Localization of BCR-ABL to F-actin regulates cell adhesion but does not attenuate CML development

P210^BCR-ABL, the oncogene formed by a balanced rearrangement between the c-abl gene on chromosome 9 and the bcr gene on chromosome 22, leads to the formation of the Philadelphia (Ph) chromosome, a diagnostic marker of chronic myelogenous leukemia (CML).^1,2  Expression of P210^BCR-ABL in hematopoietic progenitor cells causes deregulated cell proliferation, decreased programmed cell death, and abnormal adhesion to bone marrow (BM) stroma (for a review, see Wertheim et al³ ). Together, these processes are thought to lead to the defects in cell number and maturation that characterize CML. Defective cellular adhesion may enhance cell cycling and egress of leukemia cells out of the BM into peripheral blood, allowing cells to home to extramedullary sites.^4-6 

P210^BCR-ABL is found primarily in the cytoplasm, closely associated with F-actin through a C-terminal actin-binding domain.^7,8  Additionally, P210^BCR-ABL phosphorylates several proteins of the focal adhesion complex.⁹  Modification of linkages between cancerous cells and the extracellular matrix and interactions between P210^BCR-ABL and the cell cytoskeleton may be important for its oncogenic functions. P210^BCR-ABL likely plays a direct role in modifying cell adhesion; however, the specific functional properties of P210^BCR-ABL that cause defective adhesion have not been identified.

Phosphorylation of intracellular signaling cascades by P210^BCR-ABL tyrosine kinase is critical for cell transformation and development of a CML-like disease in mice.¹⁰  Clinical trials with the specific c-abl tyrosine kinase inhibitor STI-571 (imatinib mesylate [Gleevec]) show that hematologic and cytologic remission can be achieved by attenuating the P210^BCR-ABL tyrosine kinase activity indicating that the tyrosine kinase domain is essential for disease pathogenesis.¹¹  P210^BCR-ABL phosphorylates several proteins involved in cell adhesion—including paxillin, FAK, and CRKL— that are associated with focal adhesion contacts.^9,12  However, neither the attenuation of tyrosine kinase activity by STI-571 nor a lysine-to-arginine point mutation at amino acid (aa) 1176 rendering the tyrosine kinase domain inactive can correct for the adhesive defect caused by P210^BCR-ABL expression in transformed human and murine myeloid cell lines.¹³ 

The fusion of bcr to the N-terminal portion of c-abl leads to cell transformation by deregulating c-abl functions that are normally tightly controlled.⁷  Recent results suggest that the N-terminal coiled-coil (C-C) domain may not be required to activate the P210^BCR-ABL tyrosine kinase because mutants lacking this domain exhibit elevated tyrosine kinase activity in cell lines.^14,15  Nonetheless, the C-C region contributed by bcr leads to dimerization and tetramerization^16,17  of P210^BCR-ABL that may enhance the c-abl tryosine kinase or that may be required for actin binding.⁷ 

Although the c-abl actin-binding function is necessary for F-actin localization and transformation of Rat-1 fibroblasts,¹⁶  the contribution of F-actin localization to either P210^BCR-ABL–induced adhesion defects or CML remains unclear. Failure to oligomerize P210^BCR-ABL by deletion of the bcr C-C domain severely retards the development of myeloproliferative disease (MPD) in mice, most likely because of a decrease in tyrosine kinase activity.^14,15  However, loss of F-actin binding is an alternative explanation because the effect of F-actin localization on P210^BCR-ABL–induced disease has not been studied in a murine BM transplantation model of CML.

We used a quantitative cell adhesion device that can discriminate between changes in cell adhesion by specifically measuring the strength of binding between a population of adherent cells and fibronectin, a major component of the BM extracellular matrix.¹⁸  Using this device, we have shown that expression of P210^BCR-ABL in a myeloblast cell line, 32D, leads to increased cell adhesion in accordance with several published studies using qualitative plate and wash assays.^4,13,19  We also found that expression of a mutant lacking tyrosine kinase activity failed to normalize cell adhesion and that treatment of Meg-01 cells, a P210^BCR-ABL–expressing cell line from a CML patient in blast crisis, with STI-571 had no effect on binding to fibronectin, suggesting that P210^BCR-ABL does not enhance adhesion through inside-out signaling originating at its tyrosine kinase.¹³ 

We constructed several mutants with domain deletions or point mutations in key regions in P210^BCR-ABL and expressed these mutants in 32D cells to determine the region in P210^BCR-ABL that is responsible for enhancing adhesion to fibronectin. We identified 3 regions that are necessary for P210^BCR-ABL to cause elevated adhesion: the N-terminal C-C domain that facilitates oligomerization and F-actin localization, the bcr sequences between aa 163 to 210, and the C-terminal actin-binding domain of c-abl. We were particularly interested in the relationship between F-actin localization, elevated cell adhesion to fibronectin, and MPD development in our mouse model. Here we report that deletion of the C-terminal actin-binding domain had no effect on development of the MPD. Together, our findings indicate that localization of P210^BCR-ABL to the actin cytoskeleton is responsible for abnormal adhesion to fibronectin; however, F-actin binding, and potentially deregulated cell adhesion to fibronectin, is not a primary determinant for murine CML development.

To create Δ(1-63) BCR-ABL lacking the N-terminal C-C domain, a 1.4-kb HpaI/BamHI fragment was released from MigP210^BCR-ABL.²⁰  An ATG start codon was inserted upstream of aa 64 by generating a polymerase chain reaction (PCR) fragment using the 5′ primer CT CTC GAGGTTAAC ATG GCC AAG GAA AAG AAG AGC (boldface and underlined sequences are recognized by XhoI and HpaI, respectively) and the 3′ primer ACG TAG AAG GGC TTC TCG. This fragment was digested with HpaI and BamHI and was ligated to HpaI/BamHI-digested MigP210^BCR-ABL. The resultant plasmid was cut with BamHI, and the 1.4-kb fragment of P210^BCR-ABL was inserted in the correct orientation to construct MigΔ(1-63) BCR-ABL. Δ(1-63) BCR-ABL was released with HpaI and EcoRI and was ligated to the analogous site in the pK1 vector, a murine stem cell leukemia virus (MSCV)–based vector that is identical to MigRI except that an internal ribosomal entry site (IRES)–puromycin resistance gene was inserted in place of the IRES-green fluorescence protein (IRES-GFP) (kind gift from Karen Ehrmann and Stephen Emerson, University of Pennsylvania, Philadelphia).

The (1-162) BCR-ABL mutant was cloned through the release of full-length BCR-ABL from MigP210^BCR-ABL with EcoRI. The c-abl portion was isolated by digestion with HincII and was cloned into the HpaI–EcoRI site of the pK1 vector. The BCR-ABL breakpoint at aa 162 was generated through sequential PCR amplification of a fragment containing the first 486 bp bcr and 3′ overlapping abl sequences (fragment 1, 5′ primer 77791¹⁴  TCA CTC CTT CTC TAG GC and 3′ primer GCCGCTGAAGGGCTTTGC GGA TCC GCT CGA AG, abl sequences underlined) and a second fragment (fragment 2, 5′ primer CTT CGA GCG GAT CCG CAAAGCCCT TCAGCGGC and 3′ primer 77793¹⁴  AGC AGA TAC TCA GCG GC) containing 5′ overlapping bcr sequences. Fragments 1 and 2 were used as templates to generate a fragment that contained the BCR-ABL breakpoint at aa 162 using primers 77791 and 77793. This fragment was digested with HincII and was ligated in the correct orientation into the HpaI site of pK1 already containing the c-abl fragment. The BCR-ABL cassettes in MigP210^BCR-ABL, MigRI 1-210 BCR-ABL,¹⁴  MigRI (1-63) BCR-ABL,¹⁴  and MSCV 2.2 ΔActin BCR-ABL²¹  were released with EcoRI and were cloned into the pK1 vector at the corresponding site.

High-titer retroviral supernatants were generated by transfection into Bosc23 cells.²⁰  Viral titers were normalized by GFP expression of transduced NIH 3T3 cells (MigRI constructs) or colony formation of puromycin-resistant NIH 3T3 cells (pK1 constructs).¹⁴  32D cells²²  were maintained in 32D culture media¹³  and were transduced by spinoculation.²³  At 48 hours after transduction, 32D cells transduced with vectors containing the puromycin resistance gene were selected with puromycin (1.5 μg/mL) for 4 days and maintained in 32D media containing 10% WEHI supernatant. 32D cells transduced with GFP as a surrogate marker for P210^BCR-ABL expression were sorted for GFP expression (MigRI) or were selected for growth in the absence of interleukin-3 (IL-3) (MigP210^BCR-ABL) for 3 to 4 days, after which WEHI was reintroduced into the culture medium to maintain consistent growth rates.

Before they were spun, all cells were washed and transferred to a Tris-based buffer as described.¹³  Cells were treated with control or adhesion blocking reagents for 10 minutes before incubation with fibronectin. Functional blocking antibody PS/2 (anti-α₄ integrin) was a gift from C. Buck (Wistar Institute, Philadelphia, PA) and L. Terracio (University of South Carolina, Columbia).^24,25  BMA5 (anti-α₅ integrin) was obtained from Chemicon (Temecula, CA).²⁶ 

Circular glass coverslips (Fisher Scientific, Hampton, NH) were adsorbed with 10 μg/mL fibronectin (Becton Dickinson, Franklin Lakes, NJ) and were blocked with 1% bovine serum albumin (BSA). The cell detachment device was operated as described.²⁷  Adhesion experiments were carried out in the absence of growth factors. Coverslips were analyzed by recording the adherent fraction of cells at various radial distances from the center. The shear stress, τ, imparted on the cells was calculated by the equation τ = 0.800 r (ρμω³)^1/2, where τ is proportional to r, the radial distance; μ and ρ are the fluid dynamic viscosity and density, respectively; and ω is the angular velocity of the spinning disk. The fluid flow that is established by a spinning disk has been characterized,^28,29  and the fraction of beads or cells under shear flow that remain bound to an adhesive surface is known to follow a sigmoid-shaped model.³⁰  Adhesion profiles for this study were prepared²⁷  and were accepted for R² ≥ 0.70. This is an arbitrary value to ensure that the cell distribution data conform to the numeric model; however, increasing or decreasing this value does not significantly affect the measurement of cell adhesion (data not shown). This criterion was used for all conditions with the exception of runs using cells incubated with reagents that severely block cell adhesion or when cells were incubated on matrices of BSA alone because of the few cells that remained attached after a spin.

The critical shear stress, τ₅₀, was measured as the surface shear stress, τ, which is at an adherent fraction of 0.50 along the fitted curve. Determination of significance between mean values of critical shear stress, τ₅₀, was determined using a 2-tailed Student t test with significance defined as P ≤ .050. Data are presented as mean ± SD.

Donor BM was harvested from femurs and tibias of C57BL/6 mice that had been treated with 1 mg/kg 5-flurouracil (5-FU) 4 days previously. BM was transduced with titer-matched retroviral supernatants, recipient mice were killed when they exhibited evidence of disease, and tissues were processed for analysis as described.^14,20 

Western blots were performed on lysates from 32D cells or from primary mouse tissue. 32D cells were incubated overnight in 32D medium with 10% WEHI, and Western blots were prepared as described using the 8E9 antibody (PharMingen, San Diego, CA) to reveal BCR-ABL.¹³ 

For flow cytometry analysis, 32D cells transduced with the control vector MigRI or MigP210^BCR-ABL were handled according to the same procedure used to prepare cells for adhesion experiments except that cells were not preselected for GFP or P210^BCR-ABL expression. Integrin expression was assessed in triplicate for each population. Three populations of MigRI cells and 2 populations of MigP210^BCR-ABL were transduced separately. Anti-α₄ (R-2; PharMingen) and anti-α₅ (5H10-27; PharMingen) antibodies were used to stain for integrins, and cells were analyzed by flow cytometry.¹³ 

NIH 3T3 cells were transduced with retroviral constructs encoding P210^BCR-ABL, ΔActin, Δ(1-63) BCR-ABL, (1-63) BCR-ABL, or the pK1 control plasmid and were selected for 4 days with puromycin. Cells were stained to reveal F-actin, BCR-ABL, and the cell nucleus using confocal microscopy.¹³ 

The spinning disk cell detachment device uses fluid flow to detach cells bound to a monolayer of fibronectin. Detachment force can easily be determined by the properties of the buffer and distance of the cells from the center of the coverslip (see “Materials and methods”). The spinning disk cell detachment device measures discrete changes in cell adhesion and has been used to characterize individual activation states in α₅β₁ and α_vβ₃ integrins in cells bound to fibronectin and other extracellular matrix molecules.^30-33 

To determine which integrin receptors primarily mediate adhesion in P210^BCR-ABL-expressing cells, we transduced 32D cells with a retrovirus coexpressing P210^BCR-ABL and GFP (Mig P210^BCR-ABL) or GFP alone (MigRI). Incubation of control MigRI cells on BSA coverslips without fibronectin results in few cells remaining bound after a spin and gives a τ₅₀ of 3.1 dyne/cm² (Figure 1A). This condition represents background or nonspecific adhesion, which is low compared with that for cells bound to fibronectin. In contrast, the τ₅₀ of MigRI-transduced cells incubated on fibronectin-coated coverslips is 22.7 dyne/cm², greater than a 7-fold increase in adhesion compared with that for BSA alone (Figure 1A). The increased binding is observed as a rightward shift in the adhesion profile as cells remain attached at higher shear stress (Figure 1A). Expression of MigP210^BCR-ABL in 32D cells leads to a further rightward shift in the adhesion profile and an increase in the τ₅₀ to 36.4 dyne/cm² (Figure 1A). Comparison of 2 separate populations of MigP210^BCR-ABL and MigRI cells shows that MigP210^BCR-ABL leads to a 1.5- to 1.6-fold increase in fibronectin binding (P ≤ .003; Figure 1B). This increase in adhesion is similar to that observed in our previous report that P210^BCR-ABL caused a 1.7-fold increase in binding between 32D cells and fibronectin when P210^BCR-ABL was expressed from the pK1 vector containing the puromycin resistance gene in place of GFP.¹³ 

Binding of integrins to fibronectin requires intracellular energy stores and extracellular divalent cations as cofactors. To show that cells specifically bind fibronectin, control MigRI cells were preincubated with a soluble Arg-Gly-Asp (RGD)–containing peptide that occupies the fibronectin binding site on integrins that recognize the RGD sequence. Incubation of cells with the RGD-containing peptide led to a significant decrease (P = .001) in adhesion to fibronectin. Cells were also treated with 0.02% sodium azide (NaAZ) to reduce energy stores by blocking oxidative phosphorylation or 1 mM EDTA (ethylenediaminetetraacetic acid) to chelate free cations. Both treatments decreased binding (10% of control for NaAZ and 16% of control for EDTA) to levels achieved with BSA alone (13% of control), indicating that adhesion in our system requires energy and divalent cations, properties of all integrin-fibronectin linkages.

The α₄β₁ and α₅β₁ integrins are the primary cell surface receptors that bind fibronectin. To determine whether P210^BCR-ABL influences the relative expression of these integrins, unselected 32D cells transduced with MigP210^BCR-ABL or MigRI were stained with antibodies to the extracellular region of the α₄ or α₅ integrin subunits (Figure 2A). Comparison between GFP⁺ MigP210^BCR-ABL and GFP⁺ MigRI cells suggests that P210^BCR-ABL may cause a 1.2-fold increase in α₅β₁ expression; however, this change was not statistically significant across all populations of cells tested. Expression of α₄β₁ was essentially unchanged. These findings are comparable with those of other published reports that suggest P210^BCR-ABL expression may lead to minimal changes in integrin expression.^4,19  These small alterations in integrin levels may contribute to, but are unlikely to account for, the entire increase in adhesion in P210^BCR-ABL cells.¹³ 

To determine the extent to which α₄β₁ and α₅β₁ are involved in binding to fibronectin in our system, 32D cells selected for the expression of MigRI (Figure 2B) or MigP210^BCR-ABL (Figure 2C) were incubated with blocking antibodies to α₄β₁, α₅β₁, or both integrins simultaneously. Blocking α₄β₁ function in either MigRI or MigP210^BCR-ABL cells had no effect on cell adhesion (MigRI, P = .439; Figure 2B) (MigP210^BCR-ABL, P = .537; Figure 2C). In contrast, attenuation of adhesion by the α₅β₁ integrin reduced binding to 25% and 63% of untreated MigRI (P < .001; Figure 2B) and MigP210^BCR-ABL cells (P < .001; Figure 2C), with nonspecific adhesion representing approximately 13% and 17% of adhesion, respectively. Blocking both α₄β₁ and α₅β₁ did not significantly decrease binding to fibronectin in MigRI cells beyond that attenuated by inhibiting α₅β₁ alone (P = .163); however, adhesion in MigP210^BCR-ABL cells was significantly lower than that achieved using only a blocking antibody to α₅β₁ (P = .007). It is likely that α₄β₁ plays a minor role in adhesion in MigP210^BCR-ABL cells and is low compared with the involvement of α₅β₁.

We compared P210^BCR-ABL mutants that lack functional C-C, Y177, or actin-binding domains to test the contribution of these regions on adhesion to fibronectin (Figure 3A). Many of these mutants were tested in our murine CML model, and we found that the deletion of the C-C domain or the mutation of Y177 was sufficient to prevent induction of the wild-type MPD.¹⁴  Expression of P210^BCR-ABL in 32D cells alleviates the requirement of IL-3 for survival.³⁴  After selection in puromycin for 4 days, each cell population was assessed for the ability to grow in IL-3–free media (Figure 3B). All mutants survived without IL-3 supplementation; however, cells transduced with the control vector pK1 rapidly died within 48 hours, as assessed by trypan blue inclusion. Growth of 32D cells transduced by all mutants except Δ(1-63) BCR-ABL was similar to wild-type P210^BCR-ABL. Deletion of the C-C domain in Δ(1-63) prevents oligomerization of P210^BCR-ABL.^14,16  Although able to grow in IL-3–free media, the proliferation rate of cells transduced with this mutant was decreased relative to wild-type P210^BCR-ABL (Figure 3B). This mutant also expressed lower tyrosine kinase activity, suggesting that this may be a cause of decreased growth (Figure 3C).

Two independently transduced populations of each mutant were repeatedly spun, and the critical shear stress of each spin is presented for 1 of the 2 populations tested (Figure 4). Of the mutations known to affect F-actin binding, both Δ(1-63) and ΔActin showed significantly lower adhesion in our assay compared with 2 populations of wild-type P210^BCR-ABL (Δ(1-63), P < .001 and P < .001; ΔActin, P = .036 and P < .001; Figure 4), suggesting that association with F-actin may be critical for enhanced binding to fibronectin. To confirm that Δ(1-63) resulted in reduced localization to F-actin, the control vector pK1, Δ(1-63), and ΔActin cells were transduced into NIH 3T3 fibroblasts. Cells were stained for P210^BCR-ABL localization and F-actin filaments (Figure 5). The pattern of wild-type P210^BCR-ABL staining in NIH 3T3 cells resembled linear filament-like strands that overlap with F-actin. In contrast, ΔActin- and Δ(1-63)–transduced NIH 3T3 fibroblasts showed a diffuse pattern of anti-abl staining in the cytoplasm, consistent with previously published results.⁷  A reduction in F-actin colocalization with the ΔActin mutant had been previously demonstrated in 32D cells.²¹ 

The (1-63) mutant, in which the C-C domain was fused to c-abl, had decreased adhesion compared with wild-type P210^BCR-ABL (P < .001 and P < .001; Figure 4), but the staining pattern indicated that it retains the ability to colocalize with F-actin (Figure 5). This suggests that the mutant leads to lower adhesion through a mechanism that is distinct from Δ(1-63) because the c-abl sequences are conserved among the 2 mutants, but each retains distinct regions of bcr.

One possibility is that bcr sequences downstream of the C-C domain may recruit proteins to the cell cytoskeleton that are necessary for adhesion. To investigate this possibility, we attempted to map the minimal region in bcr between aa 64 and aa 927 that facilitates P210^BCR-ABL cell adhesion. We found that a (1-210) BCR-ABL mutant conferred high binding, similar to wild-type P210^BCR-ABL, compared with cells expressing the (1-63) mutant (Figure 4). This suggests that aa 64-210 is a critical region in bcr that leads to increased binding to fibronectin.

To narrow the critical region within aa 64-210, we constructed a (1-162) mutant that fuses the first 162 aa of bcr onto c-abl. This mutant results in transformation of 32D cells and increased whole cell phosphotyrosine levels (Figure 3B-C). When assessed for its ability to bind fibronectin in our spinning disk adhesion device, 32D cells expressing this mutant showed decreased levels of adhesion that were in the same range as the (1-63) mutant, narrowing the critical region to between aa 163 and aa 210.

Tyr177 is the only tyrosine that may bind SH2-containing proteins within this region. Mutation of tyrosine to phenylalanine (Y177F) abrogates binding to GRB2^14,35 ; however, similar to P210^BCR-ABL, it caused increased binding to fibronectin (Figure 4). This suggests that the P210^BCR-ABL–GRB2 interaction is not responsible for increased adhesion in P210^BCR-ABL–expressing 32D cells.

Together, our adhesion experiments identify 3 regions in P210^BCR-ABL that contribute to the ability of wild-type P210^BCR-ABL to increase the magnitude of binding between 32D cells and fibronectin. The regions are the N-terminal C-C domain that facilitates oligomerization and enhances F-actin localization, the bcr sequences spanning aa 163 to 210, and the C-terminal actin-binding domain mediating F-actin localization.

Although P210^BCR-ABL colocalizes with F-actin, the contribution of F-actin localization to CML development has not been characterized in a BM transplantation model of CML. Evaluation of the ability of the ΔActin mutant to induce wild-type MPD may indicate whether the failure of Δ(1-63) to cause murine CML^14,15  is attributed to its reduction in F-actin localization.

All mice given ΔActin-expressing BM developed overt disease within 14 days, indicated by elevated white blood cell (WBC) counts (109 000 ± 61 000 cells/μL) (Figure 6A). Wild-type P210^BCR-ABL mice also developed disease during this interval (Figure 6A). MigRI mice remained healthy with normal WBC counts (less than 10 000 cells/μL). At necropsy, both wild-type P210^BCR-ABL and ΔActin mice exhibited hepatosplenomegaly, pulmonary hemorrhage, and absent lymphadenopathy. Histologic examination showed infiltration of the BM, liver, spleen, and lungs by neutrophils (Figure 6B). MPD was similar between P210^BCR-ABL and ΔActin except that pulmonary hemorrhage was greater in the P210^BCR-ABL mice. The ΔActin protein could be detected in spleens of these animals at the appropriate size using an antibody to the c-abl tyrosine kinase domain, excluding the possibility of reversion to wild-type in ΔActin mice (Figure 6C). BM, spleen, and peripheral blood were collected from all mice at the time of killing, and their immunophenotypes were assessed using flow cytometry (Figure 6D). Most cells from all 3 organs of diseased mice (wild-type P210^BCR-ABL and ΔActin) were GFP⁺, indicating that the transduced donor cells became the dominant population. In both cohorts, these cells stained predominantly GR-1⁺/Mac-1⁺, indicating they were of granulocytic lineage. These cells also stained negatively for T-cell and B-cell markers, Thy 1.2 and B220, respectively (data not shown). Increased GFP^–/GR-1⁺/Mac-1⁺ cells were observed in wild-type P210^BCR-ABL and ΔActin mice compared with MigRI. This is likely because of a trans effect caused by cytokines released from GFP⁺/GR-1⁺/Mac-1⁺ cells.^10,36  Together, these results indicate that preventing the association of P210^BCR-ABL with F-actin is dispensable for CML development in our mouse model but is necessary for P210^BCR-ABL to enhance cell adhesion.

Abnormal binding between hematopoietic progenitor cells and BM stroma has been proposed, in addition to loss of programmed cell death and enhanced cell proliferation, as an important contributor to the onset of leukemia.³⁷  Several studies using BM from CML patients^5,38  or hematopoietic cell lines^4,19  expressing P210^BCR-ABL have documented an adhesion defect between cells and fibronectin, a major extracellular molecule of the BM microenvironment.¹⁸  Contact between P210^BCR-ABL–expressing cell lines and fibronectin promotes entry into the cell cycle, suggesting that enhanced binding may promote deregulated cell proliferation and tumor formation.⁴  Other studies indicate that P210^BCR-ABL may assist in the extravasation of leukemia cells out of the BM and into the peripheral circulation, where abnormally high levels of WBCs are typically present in untreated CML patients.^5,38  Fibroblasts expressing P210^BCR-ABL support this theory by exhibiting increased motility, accumulation of intracellular F-actin, and formation of filopodia compared with parental cells cultured on fibronectin.³⁹ 

In vitro studies such as these provide a context for understanding the relationship between P210^BCR-ABL expression and cell adhesion and motility. What remains to be determined is the contribution of defective adhesion to development of the CML disease. To investigate this relationship, we sought to understand how the P210^BCR-ABL protein regulates cell adhesion and then to determine whether this process is contributory to murine CML. We used 2 experimental techniques to link cell adhesion with disease development.

We identified 3 regions in wild-type P210^BCR-ABL—the N-terminal C-C domain, aa 163-210, and the C-terminal actin-binding domain—that, when deleted, lead to reduced adhesion to fibronectin compared with wild-type P210^BCR-ABL. The C-C domain enhances the c-abl tyrosine kinase and the F-actin–binding domain.⁷  It is unlikely, however, that the depreciation in tyrosine kinase activity in the Δ(1-63) mutant led to a decrease in adhesion to fibronectin because we have previously shown that the attenuation of tyrosine kinase activity with STI-571 or a point mutation in the tyrosine kinase domain does not diminish the effect of P210^BCR-ABL on adhesion.¹³  Rather, the decreased ability of the Δ(1-63) mutant to colocalize with F-actin is the likely mechanism leading to decreased adhesion with this mutant.

The observation that interactions between P210^BCR-ABL and the cell cytoskeleton are necessary for P210^BCR-ABL–mediated increased adhesion is consistent with observations in fibroblasts that indicate that P210^BCR-ABL expression leads to abnormal cytoskeletal architecture, altered F-actin formation, and increased filopodia formation.³⁹  An intact actin cytoskeleton is necessary for binding fibronectin in our system because complete depolymerization of F-actin with cytochalasin D abrogates all binding to fibronectin (Figure 2C).

Evidence from the literature shows that P210^BCR-ABL expression directly alters integrin-cytoskeletal linkages by modifying F-actin. Human megakaryoblast cells, Mo7e, expressing P210^BCR-ABL or CD34⁺ cells from CML patients are inefficient in the formation of integrin caps, or clusters, containing α₄β₁ and α₅β₁ integrins.⁴⁰  P210^BCR-ABL is thought to contribute to increased F-actin polymerization that retards integrin mobility within the cell membrane, preventing integrin clustering.⁴⁰  Treatment of CML CD34⁺ cells with low-dose cytochalasin D increased cap formation and restored adhesion to near normal levels.

Our results favor a mechanism in which P210^BCR-ABL binds to F-actin filaments to directly modify F-actin or, more likely, to recruit effector molecules potentially through the aa 163-210 region of P210^BCR-ABL to the cell cytoskeleton. The significance of the aa 163-210 region to cell adhesion, is unclear. The Y177 binding domain is located in this region and binds GRB2, which has been suggested to activate RAS through the guanine exchange factor SOS.⁴¹  Others and we have shown that wild-type P210^BCR-ABL–GRB2 binding at this region is necessary for efficient CML development.^14,15,42  However, prevention of GRB2 binding through a Y177F mutation has no effect on enhanced adhesion, suggesting that GRB2 binding to P210^BCR-ABL is noncontributory to cell adhesion. Recently, the P210^BCR-ABL/GRB2 interaction was shown to recruit the scaffolding adapter GAB2 to the SH3 domain of GRB2. This association enhances PI3K/AKT and RAS/ERK activation and is required for efficient myeloid transformation in murine BM cells.⁴³  Interestingly, binding of GRB2 to P210^BCR-ABL is required for fibroblast transformation⁴⁴  but is generally not needed to transform hematopoietic cell lines. This is likely because of alternative pathways sufficient to activate RAS in the absence of GRB2 binding at Y177 in cell lines such as 32D cells.⁴⁵  One possibility not excluded by our findings is that RAS may be activated in 32D cells expressing the Y177F mutant and may enhance adhesion through pathways that do not involve GRB2 binding at this site.

Mouse models of CML using either transgenic mice or mice that underwent BM transplantation are useful tools to evaluate how P210^BCR-ABL leads to leukemia. Leukemia induced by the ΔActin mutant is indistinguishable from wild-type P210^BCR-ABL. Our findings also provide insight into the role of the C-C domain and argue that the inability of the Δ(1-63) mutant to efficiently colocalize with F-actin is an unlikely explanation for its inability to cause a CML-like disease in mice. Rather, the effect on tyrosine kinase activity and protein-protein interactions⁴⁶  is the likely contribution of oligomerization to the efficient induction of CML. Others have shown that complete deletion of the SH3 domain or mutation of the critical intramolecular binding motifs that regulate the tyrosine kinase inhibitory role of the SH3 domain restores the tyrosine kinase activity of a C-C mutant and enables the induction of the MPD.^15,47  This suggests that the moderate reduction in tyrosine kinase activity is likely to account for the inability of the C-C mutant to cause the MPD rather than its inability to colocalize with F-actin.

Our findings contrast those from mice transgenic for the expression of P190^BCR-ABL lacking the C-terminal F-actin–binding domain. These mice were characterized by an increase in leukemia latency and a decrease in development of acute B-cell leukemia.⁴⁸  The reason for these disparate findings is likely attributed to the intrinsic differences between the 2 models that produce different diseases (MPD in our model and acute lymphoid leukemia in the model of Heisterkamp et al⁴⁸ ). Although unlikely, differences between actin-binding mutants used for each study, such as the size of the C-terminal deletion or the additional bcr regions present in P210^BCR-ABL and absent from P190^BCR-ABL, may affect disease onset.

In summary, we show that the localization of P210^BCR-ABL to F-actin is not required for MPD induction but is needed for the elevated adhesion to fibronectin seen in 32D cells expressing wild-type P210^BCR-ABL. We used 2 independent assays to assess adhesion and oncogenicity; however, we were unable to find a consistent correlation between normalization of adhesion to fibronectin and attenuation of MPD in our murine model. This suggests that abnormal cell-fibronectin binding is unlikely to be a major independent contributor to murine CML development. Alternatively, our results implicate the dysregulated signaling caused by the P210^BCR-ABL tyrosine kinase and other motifs (eg, Y177) as critical determinants of CML induction. The relative significance of the fusion of bcr sequences upstream of c-abl is best observed with the Δ(1-63) mutant, in which our results suggest that a decline in tyrosine kinase below a critical threshold blocks the ability of this mutant to induce murine CML. Together, our studies provide further evidence that even a slight disruption in the abnormal signaling patterns of P210^BCR-ABL can significantly impact the emergence of leukemia and direct attention to several regions in P210^BCR-ABL for targeted therapy.

We thank Andrea Carpenter, Bill DeMuth, Kevin Forsythe, Gladys Grays-Board, Fred Karnell, Lara Lynch, Joey Plumb, and Lanwei Xu for their technical assistance. The flow cytometry studies were performed in the University of Pennsylvania Cancer Center Flow Cytometry and Cell Sorting Shared Resource (supported in part by the Lucille B. Markey Trust and the National Institutes of Health). The Gastroenterology Cell Morphology Core at the University of Pennsylvania processed tissue sections and was supported by National Institutes of Health Center grant P30-DK50306.