FoxO1-GAB1 axis regulates homing capacity and tonic AKT activity in chronic lymphocytic leukemia

Chronic lymphocytic leukemia (CLL) cell recirculation to and from immune niches has been proposed as a feature of CLL biology,^1-7  although it has not been entirely experimentally proven. Inside the immune niches, CLL cells obtain various prosurvival and proproliferative signals leading to activation of the B-cell receptor (BCR), phosphatidylinositol 3-kinase (PI3K)/AKT, and NFκB pathways.^2,8-11  Inhibiting CLL cell homing to immune niches represents an important mechanism of action for BCR inhibitors and a possible therapeutic strategy.^12-14  One proposed model for the study of CLL cell recirculation is to use CXCR4/CD5 intraclonal subpopulations. It has been postulated that CXCR4^dimCD5^bright cells represent a proliferative fraction of cells that have recently exited the microenvironment, whereas CXCR4^brightCD5^dim cells correspond to resting cells circulating in peripheral blood for an extended time before they return back to the immune niches.^1-7  CLL cell migration to immune niches is known to be regulated by selectin CD62L^15,16 ; integrin VLA4^17,18 ; and chemokine receptors, such as CXCR4, CXCR5, and CCR7 that guide cells toward gradients of ligands.^10,17,19-23  Migration is a highly complex process, and CXCR4 levels in CLL subpopulations were shown to reflect the history of microenvironmental interactions, since binding of SDF1 chemokine to CXCR4 leads to its lower levels on the CLL cell surface, whereas CXCR4 upregulation sensitizes cells to SDF1-guided migration.^1-7,20  However, it remains largely unknown how precisely CLL cells integrate multiple migratory signals while balancing survival in peripheral blood and the decision to return to the immune niches. It has been described that CLL cells or normal B cells outside the immune niches can use “tonic” PI3K activation^24-28  or autonomous BCR signaling to sustain their survival²⁹ ; however, factors regulating this or their interplay with homing capacity are largely unclear. It is also unclear whether PI3K activation plays a role in adaptation/resistance to BCR signaling inhibition and impaired homing during ibrutinib therapy.³⁰  We hypothesized that analyzing the “resting” nonproliferative CXCR4^brightCD5^dim cells could help to understand the factors determining the return of these cells to immune niches and/or factors that balance the survival of peripheral blood CLL cells. This may have implications for CLL cell adaption to therapeutic inhibition of homing by BCR inhibitors.

In this study, we have shown that the CXCR4^brightCD5^dim cell subpopulation has a higher capacity to home into immune niches. Analysis of this subpopulation revealed differential expression of dozens of migration-related genes, including upregulation of GAB1, which is known to act as a migration regulator in nonimmune cells.^31,32  However, its role in lymphocyte migration is unknown. It has been demonstrated that GAB1 serves as an adaptor protein that integrates signals from different receptors and affects signaling cascades, such as PI3K/AKT,^33-35  MAPK/extracellular signal-regulated kinase (ERK),^36-38  PLCγ,³⁹  and Crk/CrkL.⁴⁰  The results of our study show that GAB1 positively affects CLL cell migration in response to various chemoattractants. Gradual GAB1 accumulation in CLL cells outside immune niches is mediated by FoxO1-induced transcriptional activation of the GAB1 gene. We also found that GAB1 upregulation has a role in maintaining the basal tonic AKT phosphorylation required to sustain CLL cell survival. This is important during ibrutinib therapy since CLL cells induce the FoxO1-GAB1-pAKT axis in this context, which can be targeted therapeutically by the GAB1 inhibitors tested in this study.

The study was approved by the institutional review board, and samples were obtained with written informed consent. CLL samples were purified to contain ≥95% of CD5⁺19⁺ cells (see supplemental Methods, available on the Blood Web site). MEC1 and HS5 cell lines were obtained from the German Collection of Microorganisms and Cell Cultures GmbH, and American Type Culture Collection, respectively, and the OSU-CLL was a gift from John Byrd (The Ohio State University, Columbus, OH).⁴¹ 

For RNAseq analysis, freshly isolated CLL cells were sorted according to CXCR4/CD5 expression (purity ≥99.9%), as described previously.⁷  Libraries were prepared using the TruSeq Stranded messenger mRNA (mRNA) LT Sample Prep Kit (Illumina) and sequenced with the NextSeq 500/550 High Output v2.5 Kit (supplemental Methods).

Cells transfected with siRNA against GAB1, or GAB1/FoxO1 knockout (KO) cells, or control cells were labeled with carboxyfluorescein succinimidyl ester (CFSE) or FarRed CellTrace dye (Thermo Fisher Scientific) or were traced with a plasmid encoding GFP or Azurite, and an equal number of the cells was mixed in a 1:1 ratio (validated by flow cytometry) and injected into the tail vein of NSG mice.⁴²  For a description of the in vitro competitive migration assay, see the supplemental Methods. Cells were treated with ibrutinib (1 µM), MK-2206 2HCl (1 µM), AS1842856 (0.5 µM), plerixafor (5 µg/mL; all from Selleckchem), or GAB1 pleckstrin homology (PH) domain–targeting inhibitors (GAB-001, GAB-004).⁴³  The cells were treated by goat F(ab′)₂ anti-human IgM (10 μg/mL; Southern Biotech), SDF1, or CXCL13 (Peprotech), as described previously.^2,44,45 

Apart from NGS analysis (supplemental Methods), all statistical analyses were performed with Prism v8.0.1 (GraphPad). Values of P < .05 were considered significant.

CXCR4/CD5 cell subpopulations have been suggested as a model distinguishing cells ready to reenter immune niches (CXCR4^brightCD5^dim) vs recent emigrants from lymph nodes (CXCR4^dimCD5^bright).^1-7  We aimed to first test each subpopulation’s migration potential in NSG mice; however, labeling cells with an anti-CXCR4 antibody could affect their responsiveness to the CXCR4 ligand SDF1. To overcome this problem, we used CD5 and forward scatter area (FSC-A reflecting cell size) for cell sorting, because lower FSC-A is typical for CXCR4^brightCD5^dim and higher FSC-A for the CXCR4^dimCD5^bright subpopulation (supplemental Figure 1). Subsequently, the sorted subpopulations were stained with 2 cell-tracing dyes, and their 1:1 mixture was injected into the tail vein of NSG mice for a 4-hour competitive migration analysis. In line with the model, the CXCR4^brightCD5^dim CLL cell subpopulation migrated significantly more into the spleen from the peripheral blood than the CXCR4^dimCD5^bright cells (Figure 1A). CLL cell migration into bone marrow or the liver was the same, and the number of CLL cells infiltrating these organs was low (supplemental Figure 2), as reported previously.⁴⁶  The higher CXCR4^brightCD5^dim cell migration was also confirmed in an in vitro Transwell cell migration assay toward SDF1 (CXCL12), CXCL13, or conditioned medium (CM) derived from primary human mesenchymal stromal cells (CM^hMSC; supplemental Figure 2). This finding suggested that it is possible to identify homing-related genes by analyzing CXCR4/CD5 subpopulations. We further performed transcription analysis (RNA sequencing [RNAseq]) of the CXCR4^dimCD5^bright vs the CXCR4^brightCD5^dim cell subpopulations (a sorting strategy for CXCR4/CD5 subpopulation in Figure 1B, as detailed previously^2,7 ). This process identified 1372 differentially expressed genes in CXCR4/CD5 subpopulations (fold change, >1.5; P < .001). The CXCR4^brightCD5^dim cells were characterized by BCR/NFκB signature gene downregulation (eg, MYC, CD83, FOS, CCND2, BCL2A1, DUSP2, CCL3, and CLL4), corresponding to a resting nonactivated B-cell phenotype (for the list of genes and principal component analysis, see supplemental Figure 3 and 4; supplemental Table 2). The differentially expressed mRNAs included 147 genes annotated in databases as migration-related (Figure 1C; fold change, >1.5; P < .001; supplemental Table 3), and Gene Ontology analysis confirmed a significant enrichment of migration/adhesion–related genes (supplemental Table 4). Several such randomly selected genes were validated by real-time quantitive polymerase chain reaction (qPCR; supplemental Figure 5). Among these putative migration-related genes, we noticed a GAB1 upregulation (∼1.8-fold) in CXCR4^brightCD5^dim cells compared with CXCR4^dimCD5^bright cells. In a previous study, we showed that higher GAB1 levels are associated with shorter overall survival in CLL,³⁵  and GAB1 was identified as a migration regulator in epithelial and muscle cells^31,32  and regulator of BCR responsiveness.^33-35  The higher GAB1 levels in CXCR4^brightCD5^dim were confirmed by real-time qPCR at the mRNA level (Figure 1D) and by flow cytometry and immunoblot analysis at the protein level (Figure 1E-G; supplemental Figure 6). On the contrary, GAB1 was downmodulated in CLL cells exposed to microenvironmental signals, such as stromal cell coculture or BCR activation (see below). Altogether, this shows that GAB1 expression is induced in CXCR4^brightCD5^dim CLL cells that have a relatively increased homing capacity.

We hypothesized that GAB1 upregulation in a CXCR4^brightCD5^dim subpopulation may contribute to the return of the CLL cells to the immune niches. To examine cell homing, we created a GAB1 KO MEC1 cell line, using CRISPR-Cas9 technology (Figure 2A). The wild-type MEC1 (MEC1^WT) and MEC1^GAB1-KO cell lines were labeled with 2 cell-tracing dyes mixed in a 1:1 ratio, and injected into the tail vein of NSG mice (Figure 2B). The result revealed that MEC1^GAB1-KO cells were significantly retained in peripheral blood and were relatively less able (∼30% reduction) to migrate to the spleen than were the control cells (Figure 2C). Bone marrow or liver infiltration by MEC1^WT vs MEC1^GAB1-KO did not differ, and the number of MEC1 cells infiltrating these organs was low (Figure 2C; immunohistology in supplemental Figure 7). Similarly, silencing GAB1 by siRNA in CLL cells impaired their migration to the spleens of NSG mice (Figure 2D; immunohistology in supplemental Figure 8; silencing efficiency in supplemental Figure 9). B-cell migration is known to be regulated by several chemokines, such as SDF1 and CXCL13, produced by cells in immune niches.^10,21-23  Indeed, we confirmed their presence in the bone marrow of patients with CLL, production by primary human bone marrow mesenchymal stromal cells, or the human stromal cell line HS5, and in the tissues of NSG mice (supplemental Figure 10). Treating MEC1^WT cells with SDF1, CXCL13, or CM^hMSC led to rapid (within minutes) GAB1 phosphorylation at activating tyrosines Y307 and Y627 (Figure 2E), suggesting that GAB1 is involved in response to multiple soluble factors. Indeed, in vitro MEC1^GAB1-KO cell migration toward SDF1, CXCL13, or CM^hMSC was impaired compared with MEC1^WT cells (Figure 2F). Similar data were obtained when GAB1 was silenced by siRNA in MEC1 cells, and CM from HS5 cells was used as an attractant (supplemental Figure 11). The specific CXCR4 inhibitor plerixafor largely blocked the GAB1 phosphorylation induced by recombinant SDF1 (CXCR4 ligand) but did not block the phosphorylation induced by the CM, also confirming that signaling by multiple soluble factors involves GAB1 activation (supplemental Figure 11). In primary CLL cells, silencing GAB1 by small interfering RNA (siRNA) also reduced phosphorylation of kinases activated after cell stimulation with SDF1, CXCL13, or CM^hMSC (Figure 2G; supplemental Figure 12A) and impaired migration toward SDF1, CXCL13, or CM^hMSC in vitro, compared with cells transfected with the negative control (Figure 2H). Cell viability and CXCR4 levels were not affected by siRNA against GAB1 (supplemental Figure 12B-C). Moreover, MEC1^GAB1-KO cells failed to properly polymerize F-actin (a process required for effective migration) after stimulation with CM from HS5 cells (supplemental Figure 13A). MEC1^GAB1-KO cell motility speed was significantly (P < .001) lower than MEC1^WT (supplemental Figure 13B). GAB1 KO also impaired other migration-related characteristics of the cells, as assessed by holographic microscopy (supplemental Table 5). Altogether, findings suggest that GAB1 sensitizes CLL cells to homing signals and influences the related cytoskeletal organization.

To reveal potential GAB1 expression regulators, we identified 20 annotated⁴⁸  transcription factors significantly differentially expressed in CXCR4/CD5 subpopulations (Figure 3A; supplemental Table 6). We further studied FoxO1 as a potential GAB1 regulator because of the presence of multiple putative evolutionary conserved binding sites upstream from the GAB1 promotor (supplemental Figures 14 and 15; supplemental Table 7). We also focused on FoxO1 because it has been identified as a regulator of T-cell homing⁴⁹  and normal B-cell migration within the germinal centers^50,51  and is involved in malignant B-cell biology.^52,53  Chromatin immunoprecipitation identified FoxO1 binding at 2 sites upstream from the GAB1 transcription start sites (Figure 3B; supplemental Table 8). In addition, FoxO1 downmodulation by siRNA or treatment with FoxO1 inhibitor significantly repressed GAB1 at the mRNA and protein levels (Figure 3C-D; supplemental Figure 16A-B). The existence of a FoxO1-GAB1 axis is supported by a general positive correlation of GAB1 and FoxO1 mRNA levels (n = 42) and protein levels (n = 22) in CLL cells (Figure 3E-F), and the higher levels of both molecules in CXCR4^brightCD5^dim cells (Figure 3G-I; supplemental Figure 6). Moreover, MEC1 cells without FoxO1 (MEC1^FoxO1-KO) had lower GAB1 levels (supplemental Figure 16C) and had impaired migration to immune niches in NSG mice in comparison with MEC1^WT cells (supplemental Figure 17A). The migration of MEC1^FoxO1-KO cells or CLL cells with silenced FoxO1 toward SDF1, CXCL13, or CM^hMSC was also impaired in vitro (supplemental Figure 17B-C), which mimics the MEC1^GAB1-KO cell phenotype.

To understand the GAB1 regulation via FoxO1, we further analyzed the mechanisms resulting in their relatively higher levels in CXCR4^brightCD5^dim CLL cells and lower levels in CXCR4^dimCD5^bright cells. In normal B cells, FoxO1 is known to be negatively regulated by its phosphorylation at position S256 in the context of BCR activation. This regulatory process is mediated by active AKT, leading to loss of FoxO1 interaction with DNA, translocation to cytoplasm, and subsequent degradation.^55,56  Indeed, in CLL cells, strong AKT activation by BCR cross-linking or stromal cell coculture induced inhibitory FoxO1 phosphorylation with its subsequent degradation, and concurrently, GAB1 mRNA and protein levels were decreased (supplemental Figure 18). On the contrary, selective AKT inhibitor induced dephosphorylation of FoxO1, and upregulation of GAB1 mRNA and protein levels (∼twofold increase; Figure 3J). The expression of other selected BCR/microenvironment-related proteins such as SYK, PLCγ, or CD79α, was not affected by AKT inhibition (Figure 3J; supplemental Figure 19). These data suggest that AKT is a crucial, negative GAB1 regulator acting via reducing its transcriptional activator levels, FoxO1. On the other hand, relatively lower AKT activity in CXCR4^brightCD5^dim cells (supplemental Figure 21) allows for an increase in FoxO1 and induction of GAB1. Altogether, the data show that FoxO1 transcriptionally regulates GAB1, and both proteins are upregulated in the CXCR4^brightCD5^dim intraclonal cell subpopulation contributing to increased cell migration capacity.

We and others have shown that GAB1 is phosphorylated after BCR activation and serves as an important docking protein, allowing for propagation of the signal to PI3K and AKT.^33,35  In line with this, we observed that silencing of GAB1 impaired AKT^S473 phosphorylation after BCR cross-linking (Figure 4A). In this respect, we noted that modulation of the GAB1 level also affected basal AKT phosphorylation levels, irrespective of BCR cross-linking (Figure 4A), hinting at a possible role in tonic AKT activity. Indeed, GAB1 overexpression by a plasmid construct induced AKT phosphorylation (Figure 4B). Analogically, silencing endogenous GAB1 by siRNA reduced AKT phosphorylation (Figure 4C). The effect of GAB1 on AKT activation was validated on MEC1 and OSU-CLL cells with tetracycline-inducible shRNA against GAB1 (Figure 4D). GAB1 overexpression led to the induction of phosphorylation of the PI3K subunit p85, whereas GAB1 silencing by siRNA had the opposite effect (Figure 4B-C), which indicates that a GAB1–induced increase in p85 phosphorylation is a possible mechanism behind the observed increase in AKT phosphorylation. We have also tested MEC1^GAB1-KO vs MEC1^WT cell growth capacity after engraftment in mice for 3 weeks. Similar to the short-term migration assays, the MEC1^WT and MEC1^GAB1-KO cell lines (each traced with GFP/Azurite) were mixed in a 1:1 ratio and injected into the tail vein of NSG mice. This revealed approximately twice as many MEC1^WT cells as MEC1^GAB1-KO cells in the spleen, bone marrow, and liver, a difference slightly larger than in the migration experiments (Figure 4E; immunohistology in supplemental Figure 22), suggesting that GAB1 KO also negatively affects general cell fitness. To further investigate the FoxO1-GAB1-pAKT axis, we analyzed MEC1^FoxO1-KO cells and observed diminished GAB1 levels and reduced AKT phosphorylation compared with MEC1^WT cells (Figure 4F). A similar phenotype was obtained after FoxO1 silencing by siRNA in primary CLL cells, after MEC1 cells, or CLL-cell treatment with FoxO1 inhibitor (Figure 4F; supplemental Figure 16). Altogether, this suggests that increased FoxO1/GAB1 levels in CXCR4^brightCD5^dim CLL cells contribute to tonic pAKT activity.

Ibrutinib therapy inhibits CLL cell homing,^6,57  and we hypothesized that this could affect the FoxO1-GAB1-pAKT axis. Indeed, treating MEC1 cells with ibrutinib in vitro led to FoxO1 and GAB1 protein and mRNA accumulation (Figure 5A-C), albeit the kinetics of mRNA and protein induction were not identical, and the GAB1 protein levels were sustained for longer. Similarly, treating CLL patients with ibrutinib in vivo leads to increased FoxO1 and GAB1 protein and mRNA levels (Figure 5D-E). Notably, after an initial drop in AKT phosphorylation, we observed restoration of AKT^S473 phosphorylation after several days of continuous MEC1 cell exposure to ibrutinib in vitro (Figure 5A) or after several days or weeks of ibrutinib therapy in CLL patients in vivo (Figure 5D; supplemental Figure 23). Idelalisib, as a single agent, also increased GAB1 and FoxO1 protein levels, but the induction of pAKT was less prominent (supplemental Figure 24). This result corresponds to the PI3K involvement in inducing AKT phosphorylation after GAB1 upregulation (Figures 4B-C and Figure 5D). Importantly, the AKT phosphorylation in response to ibrutinib did not lead to FoxO1 localization into the cytoplasm, and FoxO1 remained in the nucleus, enabling it to function as a transcription factor (supplemental Figure 25). This finding is in line with the data of another study suggesting that transport of FoxO1 from the nucleus upon AKT activation is impaired in B-cell lymphomas,⁵²  and the FoxO1 activity is not mutually exclusive with low-level pAKT. Importantly, the GAB1 knockdown prevented the restoration of AKT phosphorylation in ibrutinib-treated cells (Figure 5F). Altogether, this result suggests that increasing GAB1 levels in CLL cells during ibrutinib therapy provides tonic pAKT activity and may compensate for the lack of AKT activation via BCR/microenvironmental interactions during ibrutinib therapy.

We hypothesize that GAB1 upregulation during ibrutinib therapy acts as a mechanism to compensate for the lack of pAKT levels and that this effect can be targeted therapeutically. Recently, the first GAB1 inhibitors targeting the PH-domain necessary for GAB1’s docking function have been developed.⁴³  We observed that these compounds led to a dose-dependent inhibition of GAB1 phosphorylation, AKT inhibition (Figure 6A; supplemental Figure 26) and decreased CLL cell viability at higher micromolar concentrations after 48 hours (Figure 6B). To test GAB1 inhibitor specificity, MEC1^WT and MEC1^GAB1-KO cells were treated with various doses of the inhibitors. This demonstrated a greater induction of apoptosis and pERK/pAKT repression in MEC1^WT cells by GAB1 inhibitor than in MEC1^GAB1-KO cells (supplemental Figure 27). GAB1 inhibitors also inhibited CLL cell migration toward HS5 CM, SDF1, CXCL13, or CM^hMSC (Figure 6C-E). GAB1 inhibition impaired Ca²⁺ influx induced by SDF1 chemokine (Figure 6F) and BCR-induced (anti-IgM) AKT phosphorylation (Figure 6G). Importantly, the GAB1 inhibitor blocked the “compensatory” AKT phosphorylation induced by ibrutinib (Figure 6Hi) and increased the apoptosis of MEC1 cells exposed to ibrutinib for an extended time (Figure 6Hii). To test this in primary CLL cells undergoing spontaneous apoptosis within days, CLL cells were cultured in medium with interleukin 4 (IL4), which partially prevented their spontaneous apoptosis, and were simultaneously treated with ibrutinib or vehicle for 6 days, to allow time for FoxO1-GAB1 induction (supplemental Figure 28A). Subsequently, these cells were treated with GAB1 inhibitors (or vehicle) for an additional 48 hours. CLL cell killing was increased by GAB1 inhibitors combined with ibrutinib, compared with ibrutinib or GAB1 inhibitor alone (Figure 6I-J; P < .001). In culture conditions without IL4, allowing for only a short-term experiment, GAB1 inhibitors showed an ability to kill primary CLL cells within 48 hours, but this was not synergistic with ibrutinib (supplemental Figure 28B). This result was expected because in a brief time of ibrutinib exposure, the GAB1 inhibitor acts similarly to a BCR inhibitor in inhibiting tonic pAKT levels, and ibrutinib does not yet induce compensatory AKT phosphorylation via GAB1. Altogether, GAB1 inhibition can be used to inhibit CLL migration, basal or BCR-induced AKT activity, and adaptive AKT activation during ibrutinib therapy.

In this study, FoxO1 transcriptionally induced GAB1 in CLL cells, which increased their homing capacity by influencing chemokine responsiveness and F-actin polymerization. FoxO1-induced GAB1 also increased the basal/tonic AKT activity, which is underscored in the context of ibrutinib therapy. Ibrutinib leads to FoxO1-GAB1 axis induction and adaptive restoration of AKT phosphorylation. We have demonstrated that GAB1 can be targeted therapeutically by novel GAB1 inhibitors, alone or in combination with BTK inhibition.

It has been hypothesized that CLL cell recirculation can be studied by analyzing CXCR4/CD5 intraclonal CLL subpopulations, given that CXCR4^brightCD5^dim cells should be more responsive to SDF1 because of higher CXCR4 levels.^1-7  Indeed, the CXCR4^brightCD5^dim subpopulation showed a significantly higher tendency to migrate into the spleen from the peripheral blood of NSG mice than did CXCR4^dimCD5^bright cells and, in vitro, toward chemokines. We have previously analyzed CXCR4^dimCD5^bright cells to understand the pathways activated in the microenvironment.^2,3,7,59  In the current study, we hypothesized that analyzing the “resting” nonproliferative CXCR4^brightCD5^dim cells could help to understand the factors influencing their return to immune niches. RNA sequencing identified 147 genes differentially expressed in CXCR4/CD5 subpopulations and annotated in databases as migration related. We further demonstrated that one of the genes upregulated in CXCR4^brightCD5^dim cells (namely, GAB1) is a regulator of CLL cell migration. The genetic ablation of GAB1 in malignant MEC1 cells largely inhibited their migration capacity toward SDF1, CXCL13, or CM produced by bone marrow stromal cells, and a similar result was obtained with siRNA against GAB1 in primary CLL cells. This inhibitory action was confirmed by an in vivo competitive migration assay in mice, showing that GAB1 KO or GAB1 downmodulation by siRNA leads to impaired migration of malignant B cells to the spleen. This result is in line with data from muscle progenitor cells,³²  where genetic ablation of either GAB1 or CXCR4 leads to highly similar (but not identical) defects in migration to specific tissue locations, suggesting that GAB1 has a nonredundant role in the CXCR4 pathway, but also in other cell-migration mechanisms. In line with this notion, we observed that cytokines produced by cells in immune niches, including CXCR4 ligand SDF1 and CXCR5 ligand CXCL13, induced GAB1 phosphorylation within minutes, and GAB1 positively regulated F-actin polymerization and cell migration in response to chemokines. This effect could be functionally linked to GAB1’s acting as a cell-membrane–docking site for multiple proteins involved in migration such as PI3K,^34,60  ERK,³⁶  PLCγ, Crk, or CrkL proteins.⁶¹  Indeed, we have observed that GAB1 positively affects the activity of PI3K, a known regulator of F-actin polymerization, and CLL/T lymphocyte migration.^60,62,63  Altogether, the data show that GAB1 accumulation in CLL cells increases their homing capacity by influencing responsiveness to chemokines and integrating signals, resulting in actin reorganization.

We further studied the mechanism that leads to GAB1 accumulation in CXCR4^brightCD5^dim cells, showing that the transcription factor FoxO1 accumulates in these cells and transcriptionally activates GAB1. It is known that AKT activity negatively regulates FoxO1 by mediating inhibitory FoxO1^S256 phosphorylation, resulting in its subsequent translocation from the nucleus to the cytoplasm and degradation.^64,65  Indeed, a strong AKT phosphorylation induction in CLL cells after adhesion or BCR activation leads to FoxO1 degradation and concurrent GAB1 downmodulation. We also tested whether miR-150, a known regulator of GAB1, affects GAB1 levels, but we did not find anticorrelation of the miR-150 levels with GAB1 levels in CXCR4/CD5 subpopulations (supplemental Figure 29), suggesting that posttranscriptional regulation by miR-150 is not relevant in this context, but more likely contributes to the interpatient differences, as proposed previously.³⁵  Altogether, the data indicate that FoxO1 upregulation in the CXCR4^brightCD5^dim CLL subpopulation is permitted by relatively low-level AKT activity (related to the absence of BCR/adhesion–mediated signals) and leads to transcriptional GAB1 activation and supports cell homing back to “signal-rich” niches. Similar to GAB1 KO, FoxO1’s genetic ablation or its silencing by siRNA also resulted in decreased B-cell migration capacity. FoxO1’s role in influencing genes involved in cell migration corresponds to observations describing its involvement in dendritic cell homing,⁶⁶  T-cell homing,⁴⁹  and normal B-cell migration within the germinal centers.^50,51  In our data, several migration-related genes (such as ITGB2, PIK3CA, NF1, ETS1, ST14, LRRK2, and APBB2), which are putative FoxO1 targets, were differentially expressed in CXCR4/CD5 subpopulations. This finding suggests that FoxO1 contributes to regulating CLL migration by inducing GAB1 and other genes. Notably, the Fox family has been suggested to be one of the 3 transcription factors responsible for CLL cell identity.⁵³  In addition, FoxO1 is known to be aberrantly transcriptionally active in B-cell lymphomas⁵²  and to be activated by mutations in some cases.⁶⁷ 

We also identified another GAB1 function by describing its role in increasing the so-called tonic PI3K/AKT signaling by affecting the phosphorylation of the PI3K subunit p85 and AKT^S473. It has been shown that GAB1 contains multiple binding sites for SH2 domain-containing proteins³³  and has a crucial role in PI3K docking to the B-cell membrane connecting BCR-activation with AKT phosphorylation.³⁴  However, the role of GAB1 in tonic AKT activity was not anticipated. It is known that normal B cells depend on constant tonic signaling to maintain their basal PI3K/AKT for their survival²⁵ ; however, the proteins that mediate this tonic signaling remain poorly understood. Constitutive PI3K activity was observed in most peripheral blood CLL samples.^24,26-28  Our data show that upregulated GAB1 in CXCR4^brightCD5^dim CLL cells in peripheral blood contributes to their higher migration capacity and also supports tonic PI3K/AKT activity, which may help to explain the previously noted association of higher GAB1 levels with unfavorable prognosis in CLL.³⁵  The data suggest an interdependent balance of AKT activity and the FoxO1-GAB1 axis, which maintains tonic pAKT above a certain minimal level in CLL cells that do not have access to microenvironmental interactions. This balance is important during ibrutinib therapy, because CLL cells induce the FoxO1-GAB1-pAKT axis and thus partially compensate for the lack of microenvironmental interactions that normally strongly induce pAKT levels. FoxO1’s role in maintaining tonic pAKT^S473 levels is surprising, given that FoxO1 activity was shown to be mutually exclusive with PI3K activity, and FoxO1 was shown to inhibit B-cell proliferation by inducing genes such as p27.⁶⁸  However, we have shown that moderate pAKT activity is not mutually exclusive with FoxO1 nuclear localization in CLL. This finding is in line with data from some B-cell lymphomas that tend to gain an oncogenic constitutive FoxO1 nuclear activity to circumvent negative feedback regulation by strong pAKT activity.⁵²  It is plausible that in the nonproliferative CXCR4^brightCD5^dim subpopulation, FoxO1 induces genes involved in cell cycle arrest (indeed, these cells have higher p27 levels), but also upregulates GAB1 to increase homing of CLL cells and sustain their survival by maintaining tonic pAKT levels.

We have further shown that an inhibitor targeting GAB1 PH domain suppresses BCR-induced and tonic AKT phosphorylation during ibrutinib treatment, and GAB1 inhibition combined with ibrutinib increases cell killing in vitro. Moreover, GAB1 inhibitors also impaired ERK phosphorylation, which is in line with GAB1’s role in ERK activation.^36,69  In addition, GAB1 inhibitors impaired CLL cell migration to an extent similar to ibrutinib. GAB1 may represent a promising therapeutic target the inhibition of which may have a different toxicity profile compared with those of PI3K inhibitors or the CXCR4 agonists such as plerixafor. However, development of this inhibitory therapy will require more potent and selective GAB1 PH-domain inhibitor(s) because their effect on CLL migration and viability was observed only at higher concentrations. It is likely that the tested inhibitors also have off-target effects by inhibiting other PH-domain–containing proteins.⁴³  We have, for example, noted that these GAB1 inhibitors have an effect on cell migration similar to that of GAB1 silencing or KO, but a stronger impact on cell viability. This similarity could be caused by incomplete GAB1 silencing by siRNA or selection of single-cell clones tolerating GAB1 KO during its genomic editing, but could also be caused by the off-target effects of the inhibitors.

In summary, we have described a novel FoxO1/GAB1-dependent regulatory mechanism of CLL homing and maintenance mechanism for tonic AKT signaling (summarized in Figure 7). We have shown that during ibrutinib therapy, CLL cells sustain AKT activity via the FoxO1-GAB1 axis and this can be therapeutically targeted with novel GAB1 inhibitors.

The authors thank Zuzana Smrckova (University of Technology, Brno, Czech Republic) for help with cell motility assays.

This work was supported by the Czech Science Foundation (project no. 20-02566S). This project has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 Research and Innovation Programme (grant agreement no. 802644). This work was supported by the Ministry of Health of the Czech Republic, grant no. NU20-03-00292. All rights reserved. This work was also supported by MH CZ-DRO (FNBr, 65269705) and MUNI/A/1595/2020. The authors acknowledge the CF Genomics CEITEC MU supported by the National Center for Medical Genomics (NCMG) research infrastructure (LM22018132 funded by MEYS CR) and Core Facility Bioinformatics of CEITEC MU for their support in processing the data. This work was supported by the MEYS CR (Large RI Project LM2018129 Czech-BioImaging).

Contribution: V.S. designed the study, performed experiments, analyzed the data, and wrote the article; L.O., L. Kostalova, K.M.L., E.V., J.V., and S.S. performed experiments; G.M.P., K.A.C., and V.B. performed the NGS experiments; J.O. analyzed the NGS data; D.Z. performed the holographic microscopy for the cell migration; T.L. and M.B. performed the cell sorting; M.K.P. and J.K. performed the fluorescence microscopy; K.L. and L. Kren performed the immunohistology; Z.T. and S.Z. provided and discussed the use of GAB1 inhibitors; S.P., Y.B., M.D., A.P., and J.M. provided samples and clinical data; M.M. designed the study, interpreted the data, and wrote the article; and all authors edited and approved the article for submission.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Marek Mraz, Central European Institute of Technology, Masaryk University, Kamenice 5, 625 00 Brno, Czech Republic; e-mail: marek.mraz@email.cz.