quantitative approaches to understanding signaling ...-3 -2.5 -2 -1.5 -1 -0.5 0 0.5 py reduction w/...
TRANSCRIPT
Quantitative approaches to understanding signaling regulation of 3D cell migration
April 8, 2014
Metastasis is a multistep process
Nature Reviews | Cancer
Primary tumour Vascularization Detachment
Circulatingtumour cell
Adhesion toblood vessel wall Extravasation
Growth of secondary tumour
Intravasation
In the series of steps that comprise the metastatic process, cancer cells migrate or flow through vastly different microenvironments, including the stroma, the blood vessel endothelium, the vascular system and the tissue at a secondary site1,2 (FIG. 1). The ability to successfully negotiate each of these steps and advance towards the formation and growth of a secondary tumour is dependent, in part, on the physical interactions and mechanical forces between cancer cells and the microenvironment. For example, the physical interactions between a cell and the extracellular matrix — the collagen-rich scaffold on which it grows — have a key role in allowing cells to migrate from a tumour to nearby blood vessels. During intravasation and extravasation, cells must undergo large elastic deformations to penetrate endothelial cell–cell junctions. In the vascular system, the interplay between cell velocity and adhesion influences the binding of cancer cells to blood vessel walls and hence the location of sites where a secondary tumour can form and grow. A clearer understanding of the role of physical interactions and mechanical forces, and their interplay with biochemical changes, will provide new and important insights into the progression of cancer and may provide the basis for new therapeutic approaches.
Physical interactions in invasionFollowing the growth of a primary tumour, the combination of continued tumour pro-liferation, angiogenesis, accumulated genetic transformations and activation of complex signalling pathways trigger the metastatic
cascade (FIG. 2). In particular, the detachment of carcinoma cells from the epithelium and the subsequent invasion of the underlying stroma resembles, at both the cellular and molecular levels, the well-characterized epithelial-to-mesenchymal transition (EMT) in embryogenesis3. The role of EMT in cancer metastasis is being actively explored4,5. Critical to EMT is the loss of E-cadherin (an intercellular adhesion molecule) and cytokeratins, which leads to dramatic changes in the physical and mechanical properties of cells: specifically, reduced intercellular adhesion and a morphological change from cuboidal epithelial to mesenchymal6. One consequence of these changes is detachment from the primary tumour and the acquisition of a motile phenotype5. These cells also begin to express matrix metalloproteinases
(MMPs) on their surface, which promote the digestion of the laminin- and collagen IV-rich basement membrane7. After leaving the tumour microenvironment, motile tumour cells encounter the architecturally complex extracellular matrix (ECM), which is rich in collagen I and fibronectin8 (BOX 1). In the vicinity of a mammary tumour, the matrix is often stiffer than in normal tissue owing to enhanced collagen deposition9 and lysyl-oxidase-mediated crosslinking of the collagen fibres by tumour-associated fibroblasts10. Collagen crosslinking enhances integrin signalling as well as the bundling of individual fibres11. Such changes in the physicochemical properties of the matrix can enhance cell proliferation and invasion in a positive feedback loop9. Whether stiffening of the stromal matrix occurs in other solid tumours, besides mammary tumours, remains to be determined. However, despite recent technological advances (TABLE 1), remarkably little is known about the molecular and physical mechanisms that drive motile cancer cells away from primary tumour and into the stromal space, especially at the subcellular level.
Motility in three dimensions. Much of what we have learned about the physical and molecular mechanisms driving normal and cancer cell motility has come from in vitro studies using two-dimensional (2D) substrates12–14. However, the dimensionality of the system used to study cancer invasion can have a key role in dictating the mode of cell migration. This is not entirely surprising as the three-dimensional (3D) microenvironment of the ECM in vivo is characterized by many features, including the pore size and fibre orientation, features that are not found in conventional ECM-coated
O P I N I O N
The physics of cancer: the role of physical interactions and mechanical forces in metastasisDenis Wirtz, Konstantinos Konstantopoulos and Peter C. Searson
Abstract | Metastasis is a complex, multistep process responsible for >90% of cancer-related deaths. In addition to genetic and external environmental factors, the physical interactions of cancer cells with their microenvironment, as well as their modulation by mechanical forces, are key determinants of the metastatic process. We reconstruct the metastatic process and describe the importance of key physical and mechanical processes at each step of the cascade. The emerging insight into these physical interactions may help to solve some long-standing questions in disease progression and may lead to new approaches to developing cancer diagnostics and therapies.
Figure 1 | The metastatic process. In this complex process, cells detach from a primary, vascularized tumour, penetrate the surrounding tissue, enter nearby blood vessels (intravasation) and circulate in the vascular system. Some of these cells eventually adhere to blood vessel walls and are able to extrav-asate and migrate into the local tissue, where they can form a secondary tumour.
PERSPECT IVES
512 | JULY 2011 | VOLUME 11 www.nature.com/reviews/cancer
© 2011 Macmillan Publishers Limited. All rights reserved
Wirtz et al, Nat Rev Cancer, 2011
Invasive programs globally rewire cell signaling
EMT. Opposing this function, multiple pro-angiogenic
proteins were increased, notably angiopoietin-1 (ANGPT1),co-expression of thrombospondins together with the
ADAMTS1 protease [86] and VEGF-C.
Multiple receptor protein kinase networks were acti-vated in the mesenchymal state. Autocrine Gas6 stimula-
tion of Axl and Tyro3 activity were established from array
and RT-PCR RNA abundance measurements. The datasuggest inhibitors of Axl and Tyro3 activity may show
utility in limiting the survival of shed mesenchymal-like
NSCLC cells. A functional role of autocrine Axl-Gas6expression in lung [87] and breast cancer [88] progression
has been established. Autocrine activation of FGFR and
PDGFR signaling also was observed in both established(H1703) and metastable TGFb models, supporting previous
findings [25, 89]. This increase in kinase activity and
PDGFR autophosphorylation was associated withincreased PDGFR and PDGF RNA transcripts. In meta-
stable Snail and Zeb1 models, FGFR1 and PDGFR tran-
scripts were markedly increased, PDGF was modestlyincreased and FGF1 was unchanged or slightly decreased
(Supplementary Table S5). These data support the predic-
tion of kinase activation by autocrine co-expression ofreceptor–ligand transcripts. We asked what functional
EMT dependent acquisition of ITGα5β1/FN and TGFβ signaling, activation of MAPK1/Erk2 and PTK2/FAK, inactivation of GSK3
1 100 1 100
H358
Erlotinib (uM)
pErk
Erk
pAkt(473)
Akt
β-Actin
H358(TGF β)B
Glypican1/6FGFR1
FGF1/2
p
AXLTYRO3
ADAMTS1
CTGF
IKIP
TGFβR1/2BMPR2
TGFB1/2/3
LTBP2
FAM129BS628,S679,
S683
BRAF p
HDGF
RAS
FAM129A
p
pPTK2pRBM35A/B
GAS6
FGFBP1FGF1/2
ITGα5β1
FN
MARCKS
ANXA6
PEA15
PFTK1TGFBI
PXNS84/S85
T284/S288Y118
MAPK1Y187
SRCp
MST4
CAV1p
DUSP4/6G3BP1
G3BP2
ENG
CCDC80
HDGF
P2RY5
PTPN14
MAPK14p
PKN1GADD45
SGK1 GSK3InactiveS278
AKT1AKTIP
AKT3DDIT4
DIDO1
INHBB
THBS1/2
p
PDGFRA
PDGF A/B/C
STAT3
PTPN11
p
CAV1pINA p
PIK3R1
PIK3CB
p
pp
PLCG1p
ANGPT1
IL8IL18
IL7CCL26
CGB
SCG5
VEGFC
IL11
IL6CLCF1
IL6ST
IL6RCNTRF JAK2
CRLF1
CXCR7
SPARC
DISP2
GIT2
GIT2
ANXA6
MAP3K3
pSTAT3
ISL1
PTX3 YARS
Fig. 5 a Gain of new survivaland cytokine networks.Increased signaling throughIL-11/IL-6/gp130, TGFb/BMP2, Axl/Tyro3, integrina5b1, PDGFR and FGFR wereobserved. Downstreamactivation of BRAF/MAPK1and PTK2/paxillin signalingwere observed. Components ingreen were attenuated duringEMT while those in lightbrown/red were increased withEMT. b Activating MAPK1/Erkphosphorylation (T185/Y187),associated with TGFb-inducedEMT, was confirmed byimmunoblot
Clin Exp Metastasis
123
co-attenuated with the EMT-mediated reduction in Met,
Ron, EGFR and IGF1R kinase activation. These included
multiple SH2, cytoplasmic kinase and ubiquinylationadapters: pSHC, pGrb2, pGAB1, pPTK6 (Brk), pUBB,
pUBIQ and pNCKAP1. We asked whether attenuated Met
phosphorylation might be attributable to reduced HGF
binding to Met. While no change in HGF transcript abun-
dance was observed, ST14 (matriptase), an extracellularprotease which promotes proHGF activation, was markedly
reduced. SPINT2 and SPINT1, which augment ST14
EGFRY1110Y1172
EGFRERBB2ERBB3
p
METMST1R/RON
AREG
EREG
TGFA
SHC1GRB2GRB2 pp p
p
PTK6p
GAB1p
Plasma membrane
BTCIGF1R
p
NCKAP1p
UBBp
ANXA9
SHC1
UBIQp
PDCD6
PDCD6IP
p
p
SHC1p
p
SFN p
ITGα6β4
ST14HGF
SPINT1SPINT2
PRSS8
MUC1/4
GPR110
PTAFR
H19
LRIG3
HBEGF
NRG
GRB2/7p
CDC2T14/Y15
MCM2S139
CDC2CDCA7L
PCNA
SFN
CDKN2D
TMPRSS4
CCND1CCNB1
IGFBP2
IGFBP4
IGFBP7
STK39
STAP2
S100P
S100A2EZR
p
S100A14
S100A6
S100A4
TP53
S100A11
ARHGEF16
IL4R
MAP3K8
PA2G4p
PTPRK
PTPRG
macrophage
EGF family
CSF1
NKD2
DDIT4
DKK1
CTNNB1
LRP5
APOE
CTGF
TSC2
RTK tyrosine phosphorylation measured by elisa array, comparing ratios of mesenchymal-like (Mes) and epithelial (Epi) NSCLC lines: all changes p<0.01 except ErbB4 (ns) where error bars reflect the standard error of the mean
Pho
spho
tyro
sine
Con
tent Mesenchymal
Epithelial
25 Calu6
H1703
H292
H358
Calu6
H1703
H292
H358
Calu6
H1703
H292
H358
CB
L/CB
LBH
SP
D1
SH
C1/S
HC
2
-3 -2.5 -2 -1.5 -1 -0.5 0 0.5
pY reduction w/ EGFR TKI
1 hr
4 hrs
24hrs
A
B C
Fig. 4 a Loss of autocrine RTK signaling networks through EGFR/ErbB2/ErbB3, Met/Ron and IGF1R signaling networks. AttenuatedIL4R, Muc1 and Muc4 and integrin a6b4 components also wereobserved. b Attenuated RTK tyrosine phosphorylation in H292 andH358, (epithelial phenotype) Calu6 and H1703 (mesenchymalphenotype) cell models were confirmed by quantitation of RTKarrays. c EGFR is functional and responds to exogenous ligand in
both epithelial and mesenchymal states. H292 and H358, (epithelial)Calu6 and H1703 (mesenchymal) cells were exposed to EGFR kinaseinhibitor (3 lM erlotinib, 2 h) or DMSO control prior to stimulationwith exogenous EGF (10 ng/ml, 10 min). The EGFR-dependentinhibition of substrates CBL and SHC were similar in epithelial andmesenchymal-like cells, relative for control HSPD1
Clin Exp Metastasis
123
Thomson et al, Clin Exp Metastasis, 2010
Nature Reviews | Cancer
Primary tumour Vascularization Detachment
Circulatingtumour cell
Adhesion toblood vessel wall Extravasation
Growth of secondary tumour
Intravasation
In the series of steps that comprise the metastatic process, cancer cells migrate or flow through vastly different microenvironments, including the stroma, the blood vessel endothelium, the vascular system and the tissue at a secondary site1,2 (FIG. 1). The ability to successfully negotiate each of these steps and advance towards the formation and growth of a secondary tumour is dependent, in part, on the physical interactions and mechanical forces between cancer cells and the microenvironment. For example, the physical interactions between a cell and the extracellular matrix — the collagen-rich scaffold on which it grows — have a key role in allowing cells to migrate from a tumour to nearby blood vessels. During intravasation and extravasation, cells must undergo large elastic deformations to penetrate endothelial cell–cell junctions. In the vascular system, the interplay between cell velocity and adhesion influences the binding of cancer cells to blood vessel walls and hence the location of sites where a secondary tumour can form and grow. A clearer understanding of the role of physical interactions and mechanical forces, and their interplay with biochemical changes, will provide new and important insights into the progression of cancer and may provide the basis for new therapeutic approaches.
Physical interactions in invasionFollowing the growth of a primary tumour, the combination of continued tumour pro-liferation, angiogenesis, accumulated genetic transformations and activation of complex signalling pathways trigger the metastatic
cascade (FIG. 2). In particular, the detachment of carcinoma cells from the epithelium and the subsequent invasion of the underlying stroma resembles, at both the cellular and molecular levels, the well-characterized epithelial-to-mesenchymal transition (EMT) in embryogenesis3. The role of EMT in cancer metastasis is being actively explored4,5. Critical to EMT is the loss of E-cadherin (an intercellular adhesion molecule) and cytokeratins, which leads to dramatic changes in the physical and mechanical properties of cells: specifically, reduced intercellular adhesion and a morphological change from cuboidal epithelial to mesenchymal6. One consequence of these changes is detachment from the primary tumour and the acquisition of a motile phenotype5. These cells also begin to express matrix metalloproteinases
(MMPs) on their surface, which promote the digestion of the laminin- and collagen IV-rich basement membrane7. After leaving the tumour microenvironment, motile tumour cells encounter the architecturally complex extracellular matrix (ECM), which is rich in collagen I and fibronectin8 (BOX 1). In the vicinity of a mammary tumour, the matrix is often stiffer than in normal tissue owing to enhanced collagen deposition9 and lysyl-oxidase-mediated crosslinking of the collagen fibres by tumour-associated fibroblasts10. Collagen crosslinking enhances integrin signalling as well as the bundling of individual fibres11. Such changes in the physicochemical properties of the matrix can enhance cell proliferation and invasion in a positive feedback loop9. Whether stiffening of the stromal matrix occurs in other solid tumours, besides mammary tumours, remains to be determined. However, despite recent technological advances (TABLE 1), remarkably little is known about the molecular and physical mechanisms that drive motile cancer cells away from primary tumour and into the stromal space, especially at the subcellular level.
Motility in three dimensions. Much of what we have learned about the physical and molecular mechanisms driving normal and cancer cell motility has come from in vitro studies using two-dimensional (2D) substrates12–14. However, the dimensionality of the system used to study cancer invasion can have a key role in dictating the mode of cell migration. This is not entirely surprising as the three-dimensional (3D) microenvironment of the ECM in vivo is characterized by many features, including the pore size and fibre orientation, features that are not found in conventional ECM-coated
O P I N I O N
The physics of cancer: the role of physical interactions and mechanical forces in metastasisDenis Wirtz, Konstantinos Konstantopoulos and Peter C. Searson
Abstract | Metastasis is a complex, multistep process responsible for >90% of cancer-related deaths. In addition to genetic and external environmental factors, the physical interactions of cancer cells with their microenvironment, as well as their modulation by mechanical forces, are key determinants of the metastatic process. We reconstruct the metastatic process and describe the importance of key physical and mechanical processes at each step of the cascade. The emerging insight into these physical interactions may help to solve some long-standing questions in disease progression and may lead to new approaches to developing cancer diagnostics and therapies.
Figure 1 | The metastatic process. In this complex process, cells detach from a primary, vascularized tumour, penetrate the surrounding tissue, enter nearby blood vessels (intravasation) and circulate in the vascular system. Some of these cells eventually adhere to blood vessel walls and are able to extrav-asate and migrate into the local tissue, where they can form a secondary tumour.
PERSPECT IVES
512 | JULY 2011 | VOLUME 11 www.nature.com/reviews/cancer
© 2011 Macmillan Publishers Limited. All rights reserved
Invasive programs globally rewire cell signaling
EMT. Opposing this function, multiple pro-angiogenic
proteins were increased, notably angiopoietin-1 (ANGPT1),co-expression of thrombospondins together with the
ADAMTS1 protease [86] and VEGF-C.
Multiple receptor protein kinase networks were acti-vated in the mesenchymal state. Autocrine Gas6 stimula-
tion of Axl and Tyro3 activity were established from array
and RT-PCR RNA abundance measurements. The datasuggest inhibitors of Axl and Tyro3 activity may show
utility in limiting the survival of shed mesenchymal-like
NSCLC cells. A functional role of autocrine Axl-Gas6expression in lung [87] and breast cancer [88] progression
has been established. Autocrine activation of FGFR and
PDGFR signaling also was observed in both established(H1703) and metastable TGFb models, supporting previous
findings [25, 89]. This increase in kinase activity and
PDGFR autophosphorylation was associated withincreased PDGFR and PDGF RNA transcripts. In meta-
stable Snail and Zeb1 models, FGFR1 and PDGFR tran-
scripts were markedly increased, PDGF was modestlyincreased and FGF1 was unchanged or slightly decreased
(Supplementary Table S5). These data support the predic-
tion of kinase activation by autocrine co-expression ofreceptor–ligand transcripts. We asked what functional
EMT dependent acquisition of ITGα5β1/FN and TGFβ signaling, activation of MAPK1/Erk2 and PTK2/FAK, inactivation of GSK3
1 100 1 100
H358
Erlotinib (uM)
pErk
Erk
pAkt(473)
Akt
β-Actin
H358(TGF β)B
Glypican1/6FGFR1
FGF1/2
p
AXLTYRO3
ADAMTS1
CTGF
IKIP
TGFβR1/2BMPR2
TGFB1/2/3
LTBP2
FAM129BS628,S679,
S683
BRAF p
HDGF
RAS
FAM129A
p
pPTK2pRBM35A/B
GAS6
FGFBP1FGF1/2
ITGα5β1
FN
MARCKS
ANXA6
PEA15
PFTK1TGFBI
PXNS84/S85
T284/S288Y118
MAPK1Y187
SRCp
MST4
CAV1p
DUSP4/6G3BP1
G3BP2
ENG
CCDC80
HDGF
P2RY5
PTPN14
MAPK14p
PKN1GADD45
SGK1 GSK3InactiveS278
AKT1AKTIP
AKT3DDIT4
DIDO1
INHBB
THBS1/2
p
PDGFRA
PDGF A/B/C
STAT3
PTPN11
p
CAV1pINA p
PIK3R1
PIK3CB
p
pp
PLCG1p
ANGPT1
IL8IL18
IL7CCL26
CGB
SCG5
VEGFC
IL11
IL6CLCF1
IL6ST
IL6RCNTRF JAK2
CRLF1
CXCR7
SPARC
DISP2
GIT2
GIT2
ANXA6
MAP3K3
pSTAT3
ISL1
PTX3 YARS
Fig. 5 a Gain of new survivaland cytokine networks.Increased signaling throughIL-11/IL-6/gp130, TGFb/BMP2, Axl/Tyro3, integrina5b1, PDGFR and FGFR wereobserved. Downstreamactivation of BRAF/MAPK1and PTK2/paxillin signalingwere observed. Components ingreen were attenuated duringEMT while those in lightbrown/red were increased withEMT. b Activating MAPK1/Erkphosphorylation (T185/Y187),associated with TGFb-inducedEMT, was confirmed byimmunoblot
Clin Exp Metastasis
123
co-attenuated with the EMT-mediated reduction in Met,
Ron, EGFR and IGF1R kinase activation. These included
multiple SH2, cytoplasmic kinase and ubiquinylationadapters: pSHC, pGrb2, pGAB1, pPTK6 (Brk), pUBB,
pUBIQ and pNCKAP1. We asked whether attenuated Met
phosphorylation might be attributable to reduced HGF
binding to Met. While no change in HGF transcript abun-
dance was observed, ST14 (matriptase), an extracellularprotease which promotes proHGF activation, was markedly
reduced. SPINT2 and SPINT1, which augment ST14
EGFRY1110Y1172
EGFRERBB2ERBB3
p
METMST1R/RON
AREG
EREG
TGFA
SHC1GRB2GRB2 pp p
p
PTK6p
GAB1p
Plasma membrane
BTCIGF1R
p
NCKAP1p
UBBp
ANXA9
SHC1
UBIQp
PDCD6
PDCD6IP
p
p
SHC1p
p
SFN p
ITGα6β4
ST14HGF
SPINT1SPINT2
PRSS8
MUC1/4
GPR110
PTAFR
H19
LRIG3
HBEGF
NRG
GRB2/7p
CDC2T14/Y15
MCM2S139
CDC2CDCA7L
PCNA
SFN
CDKN2D
TMPRSS4
CCND1CCNB1
IGFBP2
IGFBP4
IGFBP7
STK39
STAP2
S100P
S100A2EZR
p
S100A14
S100A6
S100A4
TP53
S100A11
ARHGEF16
IL4R
MAP3K8
PA2G4p
PTPRK
PTPRG
macrophage
EGF family
CSF1
NKD2
DDIT4
DKK1
CTNNB1
LRP5
APOE
CTGF
TSC2
RTK tyrosine phosphorylation measured by elisa array, comparing ratios of mesenchymal-like (Mes) and epithelial (Epi) NSCLC lines: all changes p<0.01 except ErbB4 (ns) where error bars reflect the standard error of the mean
Pho
spho
tyro
sine
Con
tent Mesenchymal
Epithelial
25 Calu6
H1703
H292
H358
Calu6
H1703
H292
H358
Calu6
H1703
H292
H358
CB
L/CB
LBH
SP
D1
SH
C1/S
HC
2
-3 -2.5 -2 -1.5 -1 -0.5 0 0.5
pY reduction w/ EGFR TKI
1 hr
4 hrs
24hrs
A
B C
Fig. 4 a Loss of autocrine RTK signaling networks through EGFR/ErbB2/ErbB3, Met/Ron and IGF1R signaling networks. AttenuatedIL4R, Muc1 and Muc4 and integrin a6b4 components also wereobserved. b Attenuated RTK tyrosine phosphorylation in H292 andH358, (epithelial phenotype) Calu6 and H1703 (mesenchymalphenotype) cell models were confirmed by quantitation of RTKarrays. c EGFR is functional and responds to exogenous ligand in
both epithelial and mesenchymal states. H292 and H358, (epithelial)Calu6 and H1703 (mesenchymal) cells were exposed to EGFR kinaseinhibitor (3 lM erlotinib, 2 h) or DMSO control prior to stimulationwith exogenous EGF (10 ng/ml, 10 min). The EGFR-dependentinhibition of substrates CBL and SHC were similar in epithelial andmesenchymal-like cells, relative for control HSPD1
Clin Exp Metastasis
123
Targeting invasion-specific signaling may considerably improve survival
Thomson et al, Clin Exp Metastasis, 2010
Targeting AXL has minor effects on the primary tumor
Gjerdrum et al, PNAS, 2010
0 5 100
10
20
30ControlshAXL2
*
Time (weeks)
Tum
or d
iamet
er (m
m)
MDA-MB-231
Gjerdrum et al, PNAS, 2010
0 5 100
10
20
30ControlshAXL2
*
Time (weeks)
Tum
or d
iamet
er (m
m)
Targeting AXL potently blocks metastasis
0 1 2 3 4 5 6 7
Lymph node
Lungs
Kidneys
Ovaries
Liver
Lymph node
Lungs
Kidneys
Ovaries
Liver
Lymph node
Lungs
Kidneys
Ovaries
Liver
Lymph node
Lungs
Kidneys
Ovaries
Liver
Metastases detected
BALB/c
shAXL
BALB/c
Control
MB-231
shAXL
MB231
Control
7 mice/group
Gjerdrum et al, PNAS, 2010
0 5 100
10
20
30ControlshAXL2
*
Time (weeks)
Tum
or d
iamet
er (m
m)
Blocking metastasis is sufficient for improving survival
Structure of AXL/Gas6 complex
Gas6
AXL
EGF domains
Gla domain
Fn domains
SHBD domains
Ig domains
Sasaki et al, EMBO J, 2006
The systems approach to studying cell phenotype
Migration Growth
How does signal processing modulate a migration response?
Environment/Cues Cell Signaling Migration response
Quantitatively understanding 3D cell migration regulation
EGFR activation leads to AXL crosstalk
AXL with Gas6 did not elicit a substantial signaling response, and thesesignaling consequences were small compared with EGF-elicited, AXL-dependent signaling effects (fig. S3E). This observation is similar to resultsin other studies of Gas6-elicited signaling inMDA-MB-231 (44). Why dif-ferent cell lines display markedly distinct receptor activation patterns toGas6 remains a question for future studies (44–46).
To investigate the crosstalk between AXL and EGFR (as well as MET)signaling further, we next examined the ratio of fold activation (phosphoryl-ation) of various signaling proteins in the absence versus presence of AXL(Fig. 3A). The unstimulated conditions represented signaling networkactivity presumably arising from constitutive autocrine processes. Thisanalysis revealed more widespread AXL-dependent effects in EGF- orTGFa-stimulated cells compared with HGF-stimulated cells, with the largestdifference in activation observed for GSK3 (glycogen synthase kinase 3)and Akt. Further, the relative magnitude of effects across the phosphositesinvestigated was correlated between EGF- or TGFa-stimulated cells andunstimulated cells but was not correlated between HGF-stimulated andunstimulated cells (Fig. 3A, inset). These results suggest that AXL maymediate similar basal and EGFR-stimulated signaling pathways in TNBCcells, whereas HGF yields a distinct downstream AXL-mediated signature.
We then performed principal components analysis (PCA) to gain insightconcerning the network-level variation in signaling across these treatmentconditions (Fig. 3B). Principal component 1 (PC1)was found to correspondto EGF-induced signaling, and PC2 to HGF-elicited signaling, with TGFahaving an intermediate effect. Knockdown of AXL moved cells negativelyalong PC1 and reduced the magnitude of the effect of EGF stimulation. Ex-amination of the loading plot revealed separation between phosphosites
only mildly affected by knockdown [for example, phosphorylation ofSTAT3 (signal transducer and activator of transcription 3) and JNK(c-JunN-terminal kinase)] and those strongly affected (such as the phospho-rylation of Akt and GSK3), with the rest scattered at intermediate locations(Fig. 3C). Together, these data indicate that EGF and TGFa induce ErbB-mediated downstream signaling that is qualitatively similar to basalsignaling but is distinct from MET-mediated signaling, and this baseline-like signaling is disrupted by AXL knockdown. The difference betweenHGFand EGF, TGFa, and baseline signaling is likely a result of the absenceof signaling from EGFR, HER2, or AXL in the former case, because METis presumably transactivated also in the EGF- or TGFa-stimulated cases.An appealing interpretation is that autocrine EGFR ligand activity is consti-tutive and transactivates AXL.
AXL amplifies signaling in the EGFR-associated pathwaybut does not sensitize EGFR to its ligandBecause receptor activation can be quantitatively characterized in terms ofligand concentration–related sensitivity and maximal activation at satura-tion, we investigated how AXL influences the dose response of EGFR toEGF. We stimulated MDA-MB-231 cells with a range of concentrations ofEGFandmeasured the pan-phosphotyrosine on EGFR and the phosphoryl-ation of Akt (Fig. 4, A and B). Phosphorylation of Akt was chosen for mea-surement as a critical downstream signal that was strongly influenced byAXL knockdown, though not to imply that all transactivation-mediatedeffects are regulated throughAkt alone (Fig. 3C). Phosphorylation of EGFRwas unaffected by AXL knockdown except at very high (above saturating)EGF concentrations (Fig. 4A), likely as a result of altered trafficking or
0
6
12
18
EGFControl
pan
-pY
siCt
rl
siA
XL
0.0
0.5
1.0
Nor
mal
ized
AXL
FI
A B
C
pAkt
pTyk2
pSTAT3
pSrc
pIRS1
pHSP27
pcJun
pP38
pJNK
pGSK3
pERK
siAXLD
EGFR HER2 MET IGF1R0.00
0.25
0.50
0.75
1.00
1.25
siAXL surface
siCtrl surface
siAXL total
siCtrl total
Nor
mal
ized
rece
ptor
(a.u
.)
MET AXL
* *
*
Ctr
l
HG
F
EGF
Ctr
l
HG
F
EGF
z score–2 20
siCTRL
Fig. 2. EGF stimulation transactivates AXL andMET. (A) ELISA-based pan-phosphotyrosine (pan-pY) measurement of alternative receptors after EGFstimulation inMDA-MB-231 (*P<0.05, Student’s t test). (B) AXL knockdown,measured by ELISA. (C) Total and surface amounts of alternative receptorsin AXL-silenced MDA-MB-231 cells (*P < 0.05, Student’s t test). Data are
means ± SEM of three biological measurements. (D) Downstreamsignaling assessed by kinase phosphorylation in MDA-MB-231 cells5 min after stimulation with EGF, TGFa, or HGF in the presence orabsence (siAXL) of AXL. Each phosphosite was mean-centered andvariance-normalized.
R E S E A R C H A R T I C L E
www.SCIENCESIGNALING.org 6 August 2013 Vol 6 Issue 287 ra66 4
on August 8, 2013 stke.sciencem
ag.orgD
ownloaded from
Meyer et al, Sci Sig, 2013MDA-MB-231
The systems approach to studying cell phenotype
Migration Growth
AXL
The systems approach to studying cell phenotype
Migration GrowthMigration Growth
AXL
AXL knockdown broadly influences cell signaling
Meyer et al, Sci Sig, 2013MDA-MB-231
AXL with Gas6 did not elicit a substantial signaling response, and thesesignaling consequences were small compared with EGF-elicited, AXL-dependent signaling effects (fig. S3E). This observation is similar to resultsin other studies of Gas6-elicited signaling inMDA-MB-231 (44). Why dif-ferent cell lines display markedly distinct receptor activation patterns toGas6 remains a question for future studies (44–46).
To investigate the crosstalk between AXL and EGFR (as well as MET)signaling further, we next examined the ratio of fold activation (phosphoryl-ation) of various signaling proteins in the absence versus presence of AXL(Fig. 3A). The unstimulated conditions represented signaling networkactivity presumably arising from constitutive autocrine processes. Thisanalysis revealed more widespread AXL-dependent effects in EGF- orTGFa-stimulated cells compared with HGF-stimulated cells, with the largestdifference in activation observed for GSK3 (glycogen synthase kinase 3)and Akt. Further, the relative magnitude of effects across the phosphositesinvestigated was correlated between EGF- or TGFa-stimulated cells andunstimulated cells but was not correlated between HGF-stimulated andunstimulated cells (Fig. 3A, inset). These results suggest that AXL maymediate similar basal and EGFR-stimulated signaling pathways in TNBCcells, whereas HGF yields a distinct downstream AXL-mediated signature.
We then performed principal components analysis (PCA) to gain insightconcerning the network-level variation in signaling across these treatmentconditions (Fig. 3B). Principal component 1 (PC1)was found to correspondto EGF-induced signaling, and PC2 to HGF-elicited signaling, with TGFahaving an intermediate effect. Knockdown of AXL moved cells negativelyalong PC1 and reduced the magnitude of the effect of EGF stimulation. Ex-amination of the loading plot revealed separation between phosphosites
only mildly affected by knockdown [for example, phosphorylation ofSTAT3 (signal transducer and activator of transcription 3) and JNK(c-JunN-terminal kinase)] and those strongly affected (such as the phospho-rylation of Akt and GSK3), with the rest scattered at intermediate locations(Fig. 3C). Together, these data indicate that EGF and TGFa induce ErbB-mediated downstream signaling that is qualitatively similar to basalsignaling but is distinct from MET-mediated signaling, and this baseline-like signaling is disrupted by AXL knockdown. The difference betweenHGFand EGF, TGFa, and baseline signaling is likely a result of the absenceof signaling from EGFR, HER2, or AXL in the former case, because METis presumably transactivated also in the EGF- or TGFa-stimulated cases.An appealing interpretation is that autocrine EGFR ligand activity is consti-tutive and transactivates AXL.
AXL amplifies signaling in the EGFR-associated pathwaybut does not sensitize EGFR to its ligandBecause receptor activation can be quantitatively characterized in terms ofligand concentration–related sensitivity and maximal activation at satura-tion, we investigated how AXL influences the dose response of EGFR toEGF. We stimulated MDA-MB-231 cells with a range of concentrations ofEGFandmeasured the pan-phosphotyrosine on EGFR and the phosphoryl-ation of Akt (Fig. 4, A and B). Phosphorylation of Akt was chosen for mea-surement as a critical downstream signal that was strongly influenced byAXL knockdown, though not to imply that all transactivation-mediatedeffects are regulated throughAkt alone (Fig. 3C). Phosphorylation of EGFRwas unaffected by AXL knockdown except at very high (above saturating)EGF concentrations (Fig. 4A), likely as a result of altered trafficking or
0
6
12
18
EGFControl
pan
-pY
siCt
rl
siA
XL
0.0
0.5
1.0
Nor
mal
ized
AX
L FI
A B
C
pAkt
pTyk2
pSTAT3
pSrc
pIRS1
pHSP27
pcJun
pP38
pJNK
pGSK3
pERK
siAXLD
EGFR HER2 MET IGF1R0.00
0.25
0.50
0.75
1.00
1.25
siAXL surface
siCtrl surface
siAXL total
siCtrl total
Nor
mal
ized
rece
ptor
(a.u
.)
MET AXL
* *
*
Ctr
l
HG
F
EGF
Ctr
l
HG
F
EGF
z score–2 20
siCTRL
Fig. 2. EGF stimulation transactivates AXL andMET. (A) ELISA-based pan-phosphotyrosine (pan-pY) measurement of alternative receptors after EGFstimulation inMDA-MB-231 (*P<0.05, Student’s t test). (B) AXL knockdown,measured by ELISA. (C) Total and surface amounts of alternative receptorsin AXL-silenced MDA-MB-231 cells (*P < 0.05, Student’s t test). Data are
means ± SEM of three biological measurements. (D) Downstreamsignaling assessed by kinase phosphorylation in MDA-MB-231 cells5 min after stimulation with EGF, TGFa, or HGF in the presence orabsence (siAXL) of AXL. Each phosphosite was mean-centered andvariance-normalized.
R E S E A R C H A R T I C L E
www.SCIENCESIGNALING.org 6 August 2013 Vol 6 Issue 287 ra66 4
on April 3, 2014 stke.sciencem
ag.orgD
ownloaded from
PCA deconvolves transactivation effect
cellular processes induced at such nonphysiological amounts of stimulation.Other receptor-proximal components, such as the adaptor protein SHC andthe CDC2 kinase, exhibited similar phosphorylation after stimulation at theEGF dose used in the signaling studies here (fig. S4). In contrast, AXLknockdown affected the phosphorylation of Akt in response to all dosesof EGF by a shift inmagnitude (“vertically”) rather than in sensitivity (“hor-izontally”) (Fig. 4B). To deconvolve these concomitant changes in the phos-phorylation of EGFR andAkt, we plotted the abundance of phosphorylatedAkt as a function of phosphorylated EGFR in cells treated with either con-trol siRNA or AXL siRNA (Fig. 4C). This revealed a uniform downwardshift across all stimulation amounts in the absence of AXL, indicating aconsistent fold change in the magnitude of signal transduction. Each curvecould bewell described to first approximation by a Hill function, with com-parable Kd (threshold of half-maximal activation) but markedly differentmaximal activation (Fig. 4D). To identify the level at which this regulationmay occur, we fit these data to alternative models of signal transductionfrom the receptor layer (see Materials and Methods). The data were bestexplained by a model in which basal and stimulated AXL activities exist,the latter in proportion to EGFRactivation and inwhich transduction of both
signals occurs through separately saturable processes (table S3). Thismodelis consistent with our biochemical observations (Fig. 2A). The effect ofbaseline activation ofAXLcan be observed from the plot of phosphorylatedAkt as a function of pan-phosphotyrosine EGFR, where at low EGFR ac-tivation in the presence ofAXL, the phosphorylation ofAktwas higher thana simpleHill regressionwould suggest (Fig. 4C). Biologically, this indicatesthat the components downstream of the receptor are saturated by maximalEGFR activation and that, at least with respect to phosphorylated Akt, thetransactivation of AXL increases the effective amount of RTK signaling andamplifies the signaling consequence of stimulation.
Multipathway signaling correctly predicts AXLknockdown inhibition of EGF-stimulated protrusionWe next asked how the broad effects on signaling that resulted from AXLknockdown might influence the migration behavior of cells. We elected touse acute membrane protrusion as a surrogate measurement of three-dimensional migratory capacity on the basis of our previous findings thatthis assay corresponds well to growth factor–stimulated invasive motilitywithin extracellular matrix (47). Protrusion measurements from wild-type
pSTAT3pJNK
pP38
pHSP27
pSrcpTyk2
pERK
pAktpGSK3
pIRS1pcJun
5
–4
5–5
EGF HGF
Component 1 (58%)
Com
pon
ent 2
(24%
)
–2
B CsiControl
siAXL
pGSK3 pAkt pcJun pERK pP38 pTyk2 pIRS1 pSrc pHSP27 pJNK pSTAT32–2
2–1
20
21
22
Rat
io o
f fol
dac
tivat
ion
with
siA
XL
A
5
5
–4Component 1 (58%)
Com
pon
ent 2
(24%
)
0.0 1.0
* *
EGF
EGF
HGF
Vs. unstim. Effect corr.
Fig. 3. AXL knockdown attenuates downstream signaling in MDA-MB-231. (A) Ratios of fold activation after treatment with growth factorin AXL knockdown cells relative to wild-type cells: ([siAXL GF]/[siAXLUnstim]) ÷ ([siControl GF]/[siControl Unstim]). The unstimulated bar in-dicates the ratio of unstimulated abundance: [siAXL Unstim]/[siControl
Unstim]. Inset shows the Spearman correlation across all phosphositesbetween the unstimulated and stimulated ratios (*P < 0.05). (B) PCA scoreplot of signaling data after AXL knockdown. Line colors indicate stimulationconditions denoted in (A). (C) Loading plot of signaling data after AXLknockdown.
R E S E A R C H A R T I C L E
www.SCIENCESIGNALING.org 6 August 2013 Vol 6 Issue 287 ra66 5
on April 3, 2014 stke.sciencem
ag.orgD
ownloaded from
Meyer et al, Sci Sig, 2013MDA-MB-231
Crosstalk quantitatively amplifies a qualitatively distinct set of pathways
MDA-MB-231 cells were used to train a family of partial least-squares re-gression models for how protrusion activity depends on multiple phospho-proteomic signals. Minimal models that use only three signals wereexamined to ascertain the most vital pathway predictors for growthfactor–induced motility. We identified models that fit the data by cross-validation (Q2 > 0.6) and found that these were enriched in inclusion ofGSK3, STAT3, and Akt as the key predictor signals (fig. S5A). Thesemodels involved similar weights for the predictor signals in both PCs,demonstrating consistency across the ensemble of top-fitting models intheir multipathway signaling-to-protrusion relationships (Fig. 5A).
This ensemble of models was then used to predict wild-typeMDA-MB-231 protrusion by cross-validation and to a priori predict protrusion mod-ulation by AXL knockdown (Fig. 5B). EGF-stimulated protrusion waspredicted to be the most substantially attenuated response after AXLknockdown, whereas HGF-stimulated protrusion was predicted to remainessentially unaffected. These predictions were indeed correct inMDA-MB-231 cells transfected with AXL siRNA: HGF-elicited protrusion was notsignificantly affected, whereas EGF-elicited protrusionwas significantly re-duced (Fig. 5C). Treatment with R428 confirmed that EGF-stimulated pro-trusion depended on AXL-mediated signaling in another TNBC line,MDA-MB-157, but that it did not in two other breast cancer cell lines,MCF7 and T47D, which lack AXL expression (Fig. 5D). The effect ofR428 phenocopied that of AXL siRNA treatment in terms of the protrusionresponse to EGF inMDA-MB-231 cells (Fig. 5, C andD). TGFa-stimulated
protrusion was also reduced in MDA-MB-231 cells by R428 treatment, al-though to a lesser degree, which was in accord with our model predictions(fig. S5B). These results indicate that along with amplification of EGFR-induced downstream signaling, the transactivation of AXL additionallyactivates a qualitatively distinct set of signals that are important for cellmigration in response to stimuli. Moreover, our three-pathway partial least-squares regression model successfully captured the integrated effects ofthese signals on this phenotypic response.
AXL is in proximity to ErbB and MET but not IGF1R or IRWe investigated whether the transactivation of AXL (and MET) by EGFRmight involve physicochemical proximity of these RTKs. Because of tech-nical limitations in capability for distinguishing receptor colocalization byother methods (fig. S6, A and B), we used a technique in which immuno-precipitation of cross-linked receptors from lysate was performed in amultiplexed fashion on barcoded fluorescent beads. The degree of AXLcross-linking with each of various other RTKs was quantified using anAXL antibody (Fig. 6A). Across multiple cell lines, we observed a signif-icant degree of AXL cross-linking with ErbB receptors, MET, and PDGFRbut not with insulin receptor (INSR) or IGF1R (Fig. 6B and fig. S6C). Theamount of AXL cross-linkingwas roughly proportional to the abundance ofthat particular RTK—with the exception of INSR and IGF1R, neither ofwhich garnered cross-linked AXL to a measureable extent (Fig. 6C). Weconfirmed cross-linking results with reciprocal immunoprecipitation assays
0 0.5 1 2 4 8 16 32 64 27 28 290
1000
2000
3000
4000
5000
6000
7000
EGF (ng/mL)
pAkt
0 2 4 8 16 32 64 128 256 5120
1000
2000
3000
4000
EGF (ng/mL)
EGFR
pan
-pY *
*
****
* * **
*
8 16 32 64 128 256 512 210 211 212256
512
1024
2048
4096
8192siControl
siAXL
EGFR pan-pY
0
50
100
150
200
pAkt
/pY
EGFR
Kd
0
2000
4000
6000
8000
pAkt
/pY
EGFR
Bm
ax
P < 0.001n.d.
A
B
C
D
2–2
siCtrl siAXL siCtrl siAXL
pAkt
Fig. 4. AXL amplifies the EGFR signaling response. (A) ELISA of pan-pYEGFR in wild-type (siControl) or AXL-silenced (siAXL) MDA-MB-231 after5 min of treatment with varying EGF amounts. Data are means ± SEM.P < 0.05, Student’s t test. n = 3. (B) Phosphorylation of Akt in response to arange of EGF doses. Data are means ± SEM. P < 0.05, Student’s t test. n = 3.
(C) ELISA for the abundance of pan-pY on EGFR versus the phosphorylationof Akt in MDA-MB-231 cells. Lines show a Hill regression to each set of datawith SE of biological triplicate measurements. (D) Hill regression of each plotshows similar Kd values but significantly different maximal activation (F test).Error bars indicate SE of the fit.
R E S E A R C H A R T I C L E
www.SCIENCESIGNALING.org 6 August 2013 Vol 6 Issue 287 ra66 6
on August 8, 2013 stke.sciencem
ag.orgD
ownloaded from
forE, the value ofEwas calculated with lsqnonlin by solving for the valuethat minimizes:
arg minE∈ ½0,1"
D1
IC50,1
!E
1 − E
" 1m1
þ D2
IC50,2
!E
1 − E
" 1m2
þ aD1D2
IC50,1IC50,2
!E
1 − E
" 12m1
þ 12m2
− 1
#########
#########
Fitting was performed with the nlinfit function within MatLab with ini-tial parameters identified by inspection of the single drug data.Optimizationwas unconstrained, but IC50 values were in all cases significantly positive,and all m values were significantly negative as verification of effectivefitting. Confidence intervals presentedwere calculated from the empiricallyderived Jacobianwith the nlparci function. Significancewas separately ver-ified by jackknife (60).
Partial least-squares regression and PCAReplicate measurements were averaged, and each signaling variable wasmean-centered and variance-normalized before further analysis. PCAwasperformed with singular value decomposition within the pca function. Thefirst two components explained 83% of the variance.
For reduced partial least-squares modeling, the output variable wasassembled from mean protrusion measurements for each growth factor,and the unstimulated condition set to 0. Model reduction was performedby training models with all possible combinations of three input variablesets. Three variables were chosen, as it was the smallest model size withsufficient well-trained models to ensure robust variable enrichment. Modelreduction with larger reducedmodels produced qualitatively similar results.Each individual reducedmodelwas then used concomitantly, and the resultswere shown by displaying the average and SE of loading values and pre-dictions. As a result of variation in baseline signaling, predictions forknockdown cells were taken to be the prediction for the knockdown andstimulated conditionminus the prediction of the knockdown and unstimu-lated condition.
Amplification modelingEach model was fit with the nlinfit function. To ensure robustness withrespect to initial parameter selection, fitting was performed 100 timeswith randomly selected initial parameters within the range of feasiblevalues. x is 0 with AXL knocked down and 1 with AXL present. [pEGFR]and [pAkt] are from measurements of pan-phosphotyrosine EGFR and pAktacross a dose range of EGF.Models were compared using the corrected anduncorrected Akaike information criterion denoted AICc and AIC, respec-tively (61).
To fit data to a model in which activation of AXL is in proportion toEGFR activation, and signaling integration is receptor-proximal, Eq. 1was used:
½pAkt" ¼ Bmax ð½pEGFR" þ a½pEGFR"xÞKD þ ð½pEGFR" þ a½pEGFR"xÞ
þ B0
For amodel inwhich amplification ofAkt activationwith respect to a setamount of EGFR activity, and AXL only affects this proportional relation-ship, Eq. 2 was used:
½pAkt" ¼ Bmax ð1þ axÞ½pEGFR"KD þ ½pEGFR"
þ B0
For a model in which no signaling effect from AXL exists, Eq. 3 wasused:
½pAkt" ¼ Bmax ½pEGFR"KD þ ½pEGFR"
þ B0
For a model in which some baseline activation of AXL is possible inaddition to proportional activation, and signaling integration is receptor-proximal, Eq. 4 was used:
½pAkt" ¼ Bmax ð½pEGFR" þ a½pEGFR"xþ xbÞKD þ ð½pEGFR" þ a½pEGFR"xþ xbÞ
þ B0
For a model in which Akt activated by AXL and activated by EGFR issummed, with proportional activation of AXL, Eq. 5 was used:
½pAkt" ¼Bmax,1½pEGFR"KD,1 þ ½pEGFR"
þxBmax,2½pEGFR"KD,2 þ ½pEGFR"
B0
For a model in which no signaling effect from AXL through EGFRpathway exists, but there is a baseline effect ofAXLpresent, Eq. 6was used:
½pAkt" ¼ Bmax½pEGFR"KD þ ½pEGFR"
þ xbþ B0
For a model with only baseline activation of AXL, and signalingintegration is receptor-proximal, Eq. 7 was used:
½pAkt" ¼ Bmax½pEGFR" þ axKD þ ½pEGFR" þ ax
þ B0
For a model with Akt activated by AXL and activated by EGFRsummed, with proportional and baseline activation of AXL, Eq. 8was used:
½pAkt" ¼Bmax,1½pEGFR"KD,1 þ ½pEGFR"
þxBmax,2ð½pEGFR" þ bÞKD,2 þ ð½pEGFR" þ bÞ
Total receptor quantificationTotal receptor amounts were measured with a bead-based ELISA (Novagen).For quantification of AXL and MER, established ELISA antibodies andstandards (R&D Systems) were used. The capture antibody was conjugatedto unconjugated beads (Bio-Rad) and used in a multiplexed fashion with theother targets. Linearity of the assay was validated during measurement bydilution series of both the lysates and standards.
Each cell linewas seeded sparsely, and the next daywas starved for 4 hoursand lysed. Receptor measurements were normalized to total protein content toprovide a receptormass fraction (that is, femtogramof receptor permilligramofcell lysate). Thismass fractionwas used in all subsequentmodeling. For recep-tor density calculations, a subconfluent plate of cellswas trypsinized, thenumberof cells was counted and lysed, and total protein was quantified. This providedtheconversion, for each cell line, frommilligramof lysate to number of cells.Combined with the known mass of each receptor, a value could then be con-verted tonumber of receptors per cell. Finally, receptor densitywas calculatedbyusing the surface area of a HeLa cell [1600 µm2, BNID 103718 (62)].
Assay selectionTodeterminewhichmethodsmight be suitable for studying such complexeson the cell surface, we developed a simple statistical model to describe the
R E S E A R C H A R T I C L E
www.SCIENCESIGNALING.org 6 August 2013 Vol 6 Issue 287 ra66 11
on August 8, 2013 stke.sciencem
ag.orgD
ownloaded from
Meyer et al, Sci Sig, 2013MDA-MB-231, Akt pS473
siCtrlsiAXL
AXL is required for EGF-elicited protrusion
in MDA-MB-231, in which we observed the association of AXL withEGFR but not with IGF1R (fig. S6D).
On the basis of these data, we sought a quantitative framework to under-stand the respective amounts of complexing observed between AXL andeachRTKacross different cell lines.According to fundamental stoichiomet-ric considerations, the amount of AXL observed in complex with a partic-ular RTK in a particular cell line should be approximately the product of theRTK abundance in that cell line, with proportionality described by coeffi-cients constituting (i) the cross-linking and protein loading efficiency and(ii) the antibody immunoprecipitation efficiencies and extent of colocaliza-tion.Withmeasurements of RTKabundance and the amount cross-linked toAXL, we determined the remaining parameters (see Materials andMethods) to provide away to account for differences in receptor expressionwhen interpreting cross-linking data (fig. S6D). With this quantitative for-mulation, we could then calculate whether the parameter characterizingAXL/RTK colocalization deviated significantly from 0 for each RTK(Fig. 6D). Significant deviation from 0 indicates colocalization. DespiteIGF1R and INSR being substantively abundant in various cell lines, thecalculated likelihood that they localized with AXL was not significant. Al-though this parameter includes the efficiency of immunoprecipitating
IGF1R or INSR, we verified that these two receptors were detected withsimilar efficiency both by direct ELISAof the same cell lysates and by quan-tification of a recombinant standard.We additionally confirmed cross-linkedimmunoprecipitation betweenAXLandEGFR to the exclusion of IGF1Rbyreciprocal immunoprecipitation in MDA-MB-231 (fig. S6E). Our quantita-tive analysis framework ruled out the possibility that merely low abundanceof IGF1R and INSR was a trivial explanation for the absence of significantcolocalization. We therefore conclude that AXL is colocalized with ErbB,MET, and PDGFR but not with IGF1R or INSR.
The amount of EGFR-AXL complex was much greater in MDA-MB-231 than in other cell lines, likely as a result of the differences in abundanceof EGFR (Fig. 6C).MCF7 cells transfectedwithAXLand treatedwith EGFshowed no synergistic response characteristic of receptor transactivation,consistent with the relatively little EGF-elicited signaling overall (fig.S7A). We therefore considered whether we could predict the importanceof AXL transactivation induced by activation of RTKs other than EGFR.MDA-MB-453 cells have large amounts of HER2 and HER3 in complexwithAXL, so our notionwould predict that AXL signalingmight contributeto a heregulin (HRG)–stimulated response in these cells. We learned bydirect test, using AXL transfection and HRG treatment, that this is in fact
siAXL (test) siCtrl (training)0
1
2
3
Nor
mal
ized
pro
trus
ion
resp
onse
–1.2
–1.2
–1.6
–2.0
–0.8
–0.8
–0.4–0.4
0.4 0.8 1.2 1.6 2.0
0.4
0.8pGSK3
pcJun
pIRS1
pSTAT3
pAkt
Principal component 1
Prin
cipa
l com
pone
nt 2
MB-231 MB-157 MCF7 T47D
***
***
n.s.
n.s.
EGF HGF
BX/Y loadingsA
C DNo inhibitor0.3 µM R428
–400
0
400
800
1200
Are
a ch
ange
[um
2 ]
siControlsiAXL
***
n.s.
EGF HGF
–400
0
400
800
1200
Are
a ch
ange
[um
2 ]
Fig. 5. AXL signaling is required for EGF-elicited protrusion. (A) Mean load-ings of the reduced partial least-squares regression models. The red pointcorresponds to theprojectionof the phenotype. Error bars indicate the SE forthe family of reducedmodels. (B) Protrusion predictions from reducedpartialleast-squares regressionmodels for wild-type (by cross-validation) and AXLknockdown (by prediction) cells. Error bars indicate the SE of prediction
across the family of reduced models. (C) EGF-elicited protrusion responseof MDA-MB-231 cells upon AXL knockdown (***P < 0.001, Mann-Whitneytest; n= 13 to 25 from three independent experiments). (D) EGF-elicited pro-trusion responses with or without 0.3 mM R428 (***P < 0.001, Mann-Whitneytest; n = 17 to 35 from three independent experiments). MDA-MB-231 andMDA-MB-157 cells express AXL, whereas MCF7 and T47D cells do not.
R E S E A R C H A R T I C L E
www.SCIENCESIGNALING.org 6 August 2013 Vol 6 Issue 287 ra66 7
on April 3, 2014 stke.sciencem
ag.orgD
ownloaded from
in MDA-MB-231, in which we observed the association of AXL withEGFR but not with IGF1R (fig. S6D).
On the basis of these data, we sought a quantitative framework to under-stand the respective amounts of complexing observed between AXL andeachRTKacross different cell lines.According to fundamental stoichiomet-ric considerations, the amount of AXL observed in complex with a partic-ular RTK in a particular cell line should be approximately the product of theRTK abundance in that cell line, with proportionality described by coeffi-cients constituting (i) the cross-linking and protein loading efficiency and(ii) the antibody immunoprecipitation efficiencies and extent of colocaliza-tion.Withmeasurements of RTKabundance and the amount cross-linked toAXL, we determined the remaining parameters (see Materials andMethods) to provide away to account for differences in receptor expressionwhen interpreting cross-linking data (fig. S6D). With this quantitative for-mulation, we could then calculate whether the parameter characterizingAXL/RTK colocalization deviated significantly from 0 for each RTK(Fig. 6D). Significant deviation from 0 indicates colocalization. DespiteIGF1R and INSR being substantively abundant in various cell lines, thecalculated likelihood that they localized with AXL was not significant. Al-though this parameter includes the efficiency of immunoprecipitating
IGF1R or INSR, we verified that these two receptors were detected withsimilar efficiency both by direct ELISAof the same cell lysates and by quan-tification of a recombinant standard.We additionally confirmed cross-linkedimmunoprecipitation betweenAXLandEGFR to the exclusion of IGF1Rbyreciprocal immunoprecipitation in MDA-MB-231 (fig. S6E). Our quantita-tive analysis framework ruled out the possibility that merely low abundanceof IGF1R and INSR was a trivial explanation for the absence of significantcolocalization. We therefore conclude that AXL is colocalized with ErbB,MET, and PDGFR but not with IGF1R or INSR.
The amount of EGFR-AXL complex was much greater in MDA-MB-231 than in other cell lines, likely as a result of the differences in abundanceof EGFR (Fig. 6C).MCF7 cells transfectedwithAXLand treatedwith EGFshowed no synergistic response characteristic of receptor transactivation,consistent with the relatively little EGF-elicited signaling overall (fig.S7A). We therefore considered whether we could predict the importanceof AXL transactivation induced by activation of RTKs other than EGFR.MDA-MB-453 cells have large amounts of HER2 and HER3 in complexwithAXL, so our notionwould predict that AXL signalingmight contributeto a heregulin (HRG)–stimulated response in these cells. We learned bydirect test, using AXL transfection and HRG treatment, that this is in fact
siAXL (test) siCtrl (training)0
1
2
3
Nor
mal
ized
pro
trus
ion
resp
onse
–1.2
–1.2
–1.6
–2.0
–0.8
–0.8
–0.4–0.4
0.4 0.8 1.2 1.6 2.0
0.4
0.8pGSK3
pcJun
pIRS1
pSTAT3
pAkt
Principal component 1
Prin
cipa
l com
pone
nt 2
MB-231 MB-157 MCF7 T47D
***
***
n.s.
n.s.
EGF HGF
BX/Y loadingsA
C DNo inhibitor0.3 µM R428
–400
0
400
800
1200
Are
a ch
ange
[um
2 ]
siControlsiAXL
***
n.s.
EGF HGF
–400
0
400
800
1200
Are
a ch
ange
[um
2 ]
Fig. 5. AXL signaling is required for EGF-elicited protrusion. (A) Mean load-ings of the reduced partial least-squares regression models. The red pointcorresponds to theprojectionof the phenotype. Error bars indicate the SE forthe family of reducedmodels. (B) Protrusion predictions from reducedpartialleast-squares regressionmodels for wild-type (by cross-validation) and AXLknockdown (by prediction) cells. Error bars indicate the SE of prediction
across the family of reduced models. (C) EGF-elicited protrusion responseof MDA-MB-231 cells upon AXL knockdown (***P < 0.001, Mann-Whitneytest; n= 13 to 25 from three independent experiments). (D) EGF-elicited pro-trusion responses with or without 0.3 mM R428 (***P < 0.001, Mann-Whitneytest; n = 17 to 35 from three independent experiments). MDA-MB-231 andMDA-MB-157 cells express AXL, whereas MCF7 and T47D cells do not.
R E S E A R C H A R T I C L E
www.SCIENCESIGNALING.org 6 August 2013 Vol 6 Issue 287 ra66 7
on April 3, 2014 stke.sciencem
ag.orgD
ownloaded from
Meyer et al, Sci Sig, 2013
How do TAM receptors process extracellular cues?
Environment/Cues Cell Signaling Migration response
Quantitatively understanding 3D cell migration regulation
Conundrum: AXL does not robustly respond to ligand stimulation
0 5 100
5
10
15
20
25 900 ng/mL AF154100 ng/mL Gas6
Time (min)
pan-
pY A
XLc
MB453SKBR3
12Z MB231A549
A172U87 BT549
03570
1000
2000
3000
4000
Rece
ptor
(fg/m
g)
baMDA-MB-231
0 0.156 0.625 2.50
1
2
3 12Z
0
1
2
3
4
5
A549
0
1
2
3
4
BT549
0.0
0.5
1.0
1.5
2.0 A172
0
2
4
6
8
U87
0
2
4
6
8
10
pYTotal
pY/Total
Fold
Uns
timul
ated
Fold
Uns
timul
ated
Fold
Uns
timul
ated
Fold
Uns
timul
ated
Fold
Uns
timul
ated
Fold
Uns
timul
ated
Gas6 (nM)
0 0.156 0.625 2.5Gas6 (nM)
0 0.156 0.625 2.5Gas6 (nM)
0 0.156 0.625 2.5Gas6 (nM)
0 0.156 0.625 2.5Gas6 (nM)
0 0.156 0.625 2.5Gas6 (nM)
AXLTYRO3MER
0 100
1
4.5
Minutes
AXL
pY
d
Figure S1
3
MDA-MB-231
Attributes of a good sensor
0 1 2 3 4 0 1 2 3 4
Input Output
Large dynamic range
Attributes of a good sensor
0 1 2 3 4 0 1 2 3 4
Input Output
Rapid response
Attributes of a good sensor
0 1 2 3 4 0 1 2 3 4
Input Output
Memoryless
Many RTKs can be considered as “ligand concentration sensors”
100 ng/mL EGF/IGF1, 50 ng/mL HGF, hMLE-Twist1 Kim et al, Mol Cell Proteomics, 2011
Gas6 stimulation reveals complex response patterns
0 5 10
1.0
1.5
2.0
2.5
Minutes
AXL
pY
0 5 10
1.0
1.5
2.0
Minutes
BT549
A172
Gas6 (ng/mL) 100 4 20 0.8
AXL
pY
0 5 10
1.0
1.5
2.0
0 5 100.5
1.0
1.5
2.0
2.5
3.0
Minutes
U87
A549AX
L pY
AXL
pY
Minutes
AXL still poorly responsive to Gas6 at longer times
BT549
0.0
0.5
1.0
1.5
2.0
Fold
Uns
timul
ated
0 0.156 0.625 2.5Gas6 (nM)
A172
0
2
4
6
8
Fold
Uns
timul
ated
0 0.156 0.625 2.5Gas6 (nM)
pYTotal
pY/Total
A549
0
1
2
3
4
Fold
Uns
timul
ated
0 0.156 0.625 2.5Gas6 (nM)
U87
0
2
4
6
8
10
Fold
Uns
timul
ated
0 0.156 0.625 2.5Gas6 (nM)4 hrs stimulation
Structure of AXL/Gas6 complex
Gas6
AXL
EGF domains
Gla domain
Fn domains
SHBD domains
Ig domains
Sasaki et al, EMBO J, 2006
A differential equation model of receptor activation
[B](t
) ÷ [B
](t →
∞)
0
0.5
1
Time
Gas6
Ig1Ig2
A0 A1
A2A1,2
D1
D2
Rxn1
Rxn2Rxn4
Rxn3
Rxn6 Rxn5
TAM kinetic model allows mechanistic interpretation
AXL has a limited rapid response to stimulation
0 100
1
5.5
Minutes
AXL
pY
All possible outputs
Ptds Exposure is a Spatially Localized Process
Ruggiero et al, PNAS, 2012
10 µm
Exposed Ptds Membrane
5 µm
Perhaps robust receptor activation only occurs in response to local stimulation?
Cell debris
AXL expressing cell
Local stimulation shifts species abundance
A0 A1 A2 A1,2 D1 D2
10-2 10-1 100 101 102 103
Spatial inhomogeneity parameter-5
0
310
4 #/c
ell d
iffer
ence
vs.
hom
ogen
eous
stim
.
Gas6
Ig1Ig2
A0 A1
A2A1,2
D1
D2
Rxn1
Rxn2Rxn4
Rxn3
Rxn6 Rxn5
Local stimulation strongly promotes local AXL signaling
A0 A1 A2 A1,2 D1 D2
10-2 10-1 100 101 102 103
Spatial inhomogeneity parameter-5
0
3
104 #
/cell
diff
eren
ce v
s.ho
mog
eneo
us s
tim.
AXL
pY D
ensit
yat
Gas
6 Pe
ak
10-2 10-1 100 101 102 103
Spatial inhomogeneity parameter
1400
0
D [L2/min] 0 0.1 1 10
Local stimulation results in greater overall AXL signaling
AXL
pY D
ensit
yat
Gas
6 Pe
ak
10-2 10-1 100 101 102 103
Spatial inhomogeneity parameter
1400
0
D [L2/min] 0 0.1 1 10
Spatial inhomogeneity parameter
Bulk
AXL
pY3.0
2.0
1.0
0.010-2 10-1 100 101 102 103
D [L2/min] 0 0.1 1 10
A0 A1 A2 A1,2 D1 D2
10-2 10-1 100 101 102 103
Spatial inhomogeneity parameter-5
0
3
104 #
/cell
diff
eren
ce v
s.ho
mog
eneo
us s
tim.
MB231
0 10 20 300.0
0.5
1.0
1.5
2.0
2.5
Gas6, LipidGas6Gas6, 1/10 LipidLipid
Minutes
AXL
pY
A549
0 10 20 300.0
1.0
2.0
3.0
Minutes
AXL
pY
Gas6, LipidGas6Gas6, 1/10 LipidLipid
Local stimulation results in greater overall AXL signaling
20 ng/mL Gas6; 1X lipid: 100 μg/mL 5:3:2 w/w PE:PS:PC
(Adapted from Wikipedia)
Synergistic activation is specifically due to local stimulation
0 5 100
1
2
3 Gas6BSA
Pixels
pY S
tain
ing
Inte
nsity
Position of beads
17 um
Gas6-coated polystyrene beads
AXL is a sensor of ligand spatial heterogeneity
Low Gas6 High Gas6
Localized Gas6
Conclusion• AXL transactivation leads to selective pathway
amplification
• AXL-expressing cells rely on the receptor for robust migration response
• TAM receptors most robustly sense local ligand stimulation
• A quantitative model of AXL activation can be used to predict the consequence of novel interventions
Future directions
• Use the TAM signaling model to understand transactivation
• Model expansion to the full TAM family
• Rational design of improved TAM-targeted therapies
• Evaluation of TAM-mediated immune targeting
ThanksUROP/Summer Students
Jenny Kay Amalchi Castillo Ashley Powers Courtney Weldon Philip Smith Ceri Riley
CollaboratorsMiles Miller Shannon Hughes Hyung-Do Kim Michael Beste Joel Wagner
ThanksLauffenburger lab
Douglas Lauffenburger Hsinhwa Lee Miles Miller Shannon Hughes Brian Joughin Dave Clarke Hyung-Do Kim Elma Kurtagic Sarah Schrier Michael Beste Doug Jones Joel Wagner Allison Claas Butterstick Hughes
Griffith labShelly Peyton Edgar Sanchez Walker Inman
ThanksGertler lab
Frank Gertler Madeleine Oudin Guillaume Carmona Russel McConnell Michele Balsamo
Committee MembersForest White Alan Wells
FacilitiesEliza Vasile
HMS Orchestra Cluster KI Frontier Program
Thanks
Thanks
Thanks
Thanks
Questions?