ReviewEVOLUTION

ABL Tyrosine Kinases: Evolution of Function, Regulation, and Specificity

See allHide authors and affiliations

Science Signaling  14 Sep 2010:
Vol. 3, Issue 139, pp. re6
DOI: 10.1126/scisignal.3139re6

This article has a correction. Please see:

Abstract

ABL-family proteins comprise one of the best conserved branches of the tyrosine kinases. Each ABL protein contains an SH3-SH2-TK (Src homology 3–Src homology 2–tyrosine kinase) domain cassette, which confers autoregulated kinase activity and is common among nonreceptor tyrosine kinases. This cassette is coupled to an actin-binding and -bundling domain, which makes ABL proteins capable of connecting phosphoregulation with actin-filament reorganization. Two vertebrate paralogs, ABL1 and ABL2, have evolved to perform specialized functions. ABL1 includes nuclear localization signals and a DNA binding domain through which it mediates DNA damage-repair functions, whereas ABL2 has additional binding capacity for actin and for microtubules to enhance its cytoskeletal remodeling functions. Several types of posttranslational modifications control ABL catalytic activity, subcellular localization, and stability, with consequences for both cytoplasmic and nuclear ABL functions. Binding partners provide additional regulation of ABL catalytic activity, substrate specificity, and downstream signaling. Information on ABL regulatory mechanisms is being mined to provide new therapeutic strategies against hematopoietic malignancies caused by BCR-ABL1 and related leukemogenic proteins.

Introduction

ABL (1) genes were first encountered in the guise of a tumor gene in the Abelson murine lymphosarcoma virus (2). The product of the virally transduced oncogene, v-abl, showed tyrosine kinase activity (3, 4) and was determined to be an altered form of cellular Abl1 (encoded by the c-Abl gene) (5). The human ortholog of Abl1 was later identified as part of a mutationally activated fusion oncoprotein, BCR-ABL1 (6), common in chronic myeloid leukemia (CML) patients. ABL2 [also known as Abl-related gene or Arg (7)] is a paralog of ABL1 identified by sequence similarity (8). Direct biochemical analyses, cell biology observations, animal experiments, and human leukemia studies have produced working models of ABL function in normal (9, 10) and transformed cells (11).

Central to the biochemical and physiological functions of ABL proteins are their combination of a regulated SH3-SH2-TK (Src homology 3–Src homology 2–tyrosine kinase) domain cassette with cytoskeletal protein– and DNA-binding domains (Fig. 1), a combination that confers unique signaling capabilities. This Review focuses on ABL protein evolution, function, and mechanisms of regulation.

Fig. 1

ABL domain structure and motif conservation. Linear depiction of human ABL1 and ABL2 (long, b isoforms), D. melanogaster Abl, and M. brevicollis Abl1 and Abl2. my, myristoylation site (*site present but modification not verified); Tyr, tyrosine; G BD, G-actin–binding domain; and MT BD, microtubule-binding domain. Blue triangle, NES; magenta triangle, NLS; and green triangle, proline-rich motif with capacity to bind SH3 or WW domains.

The ABL Subfamily of Tyrosine Kinases

Evolution and domain structure

ABL genes are found in all metazoans, which suggests that their structure and function were fixed relatively early in tyrosine kinase evolution. Vertebrate genomes encode two closely related paralogs, ABL1 and ABL2, with conserved domain structure (Fig. 1) and intron-exon boundaries (www.ensembl.org), which suggests a gene duplication origin. Nonvertebrate metazoans (Strongylocentrotus purpuratus, Caenorhabditis elegans, and Drosophila melanogaster) have a single ABL gene, which shows strong conservation through the SH3-SH2-TK cassette and terminal actin-binding domain. The choanoflagellate Monosiga brevicollis, a unicellular protist and proximal progenitor of metazoans, encodes two nonreceptor tyrosine kinases that align throughout the ABL SH3-SH2-TK domain cassette but terminate soon thereafter (Fig. 1). This suggests an early origin for ABL kinases, with the addition of an extended carboxy terminus during the metazoan radiation.

The primary clade of human nonreceptor tyrosine kinases [see (12) for kinome dendrogram] has 22 members, all but three of which share an SH3-SH2-TK domain structure. The reiteration of this domain cassette implies strong selective pressure, which reflects multiple contributions to function and regulation (1315). Alignments using the SH3-SH2-TK sequence from human and fly nonreceptor tyrosine kinases show that ABL1 and ABL2 are more closely related to each other, and to their fly ortholog, than any other nonreceptor tyrosine kinase is related to its closest paralog or ortholog (Fig. 2). Human ABL1 and ABL2 are more than 90% identical in this domain cassette (Fig. 3).

Fig. 2

SH3-SH2-TK–family proteins. A dendrogram of the SH3-SH2-TK sequence cassettes from H. sapiens, D. melanogaster (Dm), and the protist M. brevicollis (Mb). The dendrogram was created using CLC Sequence Viewer (www.clcbio.com).

Fig. 3

Sequence alignment of human ABL1b and ABL2b. Identity (*) and strong or weak sequence conservation [(:) colon for strong conservation; (.) period for weak conservation] are indicated. Domains are indicated with the following underlines: plain, SH3; heavy, SH2; dashed, TK; or wavy, terminal F-actin binding. Shading is used to highlight the following features: sequences deleted in human leukemogenic ABL fusions (gray), phosphorylation sites with known function (red), SH3 and WW domain proline-rich ligands (yellow), DFG kinase (purple), K- or R-rich NLS motifs (pink), and NES (blue). Additional reported phosphorylation sites are indicated with a red overline. Alignment was performed using ClustalW (247).

Downstream of the TK domain, vertebrate ABL proteins contain domains for binding cytoskeletal components (F-actin, G-actin, and microtubules) and DNA. Of these, only the carboxy-terminal F-actin–binding domain is identifiable in sea urchin (S. purpuratus) and fruit fly (D. melanogaster) Abl (Fig. 4). In addition, fruit fly Abl is overwhelmingly localized to the cytoplasm and plasma membrane (1619), which suggests that it lacks the nuclear functions attributed to vertebrate Abl1. Considering proposed mechanisms of gene duplication and evolution [reviewed in (20)], it is tempting to speculate that the fixation of vertebrate Abl paralogs is attributable to the selective value of new functions, rather than to a simple partition of preexisting functions. During the “fate-determination phase” (20) following Abl gene duplication, nuclear-localization and DNA binding properties could have emerged in the nascent Abl1, whereas cytoskeletal remodeling capabilities were expanded and refined in Abl2. These changes might have been subsequently preserved by the complex demands of vertebrate development and differentiation, as well as the requirement for enhanced DNA damage repair capabilities in these longer-lived organisms.

Fig. 4

Sequence alignment of ABL proteins from representative organisms. Representative Chordates are: Primates (Homo sapiens, Hs); Rodentia (Mus musculus, Mm); Marsupialia (Monodelphis domestica, Md); Aves (Gallus gallus); Amphibia (Xenopus tropicalis, Xt); and Pisces (Danio rerio, Dr). Representative Echinoderm is sea urchin (S. purpuratus, Sp); representative Arthropod is fruit fly (Drosophila melanogaster, Dm). Sequence identity and similarity are denoted as in Fig. 3. Color shading corresponds to the description in Fig. 3. Note that the first line shows only those vertebrate ABL2 sequences for which a long (myristoylated) isoform was identified. A few large insertions (mostly in AblDm) relative to other sequences are indicated. Sequences were obtained from Ensemble and GenBank. Alignment was performed using ClustalW.

Three PxxP (21) motif SH3-binding sites belonging to class 2 [αPxαPx(R or K), reviewed in (22)] are well conserved between ABL1 and ABL2 (Fig. 3). The first is conserved in sea urchin but not fruit fly, the third is apparent in fruit fly but not sea urchin, and the second is absent from both (Fig. 4). ABL1 proteins have an additional class 2 motif that is not found in ABL2 or in nonvertebrates (Fig. 4). Human ABL2 has a single class 1 motif [(R or K)xαPxαP] that is missing from all ABL orthologs except sea urchin Abl (Fig. 4). A PPxY motif implicated in WW domain binding (23) is present in the TK domain, and extended proline-rich sequences with potential for either SH3- or WW-domain interactions are found at variable positions near the carboxy termini of metazoan ABL proteins (Fig. 4). There does not appear to be a simple progressive development of these binding motifs, which suggests a complex evolutionary path.

Three K- and R-rich nuclear localization signal (NLS) motifs have been identified in ABL1 (24), consistent with its partial distribution in the nucleus. The equivalent positions in ABL2 have less K and R content (Fig. 3), consistent with its cytoplasmic localization. Alignments with sea urchin and fruit fly Abl proteins (Fig. 4) suggest that NLS function developed relatively late. An L-, V-, and I-rich nuclear export signal (NES), probably engaged by XPO1 (exportin 1) (25), is found in the ABL1 carboxy terminus and is likely responsible for the ability of ABL1 to shuttle between nucleus and cytoplasm (26). Sequence alignment (Fig. 4) suggests that this NES developed gradually and in concert with the nuclear functions of ABL1.

ABL gene expression

In mice, both Abl genes are expressed widely, but not uniformly [(27) and http://biogps.gnf.org]. Relatively little is known, however, about the promoter elements and transcription factors that regulate the expression of ABL genes. ABL1 mRNA translation is silenced by microRNA-203 (miR203) (28), and sequence analysis suggests additional posttranscriptional regulators for ABL1 (miR196) and ABL2 (miR26 and miR1297) (http://targetscan.org).

Each human ABL gene gives rise to two primary transcripts (a and b) using alternate exons. The first exons of ABL1b and ABL2b encode target sites for cotranslational myristoylation (glycine at position 2). The relative proportions and tissue distribution of each isoform have not been established.

Role in development

There is ample evidence for ABL functions during development. The single Drosophila Abl gene is essential for viability (29, 30) and plays a key role in central nervous system development (29). Abl controls growth cone guidance and synaptogenesis, acting downstream of the Robo receptor (31, 32) and antagonistically to tyrosine phosphatases (33, 34). Abl is also required during dendritic morphogenesis (35). These developmental functions require reorganization of actin (34, 35) and microtubule (36) structures, and they involve the Abl phosphorylation substrates Enabled (31, 34, 35, 37), Abi (Abl interactor protein) (38), and Cables (32). Abl has a critical role in the developmental morphogenesis of epithelial tissues, including head involution and dorsal closure (39, 40), ventral furrow formation (18), and wing construction (41). These epithelial cell functions reflect, at least in part, the contribution of Abl to actin dynamics (18, 40). C. elegans Abl controls epithelial morphogenesis (42) but, in contrast to Drosophila, a null mutation in the C. elegans Abl gene confers only conditional phenotypes. These include sensitivity of germ cells to various types of stress (43, 44), rescue of apoptotic cell engulfment defects (45), and altered Shigella flexneri pathogenesis (46).

Both vertebrate Abl1 and Abl2 are required for normal development. Mice with a disrupted Abl1 gene show defective hematopoiesis and low viability (47, 48), osteoporosis (49), decreased systolic blood pressure (50), and cardiac hyperplasia (51). Abl1 deficits also impair the development and responsiveness of B cells (52, 53) and T cells (54). Ablation of the Abl2 gene has distinct effects, leading only to relatively subtle neuronal defects (27, 55). An Abl1−/− Abl2−/− double mutation, however, causes embryonic lethality with abnormalities in neuroepithelial cells and defects in neurulation (27). Hence, Abl paralogs make unique contributions to vertebrate development while retaining substantial functional overlap.

Biochemistry and Regulation of ABL Tyrosine Kinases

Tyrosine kinase activity

An essential property of ABL proteins is their tyrosine kinase activity. Tyrosine phosphorylation can modify substrate protein activity, localization, and partnering capacity. Consequently, knowledge of ABL catalytic regulation and substrate specificity is integral to understanding ABL function.

Structural data and biochemical studies have revealed multiple autoinhibitory mechanisms that constrain the enzymatic activity of the ABL-family kinases. A myristoyl group attached to the amino terminal glycine of ABL1 and ABL2 “b” isoform proteins can nestle into a surface pocket in the kinase domain, contributing to an autoinhibitory fold (56). A short amino-terminal “Cap” sequence stabilizes the inactive conformation of ABL1 through additional surface interactions (57). Downstream of the “Cap” peptide are SH3 and SH2 domains that cradle the kinase domain, imposing a “locked” inactive state (58, 59). Tyrosine kinase activity of ABL1 is increased by disrupting these autoinhibitory interactions, as demonstrated with the Cap domain mutation in which lysine at position 7 is replaced by alanine (K7A) (56), the SH3 domain mutation W118A (60), the SH2 domain mutations E157A and Y158D (56) and the SH2-TK linker-region double mutation P242E, P249E (also known as “PP”) (60). These key residues are identical (W118, E157, Y158, P242, and P249) or conserved (K7) in the ABL2 protein (Fig. 3).

Phosphatidylinositol 4,5-bisphosphate (PIP2), a phospholipase C–generated second messenger, inhibits ABL1 and ABL2 tyrosine kinase activity (61). PIP2 binds the ABL1 SH2 domain through residues normally required for phosphotyrosine binding (62), but the mechanism of inhibition has not been determined.

Phosphorylations in the amino-terminal half of ABL proteins, including phosphorylation within the kinase domain, can substantially alter catalytic activity (see below). These amino-terminal phosphorylation sites are conserved between ABL1 and ABL2, and phosphorylation at some sites can affect the activity of oncogenic ABL fusion proteins such as BCR-ABL1.

Covalent modification of ABL proteins

Tyrosine phosphorylation of members of the ABL family of kinases occurs in trans by ABL1 and ABL2 (63, 64), a reaction often referred to as “autophosphorylation.” ABL-family kinases are also phosphorylated by members of the SRC family of tyrosine kinases (6365) and by PDGFR (61). Tyrosine phosphorylation of ABL1-Y245 (equivalent to ABL2-Y272), which resides in the linker segment between the SH2 and kinase domains, and ABL1-Y412 (ABL2-Y439), which lies in the activation loop of the kinase domain, correlate with increased kinase activity (63, 64). Phosphorylation of ABL1-Y89 (conserved in ABL2) by members of the SRC family of kinases (66, 67) disrupts SH3 domain–based autoinhibitory interactions and intermolecular associations, such as that with ABI1, and also enhances kinase activity (66). Phosphorylation of ABL2-Y261 (conserved in ABL1) promotes ABL function through protein stabilization (68). Tyrosine phosphorylation at these same sites within the oncogenic fusion protein BCR-ABL1 correlates with its ability to transform cells (67), which demonstrates that “constitutively active” ABL mutants still respond to positive regulation. Phosphorylation of ABL1-Y272 in the kinase domain P loop inhibits ABL1 kinase activity (69) and BCR-ABL1 transforming activity (70), whereas phosphorylation of the nearby ABL1-Y276 enhances transformation and appears to promote kinase activity (70). Phosphosite entries for ABL1 and ABL2 include other phosphorylated tyrosines (Fig. 3). Possible contributions of these modifications to ABL function are not clear, although ABL1-Y158 and ABL1-Y331 contribute to intramolecular folding (56, 71), which may be altered by phosphorylation,

Several ABL phosphotyrosines are predicted SH2 domain–binding sites (http://lilab.uwo.ca/SMALI.htm). Phosphorylated ABL1-Y89 (ABL2-Y116) and ABL1-Y276 (ABL2-Y303) match well with the preferred pYDxV binding site for NCK-family SH2 domains (72) and may contribute to the interaction of ABL1 with NCK1 (73). ABL1-pY134 (ABL2-pY161) and ABL1-pY147 (ABL2-pY174) are predicted to associate with the RASA1 [Ras guanosine triphosphatase (GTPase)–activating protein 1] SH2 domain, consistent with RASA1 binding to BCR-ABL1 (74). ABL1-pY251 is predicted to bind the BLNK (B cell linker protein) SH2 domain, whereas ABL1-pY276 (ABL2-pY303) is predicted to bind the HCK hematopoietic cell kinase SH2 domain, but these interactions have not been experimentally validated.

ABL1 is phosphorylated on T754 (75), which resides at the position normally occupied by phosphoserine in type 1 consensus 14-3-3 binding sites: RSxpSxP (76). Phosphorylation of T754 leads to 14-3-3 binding, which favors the cytoplasmic localization of ABL1 (75). 14-3-3 proteins exist primarily as dimers, which allows them to cross-link multiple sites within a protein or to bridge two target proteins (76), but it is not yet known whether 14-3-3 dimerization contributes to ABL regulation. Phosphorylation of T754 can be carried out by the kinases CLK1, CLK4, MST1, MST2, and TTK in vitro (77) but has not been well studied in vivo. Although ABL2 has a conserved threonine at the same position as ABL1-T754 (Fig. 3), upstream elements of the 14-3-3–binding consensus sequence are absent, and there is no experimental evidence that ABL2 binds 14-3-3 proteins.

Phosphorylation of ABL1 residues S637 and S638 by PAK2 (p21-activated kinase 2) alters the surface charge environment immediately downstream of the third PxxP motif (PTPPKRSS638) and reduces binding to the SH3 domain protein and ABL inhibitor ABI1 (78). This may explain in part the stimulatory effect of PAK2 on ABL kinase activity and the negative feedback resulting from ABL-mediated PAK2 phosphorylation (79). These residues are conserved in ABL2 (Fig. 3), which suggests a similar type of regulation. For reasons less well understood, phosphorylation of these same residues increases ABL1 binding to CRK (78). Additional phosphorylations on ABL1, including that on S588, have been attributed to a CDC2-associated kinase and are coupled to cell division (80).

ABL1 is acetylated at K730 in the second NLS (81). This modification, performed by EP300 (E1A binding protein p300), promotes the cytoplasmic translocation of ABL1.

An amino-terminal myristoyl modification, likely carried out by a myristoyl-CoA:protein N-myristoyltransferase (NMT1 and NMT2 in humans), stabilizes the inactive ABL kinase conformational fold through an intramolecular interaction. Myristoylation of other proteins, including members of the SRC family of tyrosine kinases, has more typically been associated with lipid-membrane interactions that regulate protein localization (82). When the ABL1 myristoyl group is released from the kinase domain by treatment with GNF-2, an allosteric inhibitor compound that binds to the same pocket (83), ABL1 translocates to the endoplasmic reticulum (84). This result raises the possibility that physiological signals may also expose the myristoyl group on ABL proteins and trigger its subcellular relocalization.

Both ABL proteins are subject to polyubiquitination, which leads to their degradation (68, 85). The ubiquitin ligase CBL, itself an ABL substrate (86, 87), has been directly implicated in this modification (83). Some evidence suggests that ubiquitination of ABL2 may itself be regulated by phosphorylation of Y261 in response to oxidative stress (68). ABL1 is also targeted for caspase-mediated cleavage during some types of apoptosis (88).

Substrate specificity

Defining target site specificity is critical for understanding normal and leukemogenic ABL signal transduction pathways. ABL tyrosine phosphorylation can directly influence catalytic activity, SH2 domain binding, and subcellular localization of substrate proteins. Substantial progress has been made in defining ABL target phosphorylation sites, but the ability to accurately predict ABL substrates has remained elusive.

In vitro kinase reactions with peptide libraries suggest a preferred ABL target site: (L- or I- or V-) pYxxP (89, 90). Analysis of 119 separately validated ABL1, ABL2, and BCR-ABL1 substrates (Table 1) confirmed a strong preference for proline at position +3 and for aliphatic amino acids (L, I, or V) at position –1, as well as revealing an enrichment for acidic residues (D or E) at positions –4, –3, and +1 (Fig. 5). Global phosphopeptide analysis of cells expressing BCR-ABL1 is generally consistent with this result (91). These target site preferences presumably reflect structural requirements of the ABL catalytic site, a presumption confirmed, at least in part, by analyses of ABL active site mutations (92).

Table 1

Proteins reported as substrates of ABL1, ABL2, or BCR-ABL1 (nomenclature as in Table 2). “Substrate” shows phosphorylated tyrosine (bold black) and 10 residues flanking each side. Also highlighted: P (+1 to +5, red); D and E (–5 to +5, green) and I, L, and V (–1, blue). No sequence appears if a target has not been mapped. “AA” gives the position of target Y; # indicates targets shown for BCR-ABL1 only; * indicates sites reported for mouse or rat ortholog, but conserved in human. “Ref.” provides literature citations as PMID numbers. DNA repl/rep, DNA replication and repair.

View this table:
Fig. 5

ABL target site consensus. (Top) Sequence logo was created from 119 phosphorylation sites in Table 1 using WebLogo (248) at http://weblogo.berkeley.edu. The line at position zero represents the phosphorylated tyrosine. (Bottom) Amino acid position weight matrix generated using Python. Positions contributing most heavily to the consensus target site, as described in the text, are shown in bold.

There are caveats to identification of phosphorylation targets on the basis of consensus-derived sequences. First, peptide substrates may not accurately mimic protein substrates. Second, protein substrates identified in cells may be indirect targets, in other words, they may be substrates for kinases activated downstream of ABL1, ABL2, or BCR-ABL1. Only a few direct ABL targets have been experimentally determined using mutations that render the kinase dependent on an ATP analog (93) (94).

The SH2 domain contributes to ABL catalytic activity (71) and target site specificity (15). When the ABL SH2 domain is replaced with the SH2 domain from another protein, there is a shift in substrate profile (95). Moreover, SH2 binding preference correlates with target site preference, and juxtaposition of an appropriate SH2 ligand site enhances phosphorylation of an ABL target site. This result suggests a processive phosphorylation mechanism, i.e., successive phosphorylation of different tyrosines on the same substrate (96, 97). Processivity would also explain why ABL-mediated phosphorylation shows accelerated, autocatalytic kinetics when substrate peptides are surface bound at high density (98). In this model, a newly phosphorylated tyrosine moves from the ABL catalytic site to the SH2 pocket, positioning another tyrosine residue for efficient phosphorylation.

Recent work in the field of histone modification has uncovered proteins such as Clr4 that contain both a catalytic site (histone modification “writer”) and a separate binding site (histone modification “reader”) that coordinate processive histone modification (99, 100). By analogy, the combination of a tyrosine kinase domain (pY “writer”) with an SH2 domain (pY “reader”) in ABL and other nonreceptor tyrosine kinases may have evolved in part to increase phosphorylation efficiency of multitarget proteins or complexes. It may also facilitate the phosphorylation of “poor” target sites by docking them into the catalytic pocket. A corollary of the processivity model is that consensus ABL kinase targets based on a database of known phosphorylation sites, may overrepresent amino acids that contribute little to substrate affinity (Km) but facilitate SH2 binding.

Processive phosphorylation is probably integral to ABL function. Multiple target sites exist in many ABL substrates, including BCAR1 (96), CAT (101), CBL (86), DOK1 (102), GAB2 (103), CTTN (104), MDM2 (105), PIK3AP1 (106), PLCG1 (107), PTPN11 (108), PXN (109), POLR2A (RNA polymerase) (110), and RAD51 (111). Indeed, the ABL SH2 domain is essential for efficient phosphorylation of the 52 target sites in the repetitive carboxy-terminal domain of POLR2A (112). In addition, the list of known ABL substrates has many adaptor proteins (including CRK, CRKL, DOK1, GRB2, and NCK1) that, subsequent to phosphorylation, might deliver their associated proteins to the ABL catalytic site after docking at the ABL-SH2 domain.

This simple processive phosphorylation model implies that the ABL catalytic site and SH2 pocket have coevolved to recognize the same sequences (in other words, the best kinase target sequences, once phosphorylated, would also be the best SH2-binding targets). However, some targets within a multitarget substrate appear to require the prior phosphorylation of another target site. RAD51, an established ABL substrate (113, 114) has two ABL target sites: AY54APK (poor consensus) and IY315DSP (good consensus). In transfected cells, a RAD51Y54F mutant was efficiently phosphorylated on Y315, but a RAD51Y315F mutant had no detectable Y54 phosphorylation (115). This suggests that Y315 phosphorylation must occur first, followed by reorientation of the RAD51 substrate so that pY315 is bound to the ABL-SH2 domain and Y54 is positioned at the ABL catalytic site.

The processive phosphorylation data and RAD51 results suggest a “hierarchical processivity” model in which the substrate target site most compatible with ABL kinase domain preferences is phosphorylated with greatest efficiency. If this site is also compatible with the ABL SH2 domain specificity, it will then reposition and dock in the SH2 pocket. This mechanism could enable ABL kinases to phosphorylate otherwise poor targets on the same substrate if they are properly positioned (Fig. 6A). By extension, relatively poor substrate proteins might be recruited to ABL through a complex with strong substrates that can also dock with the SH2 pocket (Fig. 6B). A corollary to the hierarchical processivity model is that consensus ABL phosphorylation sites determined by statistical analysis may actually represent an average of “primary” sites, with high specificity for the catalytic pocket, and “secondary” sites that require the prior SH2 docking of an associated primary site to guide them into the catalytic site.

Fig. 6

Hierarchical processivity model. (A) The catalytically active conformation of ABL1 is depicted with SH3, SH2, and TK domains labeled (carboxy-terminal domains not shown). A “primary,” consensus, tyrosine target (1) is phosphorylated, then relocated to the SH2 domain. This guides a “secondary,” nonconsensus, tyrosine (2) into the catalytic site. (B) Same steps as in (A) except that the secondary tyrosine is in a separate protein associated with the initial substrate.

Target-site phosphorylation of some ABL substrates requires an escort for delivery to the ABL catalytic site, independent of processive phosphorylation. Enah (also known as Mena) phosphorylation at the target site AY296ASA is enhanced by Abi1, which binds to both Enah and Abl1 (116). Similarly, Cables enhances ABL-mediated tyrosine phosphorylation of CDK5 (117) and CTNNB1 (β-catenin) (32) by directly linking kinase and substrate. This use of adaptor proteins to recruit targets further extends the reach of ABL kinases beyond the restrictive preferences of its catalytic site and SH2 domain.

In summary, ABL substrate specificity is driven by both target sequence and domain-guided protein-protein interactions. The role of other determinants is suggested by quantitative shifts in substrate phosphorylation patterns among point mutation variants of BCR-ABL1 (70).

Regulation through tyrosine phosphatases

Given the importance of tyrosine phosphorylation in ABL activation and downstream signaling, the regulatory role of protein tyrosine phosphatases (PTPs) is perhaps not surprising. BCR-ABL1 transforming properties are blocked by PTPN1 (118) and by a variant form of PTPRO (119), which suggests that some PTPs function as negative regulators by reversing ABL-mediated tyrosine phosphorylation. In response to ionizing radiation treatment, ABL1 phosphorylates PTPN6 (also known as SHP-1), which binds to the SH3 domain of ABL1, on Tyr536 and Tyr564 (120). PTPN6 is also associated with BCR-ABL1, and PTPN6 overexpression diminishes some transformation-associated phenotypes of the CML-derived cell line K562 (121). PTPN11 (SHP-2) is an ABL1 substrate that facilitates growth factor–induced mitogenic responses (122). However, delivery of the PTPN11 phosphatase domain to BCR-ABL1 through an ABL-binding domain causes a transformation block (123), which indicates that this phosphatase can also inhibit ABL-mediated tyrosine phosphorylation signaling. PTPN12 (PTP-PEST) and PTPN18 (PTP-HSCF) are recruited by the PSTPIP1 adaptor to dephosphorylated ABL and suppress its function (124).

Interaction Partners That Regulate ABL Activity and Function

ABL self-association

Any discussion of ABL protein partners must consider self-association. Wild-type ABL proteins in cultured cells can exist as oligomers, the formation of which requires their kinase activity and an intact amino-terminal domain (125). Chemically induced dimerization of ABL1 increases both its tyrosine kinase activity and transformation capacity (126), and homophilic interaction domains encoded by translocation partners markedly enhance the catalytic activity and leukemogenic potential of ABL fusion proteins such as BCR-ABL1 and ETV6-ABL1 (discussed below). These data suggest that close proximity of partners within a dimer or oligomer enhances transphosphorylation that directly activates ABL kinases. ABL1-ABL2 heterodimers can form in response to oxidative stress and mediate apoptosis (127), which suggests functional cooperativity between ABL-family members.

Cytoskeleton components

The capacity to directly bind cytoskeletal elements is a defining characteristic of ABL proteins (Fig. 1). A conserved calponin homology (CH)–type F-actin–binding domain is located at the C terminus of ABL proteins, and ABL1 also has G-actin binding properties (128). An (I or L)WEQ (talinlike) domain, found in ABL2 only, provides additional F-actin binding properties (129). These domains mediate the formation of F-actin bundles (128, 129), which are critical for membrane protrusions. Binding to filamentous actin inhibits ABL kinase activity (130), which suggests regulatory feedback. The microtubule-binding domain of ABL2 (131) provides an additional capability for remodeling the cytoskeleton through cross-association with both thick and thin filaments.

A common signal transduction theme among ABL partners

Numerous ABL1 and ABL2 partners have been identified (Table 2). Many of these are ABL substrates, but their association with ABL is stable enough to be detected. The ABL SH2 domain may engage phosphotyrosines on a partner or, conversely, a partner SH2 domain may bind a phosphotyrosine on ABL. Similarly, the ABL SH3 domain engages PxxP motifs on partner proteins, whereas some partner SH3 domains bind to ABL PxxP motifs. Other binding motifs are also employed, and in some cases the mode of interaction is unknown.

Table 2

ABL-interacting proteins. Human protein names (www.genenames.org) are used throughout, although in some cases the murine gene and protein was reported. “Domain” indicates site of interaction on ABL, if known. “Ref.” provides literature citations as PMID numbers. “Sub” identifies proteins that are also known ABL substrates.

View this table:

The ABL interactor proteins ABI1 and ABI2 were initially identified as proteins that bind to PxxP motifs in the carboxy half of ABL1 through their own SH3 domain, while at the same time interacting with the SH3 domain of ABL1 through their own PxxP motifs (132, 133). ABI proteins influence ABL function, although the mechanism for this is not well established. ABI1 can inhibit the transforming activity of v-Abl (133), and ABI1-derived phosphopeptides can inhibit ABL1 activity through an apparent allosteric effect (134). However, ABI1 also enhances the ABL-mediated phosphorylation of several substrates (116, 135, 136) and may facilitate the oligomerization of ABL proteins (125), normally an activating event. ABI1 and ABI2 participate in a WAVE protein complex that promotes actin remodeling at the leading edge of motile cells (137). A Drosophila ortholog, dAbi, plays an opposing role to Abl in axonogenesis and synaptogenesis (38), but appears to increase Abl kinase activity (138). A third mammalian ABI paralog, ABI3, is not well characterized.

The RAS effector protein RIN1 binds directly to ABL1 and ABL2, causing an increase in tyrosine kinase catalytic efficiency (139). The proposed activation mechanism invokes ABL-SH3–domain attachment to a PxxP motif in RIN1. Subsequent ABL kinase domain–mediated phosphorylation of RIN1-Y36 leads to ABL-SH2–domain binding at that site. By disengaging ABL autoinhibitory domains, RIN1 is thought to stabilize a catalytically active ABL conformation. Activation by RIN1 does not require prior phosphorylation of ABL by SRC-family kinases (139), which suggests that this is an independent activation pathway.

Several interacting proteins appear to function as negative regulators of ABL's catalytic activity. PRDX1 (PAG) binds to the ABL1-SH3 domain, inhibits ABL's kinase activity, and overcomes the cytostatic effects of overexpressed ABL1 (140). PSTPIP1 binds to the ABL1-P589xxP and P779xxP motifs and recruits PEST-type PTPs that inhibit ABL's kinase activity (124). ABL inhibition by TUSC2 (FUS1) involves a short (20–amino acid) peptide through an uncharacterized mechanism (141).

Many of the documented ABL-interacting proteins are adaptors that mediate signal transduction and cytoskeleton dynamics. The CRK and CRKL adaptor proteins bind to ABL1 (142144), primarily through the interaction of two SH3 domains on CRK (and on CRKL) with the P545xxP and P589xxP motifs of ABL1 (145), and promote tyrosine kinase activity (146). These interactions serve to dock CRK proteins for subsequent tyrosine phosphorylation (145). Several other multi-SH3 domain adaptor proteins also form stable interactions with ABL1. These include NCK1 (145), SORBS1 (ponsin) (145), SORBS2 (ArgBP1) (145, 147), and SORBS3 (vinexin) (148), all of which engage the P631xxP motif of ABL1 (145). All three PxxP motifs mentioned above are well conserved in ABL2 (Fig. 3), which suggests that it makes similar contacts with CRK, NCK, and SORBS proteins. Other SH3 domain–bearing adaptors implicated as ABL partners are the DOK- family (149, 150), GRB- family (151), and SHC- family (152) proteins. Many of these are mediators of cytoskeleton remodeling.

Cortactin (CTTN), another SH3 domain protein involved in actin dynamics (153), binds to the P573xxP of ABL2 (154). This sequence includes an unusual upstream iteration forming a PxαPxαPxαPxK motif, and mutation of all four prolines is needed to block cortactin binding. The additional upstream prolines are not observed in fish or frog ABL2, nor are they present in any ABL1 sequence (Fig. 4), which suggests that this extended motif evolved relatively late to mediate specialized functions. The fourth PxxP motif (P779xxP) in ABL1 is not conserved in ABL2. Rather, ABL2 encodes a PxxP motif of type 2 (KxαP949xαP) for which no specific binding partners have been identified.

The ABL-SH3 domain has a binding propensity distinct from the type 1 and type 2 categories discussed above, showing a preference for PPx(F,Y,W)xPPP(L,I,V,G,A)P peptides (155). Additional peptide-binding data and modeling were used to develop an ABL-SH3–binding site–motif matrix. WAS-family proteins (WASF1-4), actin reorganization proteins that activate ARP2/3 complexes, were among the strongest ABL-SH3–domain binders predicted with this matrix (156). These results are consistent with experimental evidence that ABL1 binds and signals through WASF proteins (157). Other ABL-SH3 domain–interacting proteins are listed in Table 1.

Both ABL1 and ABL2 have proline-rich sequences, distinct from their SH3-ligand motifs, that conform to established WW domain–ligand motifs of Group 1 (PPxY) and Group 2/3 (PPPPP) (158). The WW-domain protein APBB1 (Fe65) binds to ABL1 (159, 160). From the structure of the APBB1 WW domain (161), it likely binds to PPPPP922. The physiological relevance of this interaction, and potential interactions with APBB2 and APBB3, has not been fully explored.

Cellular Functions of ABL Proteins

Reviews focusing on ABL protein contributions to actin remodeling (162), cell adhesion and motility (9), DNA damage response (163, 164), and microbial pathogen response (165) provide excellent and comprehensive coverage of these areas. Provided below is a brief overview of ABL functions, with an emphasis on ABL partners and substrates.

Actin binding, bundling, and remodeling

ABL-family tyrosine kinases act in the cytoplasm to coordinate actin remodeling in response to appropriate stimuli. This function is mediated by tyrosine phosphorylation of actin cytoskeleton-remodeling proteins and by the ABL carboxy-terminal filamentous actin (F-actin)–binding and –bundling domain (166). ABL-mediated actin remodeling has been studied primarily in the context of cell adhesion and motility (see below), axon guidance (34), and the formation of microspikes (167) and synapses (168170).

The list of known ABL binders (Table 1) and substrates (Table 2) shows that cytoskeleton-remodeling proteins are highly represented. These include F-actin–binding proteins involved in branch formation (WASF and WASL); depolymerization and severing (CFL1); membrane anchoring (ANXA1 and ANXA2); movement (Myosin IIB); and signaling (DBN1, DBNL, CTTN, RAPH1, ENAH, VASP, and EVL). ABL also promotes the association of proteins within this group, as in the case of RAPH1 and ENAH during dorsal ruffling and axonal morphogenesis (171). Microtubule subunits (TUBA and TUBB) are ABL-kinase substrates, although the target sites and effects on polymerization are unknown. The microtubule-binding proteins MAPT (tau) and PXN (paxillin) are also ABL substrates.

Cell motility and adhesion

Mammalian cells mutated in both ABL genes (Abl1−/−Abl2−/−), or wild-type cells treated with an inhibitor of ABL's catalytic activity, show increased motility with altered adhesion and spreading (172174). Conversely, increased ABL activity leads to reduced motility (172, 173). Moreover, ABL proteins localize to sites of actin remodeling (102, 131, 154, 175, 176).

Among direct ABL partners and substrates, the CRK-family (CRK and CRKL) and CAS-family (BCAR1, NEDD9, and EFS) proteins are key regulators of cell attachment and motility (177). ABL-mediated phosphorylation of CRK disrupts its binding to CAS and leads to reduced cell migration (178). The CAS-associated proteins CASS4 and CD2AP are also ABL substrates. Of note is the phosphorylation of multiple YxxP sites on BCAR1, NEDD9, and CASS4 [12, 11, and 8 sites, respectively, in Phosphosite (www.phosphosite.org)].

Interaction of ABL1 with the integrin β2–binding and –activating protein TLN2 (179) provides a direct means for ABL proteins to influence attachment to the extracellular matrix. The role of ABL1 in cell migration involves collaboration with SRC-family kinases (SFKs) (180), which likely involves the phosphoregulation of ABL proteins by SFKs (see Covalent modification of ABL proteins). In a Drosophila epithelial cell invasion model, Abl increased SFK activity through a positive-feedback loop (41), which demonstrates the integral relationship of ABL- and SRC-family kinases.

Receptor endocytosis and autophagy

Actin remodeling has been directly implicated in receptor endocytosis (181183), and ABL proteins likely contribute to this process by coordinating cytoskeleton remodeling with phosphoregulation of receptors and endocytic factors. Multiple receptor tyrosine kinases (EGFR, ERBB2, NTRK1, PDGFRB, ROS, MUSK, and EPHB2) interact with, and are in most cases phosphorylated by, ABL proteins (Tables 1 and 2). ABL interactions promote endocytosis of EGFR (184), facilitate the formation of neuromuscular synapses through MUSK (168), and inhibit PDGFRB-mediated chemotaxis (185). ABL-family tyrosine kinases also modulate the endocytosis of activated B cell receptor complexes (186).

ABL-family tyrosine kinases phosphorylate CAV1 [caveolin, a plasma membrane scaffolding protein that regulates receptor signaling (187)], as well as RIN1 [a RAS effector and RAB5 GEF in endocytosis (188, 189)] and ITSN2 [CDC42 GEF required for caveolae endocytosis (190)]. In addition, ABL proteins functionally engage the CBL family of ubiquitin ligases that drive receptor down-regulation and actin remodeling (191), as well as the CBL-associated protein SORBS1. ABL phosphorylation of CBL leads to increased EGFR stability (184). The membrane invaginations associated with autophagy and the engulfment of apoptotic cells also depend on ABL proteins (45, 192).

The role of ABL1 in DNA damage response and apoptosis

Three NLS motifs (24) and a NES motif (26), together with a regulated 14-3-3 interaction (75), facilitate ABL1 translocation between the cytoplasm and nucleus (193). Its DNA binding domain (194) allows ABL1 to associate directly with DNA in response to damage signals. Many ABL-binding partners and substrates are known mediators of DNA repair (Tables 1 and 2). These include ATM, ATR, DDB1, DDB2, ERCC3, ERCC6, RAD9A, RAD51, RAD52, and WRN. In addition, ABL proteins are found in complexes with MLH1 (195) and other functionally relevant DNA repair proteins. The DNA binding domain and NLS and NES motifs of ABL1 are weak or unrecognizable in fruit fly Abl, sea urchin Abl, and mammalian ABL2 proteins, which suggests that DNA damage-response functions evolved after ABL gene duplication.

ABL1 overexpression causes cell cycle arrest (196) and apoptosis (197) in cultured cells, which suggests that ABL facilitates a repair checkpoint following moderate DNA damage but cell death after severe damage. Several ABL targets (MDM2, MDM4, TP53, and TP73) are primary regulators for this type of damage-induced apoptosis.

Leukemogenic ABL Proteins

Several excellent reviews of ABL oncogenes in hematopoietic malignancies are available (11, 198). This section provides a brief introduction to the topic and then focuses on the current understanding of ABL function in cell transformation and disease.

BCR-ABL1 and related fusion proteins

ABL genes are activated by chromosome translocations in various hematopoietic malignancies. Chronic myeloid leukemia (CML) is characterized in almost all cases by a t(9;22)(q34;q11) translocation (199) that fuses the BCR (breakpoint cluster region) and ABL1 genes (6). Chromatin structural elements (200) and microhomologies and interspersed repeat sequences (201) may contribute to these translocations. The BCR-ABL1 fusion gene product (p210) has constitutive tyrosine kinase activity and is leukemogenic in model systems (11).

Three to 5% of childhood (202) and 20 to 30% of adult acute lymphoblastic leukemia (ALL) cases have a similar translocation (203). The resulting BCR-ABL1 (p190) fusion protein includes less BCR sequence than its CML counterpart. Additional BCR-ABL1 fusion variants have been reported in some leukemias (204, 205). Other leukemogenic ABL1 fusions include NUP214-ABL1 (191) EML1-ABL1 (206), and ETV6-ABL1 (207). In addition, ABL2 is activated by a translocation in some acute myeloid leukemias (AMLs) (208).

Although ABL fusions vary in the amount and identity of upstream partners, and even in which ABL paralog is involved, the ABL breakpoint position consistently leads to removal of the amino-terminal Cap peptide but retention of the SH3 domain (Fig. 7). What remains unclear is why the autoinhibitory SH3 domain is preserved, given that deletion of this domain enhances the catalytic activity of human ABL1. Indeed, the murine Gag-Abl1 (v-Abl) fusion disrupts the SH3 domain (Fig. 7) and is a potent oncogene (209). The presence of autoinhibitory SH3 sequences in human ABL fusion genes might be explained by a requirement for signaling through SH3-interaction partners (Table 1) during leukemia initiation or progression. Notably, murine v-Abl, with a disrupted SH3 domain, produces a phenotypically different disease from that caused by BCR-ABL1 in model systems (210).

Fig. 7

(Top) Seven translocation-derived ABL1 and ABL2 fusion oncoproteins. The following diseases are associated with each fusion: BCR-ABL1 p190 (also called p185) (ALL), p210 (CML), p230 (chronic neutrophilic leukemia or CNL), ETV6-ABL1 (CML and AML), NUP214-ABL1 (T-ALL), EML1-ABL1 (T-ALL), and ETV6-ABL2 (AML). (Bottom) The murine v-Abl oncoprotein results from a retroviral Gag gene fusion that deletes the Abl1 amino terminus and most of the SH3 domain. v-Abl causes B cell lymphomas.

There is a conspicuous absence of ABL-activating mutations in solid tumors, even though ABL fusion genes can transform human fibroblasts in culture (211) and enhanced ABL signaling may contribute to epithelial cell malignancies (212), as well as to the invasive growth of breast cancer cells (213). Furthermore, no other member of the SH3-SH2-TK type nonreceptor tyrosine kinase family is mutationally activated in spontaneous human cancers, even though SRC was first identified as the rodent oncogene in Rous sarcoma virus (214) and cultured human cells are transformed by mutationally activated SRC-family kinases (215) and TEK-family kinases (216). Overexpressed and overactive SRC-family kinases have been detected in many types of tumors [reviewed in (217)], however, and BCR-ABL1 may work in part through SRC activation (218219).

Kinase activity of ABL fusion proteins

Elimination of the ABL myristoylation site and the amino-terminal Cap domain, both of which participate in stabilizing the inactive conformation of the kinase domain, partly explains the increased and constitutive kinase activity of oncogenic ABL fusion proteins. Homophilic interaction domains in BCR and ETV6 mediate fusion protein oligomerization (220, 221), which promotes transphosphorylation and further increases ABL's catalytic activity. Reduced nuclear localization (222) may contribute further to transforming activity by ABL1 fusions by interfering with their DNA damage-response functions.

Consistent retention of the SH3 domain implies that the kinase activity of ABL fusion proteins, although constitutive, can be enhanced through derepression. Indeed, RIN1, which activates ABL1 and ABL2 by binding the SH3 and SH2 domains to relieve autoinhibition, increases the transforming ability of BCR in hematopoietic cell lines and primary bone marrow cells (223) and enhances the leukemogenic properties of BCR-ABL1 in a murine model system (223). Like ABL1 and ABL2, BCR-ABL1 proteins are subject to regulation by phosphatases, including PTPN1 (118) and PTPN6 (224). These observations show that constitutively active ABL oncoproteins remain responsive to positive and negative regulation.

ABL inhibitors in leukemia therapy

Transformation by ABL fusion proteins is inextricably tied to their tyrosine kinase activity, which suggests that targeted kinase inhibitors should be therapeutically useful. Imatinib mesylate (also known as STI571 or Gleevec) is an ATP-competitive inhibitor that stabilizes the inactive ABL kinase–domain conformation (225). Imatinib is an effective first-line treatment for CML (226), which validates the signal pathway blockade approach to cancer treatment, and other BCR-ABL1 inhibitors have been added to the CML pharmacopeia (227).

Some people with CML do not respond to imatinib (228), and even individuals with responsive disease, who must remain on the drug indefinitely, can relapse. Drug resistance is typically a consequence of mutations in the BCR-ABL1 kinase domain (229, 230), but may also result from mutations in the SH3 and SH2 domains (231, 232) or BCR-ABL1 amplification (229). BCR-ABL1–positive ALL is refractory to imatinib alone and shows high rates of resistance and relapse to chemotherapy combined with imatinib (233). Dasatinib, which is among the more effective second-generation ABL active-site inhibitors, also inhibits SRC-family kinases (234), and its efficacy in CML likely reflects the cooperation of SFKs with BCR-ABL1 in eliciting cell transformation. When different ABL inhibitors are used to treat disease sequentially, however, resistance can return through the accumulation of multiple kinase-domain mutations (235), which suggests that a combination approach might be more efficacious. One drug-resistant mutation in the kinase gatekeeper residue (236), BCR-ABL1T315I, is refractory to all established inhibitors but may respond to new drugs (237).

The discovery of allosteric (noncompetitive with ATP) ABL inhibitors (83) represents a promising new direction for the treatment of BCR-ABL1–positive leukemias. These compounds (GNF-2 and GNF-5) target the ABL1 myristate-binding pocket (unoccupied in BCR-ABL1) to stabilize an inactive kinase-domain conformation (238). Combining GNF-5 with catalytic site inhibitors suppresses the emergence of resistance mutations. GNF compounds have limited potency against BCR-ABL1T315I but worked additively with an ATP-competitive inhibitor to block this mutation in a bone marrow–transplantation leukemia model (238). Long-term effective therapy would likely benefit from pairing inhibitors of ABL's kinase activity with drugs that target ABL allosteric sites (238, 239) or drugs that target direct activators of the ABL-family kinases (139). Other combination approaches to therapy of BCR-ABL1–driven leukemias include the pairing of inhibitors of ABL's catalytic activity with drugs that reduce BCR-ABL1 expression (240) or stability (241), or that disable collaborative pathways (242245). Targeting functions required for leukemia cell survival might also provide a useful approach [reviewed in (246)].

Acknowledgments

249.Acknowledgments: The author thanks Steven Goff and Thomas Graeber for comments and criticisms and Jiyong Park for performing the position weight matrix analysis. Funding: I also acknowledge NIH CA136699 as a source of support.

References and Notes

  1. Upper case is used for human genes and proteins (ABL1 and ABL2) as well as for discussion of gene and protein families (ABL). First-letter-only upper case (Abl1 and Abl2) is used when referring to genes and proteins from all other species in this review. Human Genome Organization nomenclature (www.genenames.org) is used. More common gene names are provided in the text, with additional aliases given in Tables 1 and 2.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. Single-letter abbreviations for the amino acid residues used in this review are as follows: A, alanine; C, cysteine; D, aspartic acid; E, glutamic acid; F, phenylalanine; G, glycine; H, histidine; I, isoleucine; K, lysine; L, leucine; M, methionine; N, asparagine; P, proline; Q, glutamine; R, arginine; S, serine; T, threonine; V, valine; W, tryptophan; Y, tyrosine; x, any amino acid; α, hydrophobic amino acid.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
  32. 32.
  33. 33.
  34. 34.
  35. 35.
  36. 36.
  37. 37.
  38. 38.
  39. 39.
  40. 40.
  41. 41.
  42. 42.
  43. 43.
  44. 44.
  45. 45.
  46. 46.
  47. 47.
  48. 48.
  49. 49.
  50. 50.
  51. 51.
  52. 52.
  53. 53.
  54. 54.
  55. 55.
  56. 56.
  57. 57.
  58. 58.
  59. 59.
  60. 60.
  61. 61.
  62. 62.
  63. 63.
  64. 64.
  65. 65.
  66. 66.
  67. 67.
  68. 68.
  69. 69.
  70. 70.
  71. 71.
  72. 72.
  73. 73.
  74. 74.
  75. 75.
  76. 76.
  77. 77.
  78. 78.
  79. 79.
  80. 80.
  81. 81.
  82. 82.
  83. 83.
  84. 84.
  85. 85.
  86. 86.
  87. 87.
  88. 88.
  89. 89.
  90. 90.
  91. 91.
  92. 92.
  93. 93.
  94. 94.
  95. 95.
  96. 96.
  97. 97.
  98. 98.
  99. 99.
  100. 100.
  101. 101.
  102. 102.
  103. 103.
  104. 104.
  105. 105.
  106. 106.
  107. 107.
  108. 108.
  109. 109.
  110. 110.
  111. 111.
  112. 112.
  113. 113.
  114. 114.
  115. 115.
  116. 116.
  117. 117.
  118. 118.
  119. 119.
  120. 120.
  121. 121.
  122. 122.
  123. 123.
  124. 124.
  125. 125.
  126. 126.
  127. 127.
  128. 128.
  129. 129.
  130. 130.
  131. 131.
  132. 132.
  133. 133.
  134. 134.
  135. 135.
  136. 136.
  137. 137.
  138. 138.
  139. 139.
  140. 140.
  141. 141.
  142. 142.
  143. 143.
  144. 144.
  145. 145.
  146. 146.
  147. 147.
  148. 148.
  149. 149.
  150. 150.
  151. 151.
  152. 152.
  153. 153.
  154. 154.
  155. 155.
  156. 156.
  157. 157.
  158. 158.
  159. 159.
  160. 160.
  161. 161.
  162. 162.
  163. 163.
  164. 164.
  165. 165.
  166. 166.
  167. 167.
  168. 168.
  169. 169.
  170. 170.
  171. 171.
  172. 172.
  173. 173.
  174. 174.
  175. 175.
  176. 176.
  177. 177.
  178. 178.
  179. 179.
  180. 180.
  181. 181.
  182. 182.
  183. 183.
  184. 184.
  185. 185.
  186. 186.
  187. 187.
  188. 188.
  189. 189.
  190. 190.
  191. 191.
  192. 192.
  193. 193.
  194. 194.
  195. 195.
  196. 196.
  197. 197.
  198. 198.
  199. 199.
  200. 200.
  201. 201.
  202. 202.
  203. 203.
  204. 204.
  205. 205.
  206. 206.
  207. 207.
  208. 208.
  209. 209.
  210. 210.
  211. 211.
  212. 212.
  213. 213.
  214. 214.
  215. 215.
  216. 216.
  217. 217.
  218. 218.
  219. 219.
  220. 220.
  221. 221.
  222. 222.
  223. 223.
  224. 224.
  225. 225.
  226. 226.
  227. 227.
  228. 228.
  229. 229.
  230. 230.
  231. 231.
  232. 232.
  233. 233.
  234. 234.
  235. 235.
  236. 236.
  237. 237.
  238. 238.
  239. 239.
  240. 240.
  241. 241.
  242. 242.
  243. 243.
  244. 244.
  245. 245.
  246. 246.
  247. 247.
  248. 248.
View Abstract

Stay Connected to Science Signaling

Navigate This Article