This article appeared in a journal published by Elsevier. The attachedcopy is furnished to the author for internal non-commercial researchand education use, including for instruction at the authors institution

and sharing with colleagues.

Other uses, including reproduction and distribution, or selling orlicensing copies, or posting to personal, institutional or third party

websites are prohibited.

In most cases authors are permitted to post their version of thearticle (e.g. in Word or Tex form) to their personal website orinstitutional repository. Authors requiring further information

regarding Elsevier’s archiving and manuscript policies areencouraged to visit:

http://www.elsevier.com/authorsrights

Author's personal copy

The regulation and functions of MAPK pathways in Drosophila

Ben-Zion ShiloDepartment of Molecular Genetics, Weizmann Institute of Science, Rehovot 76100, Israel

a r t i c l e i n f o

Article history:Available online 14 February 2014

Keywords:MAPKERKReceptor tyrosine kinasesEGF receptorFGF receptor

a b s t r a c t

The Mitogen-Activated Protein Kinase (MAPK) pathway represents one of the most conserved signalingcascades in multicellular organisms, since its cytoplasmic components can also be found in single-celledeukaryotic organisms such as yeast. With this broad view in mind, we can ask ourselves not only what arethe seminal features and functions of the pathway in Drosophila, but also what have the studies in Dro-sophila taught us about the pathway in the wide range of organisms where it functions. We discuss thelinearity of the MAPK signaling pathway in developmental decisions, and the ability of the developingorganism to discriminate between different receptor tyrosine kinases that converge on the commonMAPK pathway. Diverse modes of regulating the dynamics and level of MAPK signaling are presented.Finally, the convergence of MAPK signaling with other pathways is reviewed.

� 2014 Elsevier Inc. All rights reserved.

1. Universality and architecture of MAPK pathways

The Mitogen-Activated Protein Kinase (MAPK) pathway repre-sents one of the most conserved signaling cascades in multicellularorganisms, since its cytoplasmic components can also be found insingle-celled eukaryotic organisms such as yeast [1]. In fact, evenin yeast the pathway is involved in sensing and relaying externalinputs including osmolarity stress. With this broad view in mind,we can ask ourselves not only what are the seminal features andfunctions of the pathway in Drosophila, but also what have thestudies in Drosophila taught us about the pathway in the diverserange of organisms where it functions.

In all multicellular organisms the linear order of the pathway isfixed. Ligand binding to receptor tyrosine kinases triggers dimer-ization of the receptors. This process relays external informationto the cytoplasmic side of the receptor [2]. Broadly speaking, thecytoplasmic domain is structurally and functionally separated intothe tyrosine kinase domain, and tyrosine phosphorylation sitesthat will serve as docking anchors. Alteration in the position andconformation of the cytoplasmic domains upon receptor dimeriza-tion leads to trans-phosphorylation of tyrosine residues [3]. Adap-tor proteins such as Grb2 and Shc bind phosphorylated tyrosineson the receptor, and provide a new docking site, usually by virtueof an SH3 domain. Most notable is the Guanine exchange factorSos, which then triggers membrane associated Ras protein. In turn,Ras initiates a cascade of phosphorylation of three kinases, Raf,

MAPKK (MEK) and MAPK (also termed ERK). Raf and MEK have avery restricted substrate specificity that defines the linear arrange-ment of the pathway. In contrast, MAPK has an extremely broadsubstrate specificity, including both nuclear and cytoplasmic pro-teins. It functions as the ‘‘workhorse’’ of the pathway, triggeringin parallel a range of protein activities that are necessary to exe-cute pathway task [4,5] (Fig. 1).

The MAPK pathway relays information following activation ofall receptor tyrosine kinases (RTKs) in Drosophila. There is somevariability in terms of the adaptor proteins used by different RTKs.In some cases (e.g. EGFR or Torso) the adaptor Grb2 is directly asso-ciated with the receptor, while in other cases an intermediateadaptor is used to bind Grb2 (Dof in the case of FGF receptorsand Shc for the insulin receptor). However, once Sos is recruitedby Grb2, all pathways utilize exactly the same downstream cas-cade [2].

There are several implications to this convergence of signalingmodules. First, it means that the biological function of a particularsignaling pathway can best be uncovered genetically by disruptingthe pathway at the level of the ligand or the receptor. Mutations incytoplasmic components may reveal phenotypes reflecting lack ofsignaling by more than one receptor type. Only once the biologicalfunction of a module has been uncovered spatially and temporally,can the downstream components be manipulated in the relevantspace and time, to yield meaningful conclusions. This convergingarchitecture raises the issue of specificity and the capacity of a cellto distinguish which RTK led to the activation of the commonMAPK pathway at that particular point in time.

http://dx.doi.org/10.1016/j.ymeth.2014.01.0201046-2023/� 2014 Elsevier Inc. All rights reserved.

E-mail address: benny.shilo@weizmann.ac.il

Methods 68 (2014) 151–159

Contents lists available at ScienceDirect

Methods

journal homepage: www.elsevier .com/locate /ymeth

Author's personal copy

2. Linearity of developmental RTK pathways

The striking feature regarding signaling triggered by most RTKsin Drosophila is the linearity of the signaling pathway that is em-ployed downstream of the receptor. In contrast, multiple parallelpathways are activated downstream of the activated RTK in verte-brates, including PLCc and PI3K. How do we know that this is notthe case in Drosophila? First, removal of cardinal pathways thatfunction as auxiliary ones in vertebrates, such as PLCc, has no ef-fect on RTK signaling downstream to the receptor. Second, elimina-tion of successive components in the canonical RTK pathway givesrise to identical phenotypes, again strengthening the notion of alinear pathway. A third experimental line of support for the linear-ity of the pathways comes from the expression of consitutively ac-tive elements. Similar phenotypes are again observed,strengthening the notion of pathway linearity.

The linear RTK pathways, including Torso [6], EGFR [7] (Fig. 1),FGFR (Heartless (Htl) and Breathless (Btl)) [8–10], Alk [11] andSevenless [12] are all critical for cell fate determination during

development. Therefore, there may be an operational logic for thisarchitecture. The developmental decisions are irreversible andshould be triggered in an unambiguous fashion. Furthermore, incases where the pathways do not function as an ‘‘On/Off switch’’but rather the actual strength of signaling is informative, the line-arity may facilitate a more direct and less muddled quantitativeoutput. For example, patterning along the leg proximal–distal axisis driven by graded activation of EGFR, which gives rise to differentcell fates [13].

Linearity of RTK signaling is supported by a scaffold protein thatlinks three consecutive elements, thus enhancing the linear relay ofinformation, as well as the speed of signaling. Kinase suppressor ofRas-1 (KSR1) is a protein that displays homology to the kinase do-main of Raf, but does not possess an active kinase domain. Distinctbinding sites in KSR1 for Raf, MEK and MAPK were identified, sup-porting the notion that the three proteins could be simultaneouslybrought together by a scaffolding activity of KSR1. The possibilitythat some of these binding events would be promoted by signalingprovides an additional mode of regulating the pathway. In

ActiveLigand

EGFR

(inactive)(inactive)

Sprouty

Transcription

Kekkon

Argos

Sending cell

Receiving cell

Cytoplasm

Nucleus

RAS

Raf

MEK

ERK

Ksr1

SOS

Grb2

P P

P

YAN

PntP2

argos, sprouty, kekkonpntP1

PYAN

PCic

Cic

P

Fig. 1. Signaling by the EGF receptor pathway. On the cell surface of the signal-receiving cells, the ligands encounter the EGF receptor, which upon dimerization triggers thecanonical Sos/Ras/Raf/MEK/MAPK pathway. Ksr1 functions as a scaffold, to increase the efficiency of signaling. The cardinal transcriptional output of the pathway is mediatedby the ETS protein Pointed (Pnt). One isoform (PntP2) is activated by MAPK phosphorylation, while the second isoform (PntP1) is constitutively active, and its transcription isactivated by MAPK. The MAPK-dependent ETS protein that triggers pntP1 expression (marked by?) may be PntP2 in the context of the eye [83], and is unknown in the settingof the embryo where PntP2 is not expressed in the same tissue. In addition, Yan provides a constitutive repressor which competes for Pnt-binding sites, and can be removedfrom the nucleus and degraded upon phosphorylation by MAPK. Capicua (Cic), an HMG-box protein represents a second repressor that can be removed upon MAPKphosphorylation. Several negative regulators keep the pathway in check. Especially important are inducible components, which constitute a negative-feedback loop. Argos isa secreted molecule which sequesters the ligand Spitz, while Sprouty and Kekkon attenuate signaling within the signal-receiving cell.

152 B.-Z. Shilo / Methods 68 (2014) 151–159

Author's personal copy

accordance with the scaffolding role, increasing levels of KSR1gives rise to a bell shaped effect on signaling levels, since high lev-els of KSR1 act as a dominant-negative by sequestering signalingcomponents instead of combining them in the same signaling com-plex [14]. Interestingly, it was also demonstrated that despite lack-ing an active kinase domain, KSR1 can form side-to-side dimerswith Raf, thereby triggering activation of Raf independent of up-stream signaling [15].

It is interesting to contrast this linearity feature with two Dro-sophila RTK pathways which appear to activate branched signalingcascades, triggered by the insulin receptor or the PDGF/VEGFreceptor (PVR). These pathways play a metabolic role and are thuscontrolling reversible processes. Since they maintain the homeo-stasis of ATP levels, the more elaborate circuitry may facilitate aversatile and adaptable outcome [2].

3. Cellular view of MAPK signaling

The canonical players in the MAPK pathway were uncovered atthe level of the whole organism, by forward genetic screens or byreverse genetics. For many of these genes both maternal and zygo-tic contribution needs to be removed, in order to expose the fullfledged phenotype. Genetic epistasis screens elucidated the orderby which the components function in the pathway. While the ge-netic assays proved to be extremely powerful, a troubling questionloomed regarding the possibility of additional components thatwere not uncovered. Those may represent players that functiononly in specific cell types, or alternatively elements that modulateand fine tune the activity of the pathway, but are not part of thecore machinery.

To uncover such elements, a new approach was required, wherethe output of the pathway would be monitored quantitatively, andeven small changes in the efficiency of signaling could be detected.With the introduction of high throughput technologies to manipu-late and monitor signaling in cultured Drosophila cells, this becamepossible [16].

Initially, quantitative outputs for MAPK signaling were assayedby monitoring the level of dual phosphorylation of MAPK followingstimulation of the EGF or insulin receptors. The availability of acomprehensive library of RNAi molecules covering most if not allthe Drosophila genes, allowed to systematically test their effecton the level of MAPK signaling [17]. In addition to the knowncanonical members of the pathway, tens to hundreds of new com-ponents affecting the level of signaling were identified, with a con-tinuous and broad range of effects. Why are so many new elementspicked up? They may fall into several categories. First, some ofthem may represent components that impinge on the pathwayand modulate it, but are not part of the canonical hardware. Sec-ond, we should note that the developmental signaling pathwaysdo not operate in a vacuum, but rather in the elaborate contextof the cell. It is thus possible that a more general effect on cell bio-logical properties such as trafficking or endocytosis, will also affectthe activity of a given signaling pathway that utilizes this generalmachinery. Finally, since the assays are conducted in cultured cells,it is possible that there may be elements that operate only in thisparticular cell type.

Cell homeostasis in culture, which reflects metabolic steady-state, may be quite distinct from normal developmental signalingin the context of the whole organism. The jury is still out whetherthe large number of modulators identified in cultured cells callsinto question the linearity of developmental RTK signaling path-ways highlighted above. Alternatively, it may represent adjust-ment mechanisms that are exogenous to the linear core signalingpathway.

Following the cell-based screens, the challenge then becomes toidentify significant regulators that were missed by the conven-

tional genetic screens carried out at the level of the whole organ-ism. One example of such a regulator is PLCc (Small wing). Asmentioned earlier, in flies the pathways triggered by RTKs aremostly linear, and PLCc is not an essential component. In fact, PLCcmutant flies are viable, and display small wings and rough eyes[18]. However, an RNAi screen conducted originally in cell culture,has uncovered it as a more subtle regulator of the pathway: A curi-ous feature of EGFR ligand processing in flies that will be discussedlater, is that the cleaved form of the ligand Spitz can be retained inthe ER. In a screen for RNAi molecules that compromise the ERretention of cleaved Spitz, PLCc was picked up. Further biologicalcharacterization has indeed demonstrated that it is required inthe eye in the cells that process the ligand, rather than in the cellsthat receive the signal [19].

To expand the RNAi screens, they were combined with proteo-mic analysis of protein complexes that are generated with the keyelements of the pathway following signaling [20]. While the RNAiscreens detect functionally relevant elements that could act indi-rectly, protein–protein interaction (PPI) networks provide anorthogonal representation of network regulators by revealingphysical associations. The PPI analysis proved to be highly effec-tive, and identified, for example, a direct interaction of MAPK withthe cyclin-dependent kinase cdc2 implying a direct regulation ofthe cell cycle. Most of the novel elements that were uncoveredand are shared between different RTKs represent adaptor proteins.

Interestingly, expression of 25% of the genes identified by PPIwas altered following pathway modulation, implying a high levelof feedback regulation. When examining the global compilationof all the results, only a small fraction of the total number of hitswas isolated independently by the different assays, and those rep-resented the known canonical MAPK pathway components. Thisresult supports the conclusions from the original genetic analysis,that implicated a linear canonical pathway. The additional compo-nents identified can thus be regarded as cell-type specific or RTKspecific modulators, rather than obligatory hubs.

4. Specificity of RTK pathways

Convergence of different RTK pathways on the same linearMAPK cascade raises the question of specificity. Can cells ‘‘tell’’by the mode of MAPK activation which RTK triggered the stimulusand respond accordingly? This issue was tested by replacing thecytoplasmic domain of the FGF receptors Btl and Htl, guiding tra-cheal cell migration and mesodermal spreading, respectively, withthe corresponding domains of Torso and EGFR. In all cases themigration defect was rescued by the hybrid receptors, implyingthe absence of specificity at the level of cytoplasmic signaling. Asexpected, when replacing the cytoplasmic domains, the specificobligatory downstream adaptor for FGF receptors Dof, was no long-er required [21].

The apparent lack of RTK signaling specificity brings forward analternative model, where the timing and context of signaling dic-tates the outcome of cellular responses. An obligatory requirementof this model is that the activation of different RTKs would be seg-regated in time, with no overlap in RTK activation within the samecell. Examination of air sac development during the larval stagesidentified distinct roles for the Btl/FGF-receptor and EGF-receptor,regulating cell migration vs. cell division and maintenance, respec-tively. The analysis of mutant phenotypes is compatible with anearly activity of the FGF receptor, followed by a distinct subsequentphase of EGFR activation. It is possible that the activation by FGFtriggers the expression or processing of ligand for EGFR [22].

The global timing of MAPK activation by RTKs was examined di-rectly when the atlas of MAPK activation was compiled. This wasmade possible by an antibody that specifically recognizes the dual

B.-Z. Shilo / Methods 68 (2014) 151–159 153

Author's personal copy

phosphorylated form of MAPK (dpERK), and can thus provide asnapshot of all activation events at any given point in time[23,24]. Obviously, this general tool displayed complex and dy-namic activation patterns, during embryogenesis and at subse-quent developmental stages. However, since principal RTKs werealready characterized molecularly and genetically at that time, itwas possible to attribute each of the dpERK patterns to the corre-sponding RTK, by examining the missing patterns in mutants for agiven RTK or its ligand.

A sweeping conclusion was that for each of the RTK mutantstested, the corresponding dpERK pattern could be completely elim-inated, even in cases where two RTKs converged on the same tis-sue. For example, EGFR and Btl/FGFR are required for earlytracheal development and migration. However, the EGFR-induceddpERK pattern is observed in the tracheal placodes before invagi-nation, while the FGFR-induced pattern is seen after invaginationand throughout the migration phase [23]. Despite the lack of tem-poral overlap, a recent analysis of the processes that uncovered indetail the steps of tracheal invagination, indicates some functionaloverlap between EGFR and Btl signaling [25] (Fig. 2). Mitoticrounding of two cells at the center of the tracheal placode releasesthe resistance of these cells to centripetal forces. EGFR is activated

by processing of ligand at the center of the placode, and triggersrecruitment and activation of MyosinII in the adjacent epithelialcells, leading to rapid epithelial buckling. Active cell motility, trig-gered by activation of Btl in the tracheal cells by expression ofBranchless (Bnl) in the adjacent non-tracheal cells [26], expandsthe invaginated placode further (Fig. 2).

Once the tracheal cells begin migration, the source of the Bnl li-gand is dynamic [26]. All tracheal cells express the Btl receptor, butthey encounter the ligand in a graded manner, leading to restrictedactivation of MAPK signaling only in the tip cells that are closest tothe ligand source [23] (Fig. 2). Two additional mechanisms thenutilize and maintain the graded activation pattern. High levels ofBtl activation at the tip induce a further induction of Btl expression,generating an effective receptor pool that can trap the ligand, pre-venting it from diffusing to more distant tracheal cells [27]. Inaddition, high and intermediate levels of Btl activation inducethe expression of Sprouty, which in a cell-autonomous mannerattenuates signaling of the pathway, such that only the cells receiv-ing maximal signaling will overcome this repression and attain a‘‘tip cell’’ fate [28,29].

The dpERK atlas displayed visually what was previously impli-cated by genetic studies, regarding the pleiotropy of RTKs. Some

Fig. 2. Successive activation of EGFR and FGFR during tracheal development. (A) EGFR is activated in the cells constituting the tracheal placode before it invaginates, followingproduction of active ligand by Rhomboid expressed by the placode cells, and leading to apical constriction of the cells. The mitotic rounding of the cells at the center of theplacode (marked in green) facilitates invagination, in conjunction with the EGFR-induced constriction. (B) After EGFR-mediated placode invagination is terminated, thesystem is regulated by activation of the FGF receptor Breathless (Btl). All tracheal cells express Btl, and the activating ligand Branchless (Bnl) is produced by defined clusters ofcells outside the placode. The graded presentation and activation of Btl leads to cell migration towards the ligand source. Since the expression of Bnl is dynamic, the trachealcells continue to move towards the source, as it progresses. To maintain the graded activation, expression of Btl is triggered by its activation, generating a sink for the ligand.In parallel, induction of the inhibitory protein Sprouty in a broader domain attenuates signaling, and restricts the highest levels of activation to the leading tracheal cells.Subsequently, only these cells assume the fate of tip cells.

154 B.-Z. Shilo / Methods 68 (2014) 151–159

Author's personal copy

RTKs like Torso, are specialists that act only at a single stage duringdevelopment. Others, operate multiple times in a highly dynamicmanner, EGFR being the champion. The cytoplasmic componentsof the pathway are maternally provided and ubiquitously ex-pressed zygotically, providing the cells with the competence to re-spond to the different RTKs at all times. Some receptors, such asEGFR or Torso are broadly expressed, while others such as Htl/FGFRor Btl/FGFR are restricted to particular tissues. Still as a rule, themajor determinant in the time and place where each pathway isactivated is the expression or availability of active ligands. Delin-eating the dynamic activation of each pathway thus focuses onthe cells that provide the ligand.

Another aspect that was revealed by the dpERK staining relatesto the range of signaling by RTKs at each phase. The observed pro-file represents the final outcome of ligand diffusion and presenta-tion, as well as the combined activity of the different negativefeedback loops that may modulate and stabilize the signaling pro-file. Indeed, differences in the range of signaling were observed fordifferent pathways, or even for the same pathway in different tis-sues. In some cases, prominent signaling was observed several celldiameters away from the ligand source, e.g. activation of EGFR inthe embryonic ventral ectoderm by a ligand that emanates fromthe midline [30]. In other cases such as the wing veins in the imag-inal disc, activation of EGFR is restricted to the vein cells that pro-duce the active ligand [24].

5. Regulating the level of RTK signaling

The key determinant for the time and place of activation of a gi-ven RTK is the expression and presentation of active ligand. Forsome ligands, such as the FGF ligands Thisby, Pyramus [31,32]and Bnl [26], dynamic expression patterns have been documented,while the actual processing and secretion of the ligand does not ap-pear to be limiting. The dynamic expression pattern is especiallystriking in the case of Bnl, since it is stereotypically expressedahead of the migrating tracheal branches and prefigures the struc-ture of the tracheal tree.

The ligand of the Torso RTK, Trunk, is ubiquitously expressed inits precursor form. However, the ligand is processed and presentedto the receptor only at the termini of the embryo. This restrictiondepends on the Torso-like protein, that is specifically expressedduring oogenesis in the terminal follicle cells, leading to the condi-tioning of the vitelline membrane at these domains [33].

In the case of the EGFR ligands, both restricted localization of li-gand expression and tight regulation of ligand processing wereshown to dictate the final pattern of receptor activation. The sym-metry-breaking event in dorso-ventral patterning during oogenesisis the random migration of the oocyte nucleus to an anterior cornerof the egg. The transcripts of the EGFR ligand Gurken, which areproduced in the nurse cells, are trafficked and anchored aroundthe oocyte nucleus [34]. Tight localization of gurken transcriptsleads to local translation of the protein that is secreted out of theegg, generating a dorsally-centered gradient of EGFR activation inthe adjacent follicle cells [35]. Patterning the follicle cells is laterrelayed to the developing embryo.

EGFR has four ligands. Gurken is dedicated to oogenesis, whileSpitz (Spi), Keren and Vein operate at other developmental stages[36–38]. Except for Vein, which is produced as a secreted protein,the other three are produced as membrane precursors that need toundergo cleavage, in order to produce an active ligand [39,40]. Themost extensive work was done on the processing of Spi, which rep-resents the chief EGFR ligand at most stages of development. Com-bined analysis in cell culture and in flies demonstrated that the Spiprecursor is broadly expressed, but is retained in the ER, thusshielding it from the non-specific activity of proteases at the cell

surface. In the ER, Spi associates with the transmembrane proteinStar, which facilitates its trafficking to a late secretory compart-ment by inhibiting retrograde trafficking to the ER [41,42].

Once in the late secretory compartment, Spi precursor encoun-ters Rhomboid, a seven-transmembrane domain intra-membraneprotease. Rhomboid cleaves Spi within its transmembrane domainto release the cleaved form that is biologically active [43] (Fig. 3A).The course of ligand processing relies heavily on intracellular traf-ficking between compartments. Rhomboid and Star are exclusivelydedicated to the processing of ligands in the EGFR pathway, andhence their mutant phenotypes are similar to spi [44,45]. Whilespi and Star are broadly expressed, the expression of rhomboid ishighly dynamic, and was shown to correlate with the sites of MAPKactivation by EGFR [24,46]. Ectopic expression of rhomboid leads tothe corresponding activation of EGFR, implying that it is the onlyrestrictive element [47]. Thus, the spatial and temporal key tothe dynamic activation of EGFR lies in the elaborate promoter/en-hancer structure of rhomboid. The central role of Rhomboid is man-ifested most dramatically in the differentiation of photoreceptorcells in the eye disc, where multiple rounds of EGFR activation leadto successive induction of photoreceptor cell fates [48]. At eachround, the receiving cells where EGFR activation took place, nowinduce the expression of Rhomboid. These cells now become asource for active processing of ligand that will recruit the next co-hort of photoreceptor cells.

It is interesting to note that in some cases the elaborate ligandprocessing machinery is also utilized to modulate the amount of li-gand that will be secreted, and hence the range of EGFR activation.Two additional members of the Rhomboid family, Rhomboid-2 andRhomboid-3 are expressed in the germline and the eye, respec-tively [49]. While possessing the same structure and ligand speci-ficity as Rhomboid-1, their intracellular localization includes notonly the late secretory compartment but also the ER. The abilityof the processing machinery to operate also in the ER has two ma-jor consequences. First, active ligand is generated already in the ER,but this ligand is not utilized since it is retained in the ER by an un-known machinery that also requires PLCc [19]. The ligand precur-sor appears to be in excess so that this process has noconsequences for signaling. The chaperone Star is also a substratefor Rhomboid proteases [50]. When Star is cleaved and inactivatedin the ER, the effective level of chaperone drops, and as a result theamount of ligand that is trafficked to the secretory compartment iscompromised (Fig. 3B). This restriction is significant in the germ-line and eye, where the range of EGFR activation needs to be highlyconfined [51,52].

The process of intramembrane proteolysis entails specific struc-tural and mechanistic constrains in terms of substrate recognitionby Rhomboid proteases, and the capacity to perform the cleavagereaction that relies on hydrolysis within the hydrophobic mem-brane environment. A detailed examination of the process in areconstituted system demonstrated that gating of the potential li-gand into the catalytic site of the Rhomboid protease is the limitingevent, rather than substrate binding. Moreover, catalysis of eachmolecule is a slow process lasting on the order of minutes, provid-ing another level of interrogation of potential substrates and exclu-sion of non-substrates [53]. The functional implications are thatsuch a system can only generate a limited number of active ligandmolecules, even when active at maximal capacity.

Following EGFR activation, three inducible feedback loops stabi-lize the spatial pattern of activation. Argos is a secreted proteinthat is induced by high levels of EGFR activation, and forms aninactive complex with Spi [54–56]. While produced only in cellsreceiving the highest signaling levels, by virtue of its effect onSpi levels and distribution, Argos limits long-range effects byrestricting ligand spread. Two additional proteins operate withinthe cell. Kekkon is a transmembrane protein that forms inactive

B.-Z. Shilo / Methods 68 (2014) 151–159 155

Author's personal copy

complexes with EGFR [57], and Sprouty is a cytoplasmic proteinthat inhibits the canonical RAS/MAPK pathway at several junctions[28,29,36,58]. Since Sprouty affects the canonical pathway, it is uti-lized by several RTKs in Drosophila, most notably EGFR and Btl/FGFR, where it was first discovered. Sprouty is conserved and itsvertebrate homologues play prominent roles in RTK signaling andin cancer.

6. Competition between MAPK substrates

MAPK has a broad specificity of substrates in the nucleus and inthe cytoplasm. Since the levels of activated MAPK may be limiting,the concentration of each substrate should be considered not onlywith respect to its affinity to MAPK, but also regarding its effect onthe capacity of MAPK to phosphorylate other substrates at thesame time. Most MAPK substrates contain, in addition to the phos-phorylation site, a separate docking site that facilitates the associ-ation between the two proteins. Depending on the affinity ofbinding through the docking site, such an interaction may not betransient, thus impinging on the pool of free activated MAPK [4].

A prominent example for the effect of multiple substrates onthe activity of MAPK was described for early patterning of theembryonic termini. The Torso pathway is activated at both polesof the embryo, and the level of dpERK indicates a similar level ofactivation at each pole. On the other hand, when the consequencesof MAPK activation are monitored by the phosphorylation and deg-radation of the ubiquitous transcriptional repressor Capicua, theeffects are much more pronounced at the posterior pole. At thisearly stage of embryogenesis, there are very few asymmetriesalong the anterior-posterior axis, the most prominent being thetranslation of Bicoid (Bcd) at the anterior pole. It turns out thatBcd is a substrate for MAPK, and thus coopts some of the MAPKactivity at the anterior pole [59,60]. The effect of a given substrate

on the availability of active MAPK towards other substrates obvi-ously depends on the relative abundance of that substrate, andthe total number of substrates that are displayed at a given point.

7. How MAPK affects its substrates

Two types of transcriptional outputs were identified for RTKpathways. Most prominent is the ETS-domain protein Pointed(Pnt), which is the principal activator of transcription downstreamof RTKs. Pnt itself is produced in two alternative forms, wherePntP2 represents a typical ETS protein that needs to be activatedby MAPK phosphorylation, while PntP1 is a constitutively-activeform that requires MAPK for its transcriptional induction [61–65].Most tissues monitored use only one of the Pnt forms.

Another avenue by which MAPK impinges on transcription isthrough the phosphorylation and inactivation of transcriptionalrepressors. Yan is an ETS domain protein that lacks a transcriptionactivation domain, and hence blocks the binding of Pointed. Phos-phorylation of Yan by MAPK leads to cytoplasmic export and deg-radation [66–68]. The combined output of Yan and Pnt facilitates arobust bistable switch of target gene expression, triggered byMAPK activation. Indeed, in multiple tissues where strong MAPKactivation was monitored, the presence of Yan and Pnt proteinsis mutually exclusive, displaying a clear complementary pattern[69,70]. There are instances, however, where the presence of bothproteins is found. It is possible that in these instances anotherrepressor of transcription is playing a more central role.

Indeed, a second repressor of MAPK-target genes was identified.Capicua (Cic) is an HMG-box transcriptional repressor. It is notclear if the binding of Cic exerts a local effect on binding of othertranscription factors, or a longer range effect [71–73]. Again, phos-phorylation of Cic by MAPK leads to its inactivation and subse-quently to nuclear export [74]. Detailed kinetic measurements of

Fig. 3. Modulation of EGFR ligand levels by the Rhomboid processing machinery. (A) Trafficking of the ligand precursor mSpi from the endoplasmic reticulum (ER) to thesecretory compartment is facilitated by the chaperone Star. Rhomboid-1 is located only at the secretory compartment, and mediates high signaling levels by cleaving theligand precursor and releasing the active, secreted form (cSpi). (B) Rhomboid2/3 are localized both to the ER and the secretory compartment. Their activity in the twocompartments leads to an attenuated signal, primarily due to cleavage of the ligand chaperone, Star, in the ER. Thus, only the residual Star molecules that escaped cleavage inthe ER can traffic the ligand precursor to the secretory compartment, where cleavage will lead to secretion of active ligand. The cleaved ligand (cSpi) generated in the ER, isretained by a Small wing (Sl) dependent mechanism, and will not be secreted.

156 B.-Z. Shilo / Methods 68 (2014) 151–159

Author's personal copy

functional inactivation of Cic have demonstrated that the proteincan be inactivated rapidly, even before export from the nucleus[75,76].

MAPK also phosphorylates a multitude of cytoplasmic sub-strates in Drosophila and in other organisms [4,77]. A future chal-lenge is to uncover the consequences of these phosphorylationevents. For example, MAPK phosphorylation of the RNA-bindingprotein HOW was shown to facilitate its dimerization and capacityto bind to target RNAs [78]. There are cases where the primary con-sequence of activating the RTK signaling pathway is not transcrip-tion, but cell migration. For example following highly polarizedactivation of the Btl FGF receptor by Bnl during tracheal migration[23,26], or during the migration of the border cells of the ovary thatis mediated by EGFR and PVR [79–82]. This raises the questionwhether guided migration is triggered by MAPK phosphorylationof cytoplasmic substrates, or whether a branch which is indepen-dent of MAPK is the primary effector.

8. Dynamics of MAPK signaling

While the MAPK atlas provides a high spatial resolution of acti-vation, another important aspect is the dynamics of signaling. Ininstances where MAPK activation continues for an extended periodof time, such as the embryonic ventral ectoderm, the signal reachessteady-state and is stabilized by the employment of negative-feed-back loops.

There are, however, cases where MAPK activation is transient.In these cases it is important not only to examine in detail thekinetics of activation, but also to understand how such a transientactivation can lead to sustained and robust signaling outcomes.Further, deciphering which aspects of the transient signaling arecrucial, for instance the ability to reach a critical amplitude, orrather signaling for a minimal time duration, is imperative. In theembryonic ventral ectoderm, induction of rhomboid expressionleads to activation of the EGFR pathway, and eventually to the def-inition of distinct neuronal fates. The ability to morphologically de-fine the stage of the embryo with a resolution of minutes, allowsmonitoring of the kinetics of MAPK activation at an unprecedentedlevel of resolution and examination of the factors that impinge onits dynamic properties [76].

When activation of MAPK is transient, sustaining the responsebecomes a challenge. This issue is especially pertinent in view ofthe fact that MAPK activation not only presses the ‘‘gas’’, but alsorelieves the ‘‘brake’’ by removing proteins such as Yan and Cic. Inthe absence of sustained activation, is the brake perpetually on?This issue was examined for eye development, where the activa-tion of the EGFR pathway at each cycle of photoreceptor recruit-ment is transient and the levels of active ligand are low. It wasshown that the two forms of Pnt are utilized successively. Lowand transient levels of MAPK activation trigger PntP2, which leads(directly or indirectly) to the transcription of pntP1. Since the PntP1protein is consitutively active, the activation of target genes willpersist as long as the protein and its RNA remain [83]. The systemthus switches from a regimen that depends of the kinetics of MAPKphosphorylation and dephosphorylation, to one that relies only onthe stability of the PntP1 protein, allowing transcription to con-tinue even after the EGFR/MAPK pathway was shut off.

Transient activation of the MAPK pathway can also be sustainedin cases where repressors of transcription are inactivated. Groucho(Gro) is a co-repressor that is involved in numerous signaling path-ways and biological processes, including MAPK. When Gro is phos-phorylated by MAPK its repressor activity is compromised [84].Interestingly, this phosphorylation is highly efficient, such thatthe majority of the Gro pool is phosphorylated. By employing anti-bodies that specifically recognize the phosphorylated vs. non-phosphorylated form of Gro, it was shown that the ‘‘memory’’ of

a transient phosphorylation event persists for a long time, thusallowing the expression of target genes that are normally repressedby Gro [85].

The temporal aspects of the response to EGF receptor activationwere recently analyzed by quantitative mass spectroscopy in Rat-2cells stimulated by EGF [86]. At a resolution of minutes, an extre-mely intricate set of responses was identified, mediated by multi-ple phosphorylation waves of the scaffold protein Shc1 thatregulate the temporal flow of signaling information. The first wavetriggers recruitment of the adaptor Grb2, leading to activation ofpro-mitogenic pathways. The second set of Shc1 phosphorylationleads to AKT-mediated negative feedback, by recruitment of aphosphatase. The final wave leads to cytoskeletal reorganizationand signal termination of Shc1. A future challenge is to connectsuch a detailed analysis carried out at the cellular level to an anal-ysis with a similar temporal resolution at the organismal level, andthe inter-cellular communication events it involves.

An experimental system recently developed in mammalian cellculture has demonstrated the ability to trigger reversibly by lightthe translocation of Sos to the membrane, leading to the specificactivation of the Ras/MAPK cascade [87]. The speed and reversibil-ity of this reaction suggest that it would be possible to use optog-enetics in the whole organism to trigger the MAPK pathway, togain a deeper understanding of the dynamic control. For example,one could ask for how long you need to activate the pathway with-in a given cell or tissue in the fly, in order to elicit a response. Itwould also allow to test if the developmental outcome of MAPKactivation depends only on the tissue and time of activation, orwhether the activation dynamics also impinges on the finaloutcome.

9. Cross talk between RTKs and other pathways

The linear nature of RTK signaling raises the issue of conver-gence with other developmental signaling pathways. As a rule,extensive convergence takes place at the promoter/enhancer levelof target genes. These regulatory sequences become hubs for mul-tiple inputs. The RTK-induced outputs include binding of Pnt and/or removal of Cic and Yan. These outputs can trigger a convergenceof transcription factors at the same promoter, for example withtranscription factors activated by other developmental pathways,which contribute to defined temporal responses. In addition, bind-ing of tissue-specific transcriptional activators at the same pro-moter, defines the cell type where transcriptional activationtakes place [88–90]. This combinatorial action provides the activa-tion of gene expression with a high level of specificity, by the con-vergence in time and space of multiple developmental pathwayscoupled to a tissue context. It should be emphasized that havingthe major hub for convergence only at the final step of the pathwayprovides an enormous flexibility, since the architecture of everypromoter/enhancer will essentially define its expression pattern.This regulatory logic allows the adaptation of each promoter tothe relevant signaling pathways that are operating at the relevanttime in the tissue.

The specification of cone cell fate in the developing eye high-lights the spatial and temporal integration of information at thepromoter/enhancer level [91]. All the non-differentiated cells ex-press the Runt-family transcription factor Lozenge. The differenti-ation to a cone cell requires expression of the D-Pax2 gene, andthus its regulatory region comprises the hub that integrates thedistinct signals. It contains binding sites for the ETS transcriptionfactor Pnt activated by EGFR, and for Su(H) activated by Notch[88]. When the EGFR pathway is triggered in an undifferentiatedcell, induction of rhomboid expression converts this cell to a sourceof active and secreted Spi ligand. Activation of EGFR in the adjacent

B.-Z. Shilo / Methods 68 (2014) 151–159 157

Author's personal copy

cell by this ligand also triggers expression of Delta, affecting thenext cell that is now exposed, to ligands that activate both EGFRand Notch pathways. The simultaneous requirement for thesetwo pathways generates a feedforward mechanism, whereby theinitial EGFR activation will induce Delta expression with a delay,but will have to persist after Delta appears, in order to induce D-Pax2 expression. Finally, the requirement for Lozenge binding pro-vides the tissue context, and distinguishes the cells that experi-enced the EGFR and Notch signaling from the other cells in thesame tissue (Fig. 4).

Some cross regulation upstream of the promoter/enhancer levelhas also been identified between RTKs and other developmentalpathways. Gro is a universal repressor of gene expression. It doesnot contain a DNA-binding domain, and hence relies on associationwith other DNA-binding proteins that operate in the context of dif-ferent signaling pathways, such TCF in the Wnt pathway and bHLHproteins impinging on the Notch pathway. Phosphorylation of Groby MAPK was shown to attenuate its activity [84]. In cases whereGro inhibits RTK target genes, this may trigger the RTK transcrip-tional output, e.g. for genes induced by the Torso terminal pathway[85]. In addition, attenuation of Gro activity by MAPK phosphory-lation can impinge on the other signaling pathways that utilizeGro in the same cells.

Acknowledgements

I thank members of my lab and Stas Shvartsman for insightfulcomments. B-Z.S. is an incumbent of the Hilda and Cecil Lewischair in Molecular Genetics.

References

[1] R.E. Chen, J. Thorner, Biochim. Biophys. Acta 1773 (2007) 1311–1340.[2] R. Sopko, N. Perrimon, Cold Spring Harbor Perspect. Biol. 5 (2013). pii:

a009050.[3] A. Arkhipov, Y. Shan, R. Das, N.F. Endres, M.P. Eastwood, D.E. Wemmer, J.

Kuriyan, D.E. Shaw, Cell 152 (2013) 557–569.[4] A.S. Futran, A.J. Link, R. Seger, S.Y. Shvartsman, Curr. Biol. 23 (2013) R972–

R979.[5] I. Wortzel, R. Seger, Genes Cancer 2 (2011) 195–209.[6] M. Furriols, J. Casanova, EMBO J. 22 (2003) 1947–1952.[7] B.Z. Shilo, Development 132 (2005) 4017–4027.[8] M. Beiman, B.Z. Shilo, T. Volk, Genes Dev. 10 (1996) 2993–3002.[9] S. Gisselbrecht, J.B. Skeath, C.Q. Doe, A.M. Michelson, Genes Dev. 10 (1996)

3003–3017.[10] C. Klambt, L. Glazer, B.Z. Shilo, Genes Dev. 6 (1992) 1668–1678.[11] D. Popichenko, F. Hugosson, C. Sjogren, M. Dogru, Y. Yamazaki, G. Wolfstetter,

C. Schonherr, M. Fallah, B. Hallberg, H. Nguyen, R.H. Palmer, Development 140(2013) 3156–3166.

[12] T. Raabe, Biochim. Biophys. Acta 1496 (2000) 151–163.[13] M.I. Galindo, S.A. Bishop, S. Greig, J.P. Couso, Science 297 (2002) 256–259.

Fig. 4. Cooperation between EGFR and Notch pathways in the developing compound eye. Once a photoreceptor cell fate is induced, the cell expresses Rhomboid and becomesa source for the active EGFR ligand. Activation of EGFR in the neighboring cell induces photoreceptor cell fate and triggers expression of Delta. The next undifferentiated cellencounters activation of the Notch pathway by Delta, as well as triggering of the EGFR pathway. These cells also express Lozenge (Lz). Combinatorial integration of the threeinputs at the regulatory region of the Pax2 gene induces its transcription and differentiation of the cell into a cone cell.

158 B.-Z. Shilo / Methods 68 (2014) 151–159

Author's personal copy

[14] W. Kolch, Nat. Rev. Mol. Cell. Biol. 6 (2005) 827–837.[15] T. Rajakulendran, M. Sahmi, M. Lefrancois, F. Sicheri, M. Therrien, Nature 461

(2009) 542–545.[16] A. Friedman, N. Perrimon, Cell 128 (2007) 225–231.[17] A. Friedman, N. Perrimon, Nature 444 (2006) 230–234.[18] J.R. Thackeray, P.C. Gaines, P. Ebert, J.R. Carlson, Development 125 (1998)

5033–5042.[19] A. Schlesinger, A. Kiger, N. Perrimon, B.Z. Shilo, Dev. Cell 7 (2004) 535–545.[20] A.A. Friedman, G. Tucker, R. Singh, D. Yan, A. Vinayagam, Y. Hu, R. Binari, P.

Hong, X. Sun, M. Porto, S. Pacifico, T. Murali, R.L. Finley Jr., J.M. Asara, B. Berger,N. Perrimon, Sci. Signal. 4 (2011) rs10.

[21] C. Dossenbach, S. Rock, M. Affolter, Development 128 (2001) 4563–4572.[22] C. Cabernard, M. Affolter, Dev. Cell 9 (2005) 831–842.[23] L. Gabay, R. Seger, B.Z. Shilo, Development 124 (1997) 3535–3541.[24] L. Gabay, R. Seger, B.Z. Shilo, Science 277 (1997) 1103–1106.[25] T. Kondo, S. Hayashi, Nature 494 (2013) 125–129.[26] D. Sutherland, C. Samakovlis, M.A. Krasnow, Cell 87 (1996) 1091–1101.[27] T. Ohshiro, K. Saigo, Development 124 (1997) 3975–3986.[28] T. Casci, J. Vinos, M. Freeman, Cell 96 (1999) 655–665.[29] N. Hacohen, S. Kramer, D. Sutherland, Y. Hiromi, M.A. Krasnow, Cell 92 (1998)

253–263.[30] M. Golembo, E. Raz, B.Z. Shilo, Development 122 (1996) 3363–3370.[31] T. Gryzik, H.A. Muller, Curr. Biol. 14 (2004) 659–667.[32] A. Stathopoulos, B. Tam, M. Ronshaugen, M. Frasch, M. Levine, Genes Dev. 18

(2004) 687–699.[33] M. Furriols, G. Ventura, J. Casanova, Proc. Natl. Acad. Sci. U.S.A. 104 (2007)

11660–11665.[34] F.S. Neuman-Silberberg, T. Schupbach, Cell 75 (1993) 165–174.[35] F.S. Neuman-Silberberg, T. Schupbach, Mech. Dev. 59 (1996) 105–113.[36] A. Reich, B.Z. Shilo, EMBO J. 21 (2002) 4287–4296.[37] B.J. Rutledge, K. Zhang, E. Bier, Y.N. Jan, N. Perrimon, Genes Dev. 6 (1992)

1503–1517.[38] B. Schnepp, G. Grumbling, T. Donaldson, A. Simcox, Genes Dev. 10 (1996)

2302–2313.[39] C. Ghiglione, E.A. Bach, Y. Paraiso, K.L. Carraway 3rd, S. Noselli, N. Perrimon,

Development 129 (2002) 175–186.[40] R. Schweitzer, M. Shaharabany, R. Seger, B.Z. Shilo, Genes Dev. 9 (1995) 1518–

1529.[41] J.R. Lee, S. Urban, C.F. Garvey, M. Freeman, Cell 107 (2001) 161–171.[42] R. Tsruya, A. Schlesinger, A. Reich, L. Gabay, A. Sapir, B.Z. Shilo, Genes Dev. 16

(2002) 222–234.[43] S. Urban, J.R. Lee, M. Freeman, Cell 107 (2001) 173–182.[44] C. Klambt, J.R. Jacobs, C.S. Goodman, Cell 64 (1991) 801–815.[45] U. Mayer, C. Nusslein-Volhard, Genes Dev. 2 (1988) 1496–1511.[46] E. Bier, L.Y. Jan, Y.N. Jan, Genes Dev. 4 (1990) 190–203.[47] M.A. Sturtevant, E. Bier, Development 121 (1995) 785–801.[48] M. Freeman, Cell 87 (1996) 651–660.[49] J.D. Wasserman, S. Urban, M. Freeman, Genes Dev. 14 (2000) 1651–1663.[50] R. Tsruya, A. Wojtalla, S. Carmon, S. Yogev, A. Reich, E. Bibi, G. Merdes,

E. Schejter, B.Z. Shilo, EMBO J. 26 (2007) 1211–1220.[51] S. Yogev, E.D. Schejter, B.Z. Shilo, EMBO J. 27 (2008) 1219–1230.[52] S. Yogev, E.D. Schejter, B.Z. Shilo, PLoS Biol. 8 (2010).[53] S.W. Dickey, R.P. Baker, S. Cho, S. Urban, Cell 155 (2013) 1270–1281.[54] M. Golembo, R. Schweitzer, M. Freeman, B.Z. Shilo, Development 122 (1996)

223–230.[55] D.E. Klein, V.M. Nappi, G.T. Reeves, S.Y. Shvartsman, M.A. Lemmon, Nature 430

(2004) 1040–1044.[56] R. Schweitzer, R. Howes, R. Smith, B.Z. Shilo, M. Freeman, Nature 376 (1995)

699–702.

[57] C. Ghiglione, K.L. Carraway 3rd, L.T. Amundadottir, R.E. Boswell, N. Perrimon,J.B. Duffy, Cell 96 (1999) 847–856.

[58] S. Kramer, M. Okabe, N. Hacohen, M.A. Krasnow, Y. Hiromi, Development 126(1999) 2515–2525.

[59] Y. Kim, M.J. Andreu, B. Lim, K. Chung, M. Terayama, G. Jimenez, C.A. Berg, H. Lu,S.Y. Shvartsman, Dev. Cell 20 (2011) 880–887.

[60] Y. Kim, M. Coppey, R. Grossman, L. Ajuria, G. Jimenez, Z. Paroush,S.Y. Shvartsman, Curr. Biol. 20 (2011) 446–451.

[61] C. Klambt, Development 117 (1993) 163–176.[62] H. Scholz, J. Deatrick, A. Klaes, C. Klambt, Genetics 135 (1993) 455–468.[63] D. Brunner, K. Ducker, N. Oellers, E. Hafen, H. Scholz, C. Klambt, Nature 370

(1994) 386–389.[64] E.M. O’Neill, I. Rebay, R. Tjian, G.M. Rubin, Cell 78 (1994) 137–147.[65] L. Gabay, H. Scholz, M. Golembo, A. Klaes, B.Z. Shilo, C. Klambt, Development

122 (1996) 3355–3362.[66] Z.C. Lai, G.M. Rubin, Cell 70 (1992) 609–620.[67] I. Rebay, G.M. Rubin, Cell 81 (1995) 857–866.[68] T.L. Tootle, P.S. Lee, I. Rebay, Development 130 (2003) 845–857.[69] J.F. Lachance, N. Pelaez, J.J. Cassidy, J.L. Webber, I. Rebay, R.W. Carthew, Dev.

Biol 385 (2013) 263–278.[70] G.J. Melen, S. Levy, N. Barkai, B.Z. Shilo, Mol. Syst. Biol. 1 (2005) (2005) 0028.[71] E. Cinnamon, D. Gur-Wahnon, A. Helman, D. St Johnston, G. Jimenez,

Z. Paroush, EMBO J. 23 (2004) 4571–4582.[72] G. Jimenez, A. Guichet, A. Ephrussi, J. Casanova, Genes Dev. 14 (2000) 224–231.[73] F. Roch, G. Jimenez, J. Casanova, Development 129 (2002) 993–1002.[74] S. Astigarraga, R. Grossman, J. Diaz-Delfin, C. Caelles, Z. Paroush, G. Jimenez,

EMBO J. 26 (2007) 668–677.[75] O. Grimm, V. Sanchez Zini, Y. Kim, J. Casanova, S.Y. Shvartsman, E. Wieschaus,

Development 139 (2012) 3962–3968.[76] B. Lim, N. Samper, H. Lu, C. Rushlow, G. Jimenez, S.Y. Shvartsman, Proc. Natl.

Acad. Sci. U.S.A. 110 (2013) 10330–10335.[77] S.M. Carlson, C.R. Chouinard, A. Labadorf, C.J. Lam, K. Schmelzle, E. Fraenkel,

F.M. White, Sci. Signal. 4 (2011) rs11.[78] R. Nir, R. Grossman, Z. Paroush, T. Volk, PLoS Genet. 8 (2012) e1002632.[79] P. Duchek, P. Rorth, Science 291 (2001) 131–133.[80] P. Duchek, K. Somogyi, G. Jekely, S. Beccari, P. Rorth, Cell 107 (2001) 17–26.[81] J.A. McDonald, E.M. Pinheiro, L. Kadlec, T. Schupbach, D.J. Montell, Dev. Biol.

296 (2006) 94–103.[82] J.A. McDonald, E.M. Pinheiro, D.J. Montell, Development 130 (2003) 3469–

3478.[83] A. Shwartz, S. Yogev, E.D. Schejter, B.Z. Shilo, Development 140 (2013) 2746–

2754.[84] P. Hasson, N. Egoz, C. Winkler, G. Volohonsky, S. Jia, T. Dinur, T. Volk,

A.J. Courey, Z. Paroush, Nat. Genet. 37 (2005) 101–105.[85] A. Helman, E. Cinnamon, S. Mezuman, Z. Hayouka, T. Von Ohlen, A. Orian,

G. Jimenez, Z. Paroush, Curr. Biol. 21 (2011) 1102–1110.[86] Y. Zheng, C. Zhang, D.R. Croucher, M.A. Soliman, N. St-Denis, A. Pasculescu,

L. Taylor, S.A. Tate, W.R. Hardy, K. Colwill, A.Y. Dai, R. Bagshaw, J.W. Dennis,A.C. Gingras, R.J. Daly, T. Pawson, Nature 499 (2013) 166–171.

[87] J.E. Toettcher, O.D. Weiner, W.A. Lim, Cell 155 (2013) 1422–1434.[88] G.V. Flores, H. Duan, H. Yan, R. Nagaraj, W. Fu, Y. Zou, M. Noll, U. Banerjee, Cell

103 (2000) 75–85.[89] M.S. Halfon, A. Carmena, S. Gisselbrecht, C.M. Sackerson, F. Jimenez, M.K.

Baylies, A.M. Michelson, Cell 103 (2000) 63–74.[90] C. Xu, R.C. Kauffmann, J. Zhang, S. Kladny, R.W. Carthew, Cell 103 (2000) 87–

97.[91] L. Tsuda, R. Nagaraj, S.L. Zipursky, U. Banerjee, Cell 110 (2002) 625–637.

B.-Z. Shilo / Methods 68 (2014) 151–159 159