expression pattern, with the coexistence of CD49d-positive and -negative subpopulations within the same sample, as previously described (refs. 6, 7; Supplementary Fig. S2A). There was no significant difference in the treatment-induced lymphocytosis between samples with bimodal versus homogeneous CD49d expression (Supplementary Fig. S2B). Thus, we focused on CD49d expression changes in samples with bimodal CD49d expression (bimCD49d). At the onset of redistribution lymphocytosis, the median frequency of CD49d-positive cells increased from 42% at baseline to 53% after 3 days (Fig. 2B, P = 0.002). The percentage of CD49d-positive cells decreased from 53% at day 3 to 33% after 6 months of treatment (Fig. 2B, P = 0.004).
VLA-4 activation is retained upon BCR and chemokine signaling in acalabrutinib-treated cells
We next assessed the effect of acalabrutinib on VLA-4 activation in response to factors that regulate tumor–microenvironment interactions. As a readout of VLA-4 activation, we measured VLA-4 RO as previously described (Supplementary Fig. S3A–S3C; refs. 6, 22). On circulating CD49d+ CLL cells from patients on acalabrutinib, VLA-4 RO was consistently, albeit only moderately reduced on treatment (median VLA-4 RO, 37% at baseline vs. 32% at 6 months; Fig. 3A, P = 0.02). Upon BCR stimulation in vitro, we observed a significant increase in VLA-4 RO on CLL cells collected from patients at baseline (48% median increase, P = 0.002), and after 6 months on acalabrutinib (38% median increase, P = 0.0005; Fig. 3B). In confirmatory experiments (Supplementary Fig. S3D), we found that BCR stimulation enhanced VLA-4–binding affinity to VCAM-1 at baseline and after 6 months on acalabrutinib (Fig. 3C, P = 0.008, P = 0.04, respectively). Similarly, VLA-4 RO in response to CXCL12 stimulation in vitro increased significantly in samples collected at baseline and during acalabrutinib therapy (Fig. 3D, median increase 56% at baseline, P = 0.0002, and 45% at 6 months, P = 0.0005). Addition of acalabrutinib to in vitro cultures was also unable to block signaling-dependent VLA-4 activation (Supplementary Fig. S3E). Taken together, BCR and CXCR4 activation induced conformational changes in VLA-4 that promote cell adhesion. On circulating CLL cells from patients being treated with acalabrutinib, we observed a modest decrease of VLA-4 activation compared with baseline. Acalabrutinib was not sufficient to block BCR- and CXCR4-induced VLA-4 activation.
BCR-induced but not CXCL12-mediated VLA-4 activation can be overcome with combined BTK and PI3K inhibition
VLA-4 activation in response to BCR and CXCR4 engagement, despite acalabrutinib treatment, indicates a role for non-BTK-dependent signals. In particular, PI3K signaling is known to be involved in inside–out activation of integrins (6, 7). To determine the effects of combined BTK and PI3K inhibition, we analyzed VLA-4 RO on CD49d+ CLL cells collected after 6 months of acalabrutinib therapy, with or without in vitro addition of the PI3K-inhibitor duvelisib. Duvelisib blocked the BCR-induced increase of VLA-4 RO (Fig. 4A). However, the increase in VLA-4 RO in response to CXCL12 stimulation was not affected by the addition of duvelisib (Fig. 4B). These data indicate that inhibition of PI3K and BTK can overcome BCR-induced but not CXCR4-induced VLA-4 activation.
CD49d expression is associated with a distinct transcriptional profile in CLL
To dissect the role of CD49d expression in CLL pathobiology, and response to treatment, we analyzed the transcriptome of purified PB CLL cells from 20 patients contributing matched samples at baseline, after 1 and 6 months of acalabrutinib therapy (Supplementary Table S5). Consistent with BTK inhibition, BCR and NF-κB target genes were downregulated as previously reported (18). Cases were divided by CD49d expression, with 10 cases each representing CD49d+ and CD49d− groups. ITGA4, the gene for CD49d was consistently more highly expressed in the CD49d+ group (Fig. 5A). Differentially expressed genes (DEG) between the CD49d+ and CD49d− CLL groups were 76 at baseline, 291 at 1 month, and 120 at 6 months. Notably, the proportion of genes more highly expressed in the CD49d+ group increased with treatment from 46% at baseline to 56% at 1 month to 90% at 6 months (Fig. 5A; Supplementary Tables S6–S8).
To investigate the biologic basis for the observed transcriptional differences, we used GSEA using well-categorized lymphocytic gene signatures indicative of specific cellular functions and signaling pathways (23). Ten gene signatures were significantly enriched in the CD49d+ group compared with the CD49d− group across all time points (FDR ≤ 0.1 for all signatures and all time points; Fig. 5B; Supplementary Table S9). The transcriptional differences indicate increased immune receptor, NF-κB, and cytokine signaling in CD49d+ cases. In addition genes repressed by PAX5, genes induced by KLF2, and genes downregulated by calcium flux in activated T cells were more highly expressed in the CD49d+ cases (Supplementary Table S9). Common to these transcriptional programs is the regulation of cell activation, adhesion, migration, and cellular metabolism (28–31), all of which are more prominent in the CD49d+ cases. Interestingly, after 6 months on acalabrutinib, CD49d+ samples not only maintained a higher level of activation compared with CD49d− CLL, but differences in the degree of cytokine (IL6 and IL10) and STAT3 signaling became more prominent (Fig. 5B). In summary, CD49d+ CLL shows a distinct transcriptional profile, with higher cellular activation, and capacity for survival, adhesion, and migration.
CD49d− CLL has a low risk of disease progression on BTKi and is associated with favorable prognostic markers
Patients with CD49d+ CLL have been found to have inferior PFS on ibrutinib (6, 7). The study by Tissino and colleagues (6) included ibrutinib-treated patients from several centers, including 34 patients from our clinical trial (21). Here, we extend this analysis to 73 patients treated with ibrutinib and a median follow-up of 6.1 years; 32 (44%) were CD49d− and 41 (56%) CD49d+. Consistent with the earlier report, TTP on ibrutinib in the CD49d+ group was significantly shorter than in the CD49d− group (Supplementary Fig. S4A, P = 0.002). In contrast, in our acalabrutinib-treated patients, there was no difference in TTP based on CD49d expression using 30% as the cutoff value (Supplementary Fig. S4B, P = 0.3). We considered that variable numbers of patients with bimodal CD49d expression could explain the discrepancies between the two BTKi-treated cohorts. Notably, Tissino and colleagues (7) recently reported that bimodal CLL cases follow a clinical course similar to CD49d+ CLL irrespective of the traditional 30% cutoff value. Considering bimodal CD49d expression as a distinct category, we identified three groups: homCD49d+, homCD49d−, and bimCD49d. Patients with bimCD49d CLL had a higher risk of progression than those with homCD49d− CLL in both ibrutinib and acalabrutinib cohorts (Supplementary Fig. S4C and S4D, P ≤ 0.02), whereas the outcome for bimCD49d and homCD49d+ cases was comparable (P ≥ 0.1). To assess the effect of CD49d expression on durability of BTKi therapy in general, we merged the two treatment cohorts. Of the combined 121 patients, 48 (39.7%) progressed on treatment, including 9 with transformed disease (Supplementary Table S4). The CLL was homCD49d− in 31 (25.6%) patients, homCD49d+ in 43 (35.5%), and bimCD49d in 47 (38.8%). Median TTP for homCD49d+/bimCD49d CLL was 6.6 years and not reached for homCD49d− CLL (Fig. 6A, P = 0.0004). Interestingly, TTP for bimCD49d cases with a proportion of CD49d+ cells below the traditional 30% was no different than for those cases with ≥30% CD49d+ cells (Supplementary Fig. S4E). For homCD49d− CLL, 90% of patients were estimated to be progression-free at 8 years. In 39 patients progressing with CLL, mutations in BTK and/or PLCG2 were detected in 32 (82.1%; Supplementary Fig. S5A; Supplementary Table S4).
Next, we tested whether adding CD49d expression to our previously validated four-factor model (CLL-4) can improve on the prognostic information (5). On the basis of the 4 factors, TP53 aberration, prior treatment, β2M ≥5 mg/L, and LDH >250 U/L patients were separated into 3 groups (Supplementary Tables S1 and S2). Patients with 3 or 4 factors present were classified as having high-risk disease and had significantly inferior TTP compared with patients classified as low or intermediate risk by the CLL-4 model (Fig. 6B, P ≤ 0.008). Patients in the high-risk group were more likely to be homCD49d+ and bimCD49d (P = 0.003), whereas only two high-risk patients were in the homCD49d− group (Supplementary Fig. S5B). Among 90 patients classified as low and intermediate risk by the CLL-4 model, TTP for patients with homCD49d+/bimCD49d CLL was 7.2 years and not reached for homCD49d− patients (Fig. 6C, P = 0.003). In multivariate analysis, the CD49d expression group (homCD49d+/bimCD49d vs. homCD49d−) remained significantly associated with progression (adjusted HR, 4.99; 95% confidence interval, 1.54–16.21; P = 0.007), controlling for the 4-factor risk model and other baseline patient characteristics (Supplementary Fig. S5C; Supplementary Table S10).
Covalent BTKis are highly effective in CLL and have become a mainstay of treatment. In addition to inhibition of BCR signaling, BTKis also disrupt tumor–microenvironment interactions. CD49d integrin is a master regulator of microenvironmental interactions in CLL (14), and high CD49d expression was linked to an inferior response to ibrutinib (6, 7). Here, we investigated the role of CD49d in response to BTKis, extending prior findings using samples from patients treated with acalabrutinib and outcome data from patients treated with acalabrutinib or ibrutinib. We characterized differences in transcriptomes between CD49d+ and CD49d− CLL and show that CD49d expression provides additional and independent prognostic information to the 4-factor outcome model we previously reported (5). Taken together, our results support the prognostic value of CD49d in the setting of BTKi therapy and its contribution to the emergence of BTKi resistance by sustaining tumor–microenvironment interactions that enhance cell survival.
In ibrutinib-treated patients, CD49d+ cases showed diminished treatment-induced lymphocytosis compared with CD49d− cases and less reduction in lymphadenopathy (6). More rapid release of CD49d− CLL cells into the circulation was considered a possible explanation for these differences (6). However, here we observed equally prominent redistribution lymphocytosis in both CLL subtypes within the first two weeks of acalabrutinib treatment. These early timepoints were not analyzed in the ibrutinib cohort. In cases with bimodal expression of CD49d, the proportion of CD49d+ CLL cells increased on day 3. As we have previously shown that the early rise in lymphocytosis is due to the release of CLL cells from lymph nodes (10), the relative increase of CD49d+ cells is consistent with the redistribution of activated cells from lymphoid tissues (7, 10). Thus BTKis, at least in the first weeks of treatment, are able to disrupt adhesion of CD49d+ cells to the microenvironment. Over time, the proportion of CD49d+ CLL cells in circulation decreased, consistent with reports of decreasing VLA-4 expression during ibrutinib therapy (6, 32). Possible explanations for the more rapid disappearance of CD49d+ cells include increased death of cells that are more dependent on microenvironmental survival factors (6, 7, 10, 33, 34), downregulation of CD49d concomitant with decreased cell activation (10, 35), or a greater ability to migrate and re-enter lymphoid tissues (33). In contrast with the earlier study (6), we did not observe any differences in nodal or spleen response between CD49d+ and CD49d− cases treated with acalabrutinib. All considered, increased clearance of CD49d+ cells from circulation appears to be the most likely explanation.
In response to external stimulation, inside–out signaling induces a conformational change in VLA-4, increasing the affinity for ligands such as VCAM-1 and fibronectin. As previously reported for ibrutinib-treated patients, we found that VLA-4 affinity was downregulated in circulating CLL cells from patients on acalabrutinib compared with baseline but could be restored by in vitro BCR or CXCL12 stimulation. Insufficient BTK inhibition can be excluded given that we measured median 95% BTK occupancy by acalabrutinib at drug trough levels in patients on study (19), and the addition of acalabrutinib to in vitro cultures had no effect. PI3K signaling, bypassing BTK, has been implicated in inside–out signaling following BCR activation (6). The addition of PI3K inhibitor duvelisib to the acalabrutinib-treated CLL cells completely blocked BCR-induced VLA-4 activation. In contrast, CXCR4-controlled VLA-4 activation was unaffected. Thus the combination of BTK and PI3K inhibitors is not expected to effectively block VLA-4 activation in the microenvironment. These data are in line with a previous report, showing that dual inhibition of BTK and PI3K failed to inhibit CXCR4-controlled integrin mediated adhesion in Waldenstrom Macroglobulinemia (36).
CD49d can be an activation marker, but more generally, appears to define a distinct subset of CLL. Observations linking CD49d expression to cell activation include increased frequency of CD49d+ cells in bone marrow, compared with PB, and in the CXCR4dim/CD5 bright fraction of recent emigrants from lymphoid tissues (6, 33). Furthermore, CD49d expression correlates with increased CD38 and Ki67 expressions (6, 33). On the other hand, CD49d associates with genetic alterations, including trisomy 12, del17p, and presence of NOTCH1 mutations (7, 37, 38). To dissect differences between subsets, we compared the global transcriptomes of CD49d+ and CD49d− subsets by RNA sequencing at baseline and after 1 and 6 months on acalabrutinib. Using GSEA, the CD49d+ cases showed an overall increase in cytokine, immune receptor, and NF-κB signaling, resulting in increased cellular activation, proliferation, adhesion, and survival. Although treatment with acalabrutinib downregulates activation, proliferation, NF-κB, and cytokine signaling (18), CD49d+ cases not only maintained their higher level of activation compared with CD49d− cases but seemed to gain further. In particular, transcriptional programs, indicating JAK/STAT signaling, were further increased in CD49d+ cases over CD49d− at 6 months. Thus, the increased fitness of CD49d+ cells persisted on BTKi therapy. Considering that the overall response rate to acalabrutinib was 95.8% (17), CD49d does not directly confer BTKi resistance. Furthermore, 82.1% of the patients progressing with CLL on BTKi therapy were found to have acquired BTK and/or PLCG2 mutations.
CD49d is well established as a strong prognostic factor in CLL (15, 39). In contrast with IGHV mutational status, CD49d maintains importance in patients treated with ibrutinib (6, 7). In our acalabrutinib cohort, CD49d+ CLL, categorized using the established cutoff value of 30% CD49d-expressing CLL cells, was not associated with a risk of early progression. More recently, Tissino and colleagues (7) made the important observation that cases with bimodal CD49d expression and <30% of CD49d+ CLL cells shared the same prognosis as patients with CD49d+ CLL. Incorporating this distinction, we found that homCD49d− CLL indeed had a significantly lower risk of progression on acalabrutinib compared with CD49d+ and bimCD49d cases. Combining the acalabrutinib and ibrutinib cohorts, we observed 121 patients over a median follow-up of 4.9 years. Notably, 90% of patients with homCD49d− CLL, defined by complete absence of CD49d expression on tumor cells, were progression-free at 8 years. In contrast, patients with bimCD49d CLL had outcomes indistinguishable from CD49d+ cases with a median TTP of 6.9 years. This was true even for cases where the proportion of CD49d+ cells was clearly below 30%. Considering CD49d expression also improved on the CLL-4 prognostic model we developed in patients treated with ibrutinib (5). On the basis of the 4 factors: Treatment history, TP53 aberration, β2M, and LDH patients are divided into three risk groups. High-risk patients by this CLL-4 model were enriched for bimCD49d/CD49d+. Among low- and intermediate-risk patients, CD49d expression separated groups with starkly different risks of progression. In multivariate analysis CD49d was an independent risk factor. Thus, there may be value of adding CD49d to prognostic models, especially for patients treated with BTKi. Testing such models will require larger cohorts of patients that incorporate testing for CD49d expression.
So, how then does CD49d expression increase the risk of disease progression on BTKi therapy? Observations in patients suggest that CD49d+ cells may be more dependent on tumor–microenvironment survival signals, and therefore have a faster rate of cell death once in the periphery. On the other hand, the transcriptional profile of CD49d+ versus CD49d− CLL cells is consistent with an increased ability of CD49d+ cells to adhere, and access microenvironmental activation signals. Notably, even a minor CD49d+ subpopulation conferred a higher risk of disease progression in bimCD49d cases. We, therefore, hypothesize that these CD49d+ cells contribute to a tumor-supporting microenvironment countering some of the effects of BTKi therapy (40). Specifically, an immunosuppressive microenvironment may create a permissive niche for clonal evolution and eventual outgrowth of subclones carrying specific mutations (41). However, the underlying molecular mechanisms how CD49d expression contributes to the development of BTKi resistance remain still ill-defined.
Taken together, CD49d expression emerges as a microenvironmental factor predisposing to the emergence of clones ultimately driven by genetic resistance mechanisms. Further investigation into how CD49d could guide treatment strategies to improve durability of BTKi therapy is needed.
A. Itsara reports other support from Adaptive Biotechnologies outside the submitted work. I.E. Ahn reports personal fees from BeiGene outside the submitted work. C. Sun reports grants from Genmab outside the submitted work. E. Bibikova reports employment by AstraZeneca and equity ownership in AstraZeneca and Acerta Pharma. T.N. Hartmann reports grants from AstraZeneca and German Cancer Aid during the conduct of the study. A. Wiestner reports grants from Pharmacyclics and Acerta during the conduct of the study as well as grants from Merck, Nurix, Genmab, and Verastem outside the submitted work. No disclosures were reported by the other authors.
A. Alsadhan: Conceptualization, data curation, formal analysis, investigation, visualization, methodology, writing–original draft, writing–review and editing. J. Chen: Data curation, formal analysis, visualization, writing–original draft, writing–review and editing. E.M. Gaglione: Writing–review and editing. C. Underbayev: Formal analysis, writing–review and editing. P.L. Tuma: Conceptualization, writing–review and editing. X. Tian: Data curation, writing–review and editing. L.A. Freeman: Writing–original draft. S. Baskar: Methodology, writing–review and editing. P. Nierman: Writing–review and editing. S. Soto: Writing–review and editing. A. Itsara: Writing–review and editing. I.E. Ahn: Resources, writing–review and editing. C. Sun: Resources, writing–review and editing. E. Bibikova: Conceptualization, visualization, writing–review and editing. T.N. Hartmann: Conceptualization, resources, visualization, methodology, writing–review and editing. M. Mhibik: Conceptualization, data curation, formal analysis, supervision, investigation, visualization, methodology, writing–original draft, writing–review and editing. A. Wiestner: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, visualization, methodology, writing–original draft, project administration, writing–review and editing.
We thank our patients for participating and donating samples to make this research possible. This work was funded by King Saud bin Abdulaziz University for Health Sciences Cultural Mission of the Royal Embassy of Saudi Arabia (2085374; to A. Alsadhan), the Deutsche Krebshilfe (DKH; 70113993; to T.N. Hartmann), the Intramural Research Program of the NHLBI (HL002346–15; to A. Wiestner), and Acerta Pharma, a member of the AstraZeneca group. We thank Daniel Auclair, AstraZeneca, for comments on the article draft. The authors also thank the NHLBI Flow Cytometry Core, the NHLBI Sequencing and Genomics Core, and the NHLBI Bioinformatics and Computational Biology Core.
The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.
Note: Supplementary data for this article are available at Clinical Cancer Research Online (http://clincancerres.aacrjournals.org/).