Integration of mTOR and estrogen–ERK2 signalingin lymphangioleiomyomatosis pathogenesisXiaoxiao Gua, Jane J. Yub, Didem Iltera, Nickolas Blenisa, Elizabeth Petri Henskeb, and John Blenisa,1

aDepartment of Cell Biology, Harvard Medical School, Boston, MA, 02115; and bBrigham and Women’s Hospital, Boston, MA, 02115

Edited by Melanie H. Cobb, University of Texas Southwestern Medical Center, Dallas, TX, and approved July 25, 2013 (received for review May 14, 2013)

Lymphangioleiomyomatosis (LAM) is a destructive lung disease ofwomen associated with the metastasis of tuberin-null cells withhyperactive mammalian target of rapamycin complex 1 (mTORC1)activity. Clinical trials with the mTORC1 inhibitor rapamycin haverevealed partial efficacy but are not curative. Pregnancy appearsto exacerbate LAM, suggesting that estrogen (E2) may play a rolein the unique features of LAM. Using a LAM patient-derived cellline (bearing biallelic Tuberin inactivation), we demonstrate thatE2 stimulates a robust and biphasic activation of ERK2 and tran-scription of the late response-gene Fra1 associated with epithelial-to-mesenchymal transition. In a carefully orchestrated collabora-tion, activated mTORC1/S6K1 signaling enhances the efficiency ofFra1 translation of Fra1mRNA transcribed by the E2–ERK2 pathway,through the phosphorylation of the S6K1-dependent eukaryotictranslation initiation factor 4B. Our results indicate that targetingthe E2–ERK pathway in combination with the mTORC1 pathwaymay be an effective combination therapy for LAM.

estrogen signaling | mTORC1 signaling | ERK signaling | EMT

Lymphangioleiomyomatosis (LAM) is a destructive, progres-sive, multisystem disease in which smooth muscle-like “LAM

cells” invade the lung (1–6) and is associated with chylous pleuraleffusions, lymphatic obstruction, and renal angiomyolipomas (7).LAM occurs almost exclusively in women with onset often oc-curring in the childbearing years (5, 6, 8, 9). A hallmark of LAMis mutational inactivation of both alleles of the TSC2 gene, whichencodes tuberin (TSC2) (1, 2, 4, 5, 10, 11). Loss of TSC2 resultsin constitutively activated mammalian target of rapamycincomplex 1 (mTORC1) signaling (5), which is a primary targetfor LAM therapy because of its dominant role in regulatingcellular metabolism and growth (12). In a recent clinical trial, themTORC1 inhibitor sirolimus (a rapamycin analog) stabilizedlung function in LAM patients (12). However, sirolimus treat-ment was not curative, and the benefits were observed onlyduring the treatment period, because patients regained disease-related symptoms posttreatment (12, 13). We have focused ourattention on estrogen (E2) signaling because the disease is ex-acerbated during pregnancy. E2, upon binding to its estrogenreceptor (ER), has been reported to regulate transcription-dependent and -independent signaling events (14). Thus, inaddition to its ability to promote changes in gene expression(10, 15–18), E2 can induce the activation of signaling proteinssuch as Src, Akt, and ERK-MAP kinase (14). The importanceof ERK-MAP kinase in LAM was suggested by a recent reportshowing that E2 promoted the MEK-dependent invasion ofcells derived from Eker rat uterine leiomyoma (ELT3 cells) intothe lungs of ovariectomized mice (18). However, the molecularbasis for E2-dependent ERK contribution to the enhanced in-vasive phenotype in the presence of constitutively activatedmTORC1 was not defined, and whether E2 promoted a similarresponse in patient-derived cells remained to be determined.Here we used TSC2-null LAM patient-derived angiomyolipoma(AML) cells as a platform to examine the cellular response toE2 and the potential interaction between the E2–ERK and ac-tivated mTORC1 pathways in establishing the metastatic-likephenotype observed in LAM.

ERK is a versatile signaling molecule capable of mediatingdistinct cellular fates depending on the strength and duration ofactivation as well as the cellular localization of the active enzyme(19, 20). For example, the rapid activation of ERK and its nu-clear translocation are sufficient to induce immediate-early genes(IEGs) such as c-Fos and c-Jun, which contribute to the expressionof activator protein 1 (AP1) transcription factor complexes. In-creased AP1 levels lead to the induction of late-response genes(LRGs). However, without sustained nuclear activation of ERK,many of the products of these genes are unstable and thereforeare poorly induced. To stabilize these IEG products a strong,sustained ERK activation that promotes AP1 accumulation andLRG induction. The LRG product, Fra1, also is stabilized bysustained nuclear ERK activation (21). Our laboratory recentlyhas found that in MCF-10A human mammary epithelial cells, theRas-mediated EMT-associated change in morphology as well asthe ability to migrate and invade are mediated through sustainedERK2 signaling, which leads to expression of the LRG Fra1. Fra1expression in turn leads to elevated expression of the E-cadherintranscriptional repressor zinc finger E-box-binding homeobox 1/2(ZEB1/2), which contributes to increased cell migration and in-vasion (22).In addition to regulation of cellular responses by ERK sig-

naling, cells have evolved a very elaborate and sensitive mecha-nism to adapt to environmental changes by regulating cellularmetabolism and protein synthesis, and the mTORC1 pathway isa major orchestrator of these processes (23). mTORC1-mediatedcap-dependent translation initiation involves the phosphorylationof 4E-binding protein (4E-BP) and its detachment from the 5′ cap-binding protein eukaryotic translation initiation factor 4E (eIF4E),allowing the recruitment of the scaffolding protein eIF4G andassociated proteins such as poly-A binding protein (PABP) andthe eIF4A RNA helicase to the 5′ end of a target mRNA.mTORC1 activation also promotes the assembly of the translationpreinitiation complex, ribosome scanning, and translation initia-tion. Another level of translation control occurs for mRNAs withhighly structured 5′UTRs. This relatively small number of mRNAsis enriched for regulators of bioenergetics, cell cycle, and angio-genesis. Phosphorylation of eIF4B by the mTORC1–S6K pathwaypromotes its recruitment into the translation preinitiation complex,where it is proposed to promote significantly the RNA helicaseactivity of eIF4A leading to the unwinding of the hairpin structurespresent in the long 5′ UTRs of these mRNAs and thereby greatlyenhancing their translation efficiency (23, 24).We have identified several unique features of both ERK and

mTORC1 signaling in LAMpatient-derived cells. We observe thatE2 contributes to migration and invasion by stimulating a critical

Author contributions: X.G. and J.B. designed research; X.G. and N.B. performed research;X.G., J.J.Y., D.I., E.P.H., and J.B. contributed new reagents/analytic tools; X.G. and J.B.analyzed data; and X.G. and J.B. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1To whom correspondence should be addressed. E-mail: jblenis@hms.harvard.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1309110110/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1309110110 PNAS Early Edition | 1 of 6

CELL

BIOLO

GY

Dow

nloa

ded

by g

uest

on

July

30,

202

0

biphasic ERK2 activation these cells. We also demonstrate thatFra1 plays an important role in regulating E2-stimulated epithe-lial-to-mesenchymal transition (EMT)-like features in these cells.We provide evidence indicating that in the presence of E2 andhighly active mTORC1, ERK2, and mTORC1 signaling convergeat the level of Fra1 to orchestrate its transcription and translation,respectively, and to stimulate migration and invasion in LAMpatient-derived cells. Taken together, these observations suggestthat inhibiting E2 signaling in combination with mTORC1 in-hibitionmay target LAMcell proliferation andmigration/invasionselectively. We suggest that combination therapy along the E2–

ERK pathway (for example, with an ER antagonist) and along themTORC1 pathway with rapamycin analogs [drugs that have beenapproved by the Food andDrugAdministration (FDA)]may offera highly effective treatment for LAM.

ResultsE2-Stimulated Cell Migration and Invasion Correlate with Robust andBiphasic ERK Activation in LAM patient-derived cells. Previousresults had linked E2-stimulated ERK activity to increased in-vasion of rat ELT3 cells into mouse lungs in a xenograft model(5, 9). To test whether LAM patient-derived cells also were E2responsive along the ERK pathway, we first measured the migra-tory and invasive response of E2-stimulated cells in the presenceor absence of the MEK1/2 inhibitor AZD6244 (AZD). E2 stim-ulated LAM cell migration and invasion across Transwell supports(Fig. 1 A and B). These cellular behaviors were associated with

decreased expression of the epithelial marker E-cadherin and in-creased expression of the mesenchymal marker N-cadherin uponE2 stimulation; these effects were reversed by inhibition with AZD(Fig. 1C), indicating a role for ERK signaling in the migration andinvasion of LAM cells in vitro.To understand better how ERK contributes to cell migration,

we examined the kinetics of E2-dependent ERK activation (Fig.1D). We found that both ERK isoforms, but most prominentlyERK2 (addressed in detail in Fig. 2) displayed a reproduciblebiphasic activation pattern with an early phase around 1 h anda late phase around 6–8 h after E2 stimulation (Fig. 1D). BecauseERK activation was blocked completely by 1-h pretreatment withAZD (Fig. 1D), it is unlikely that E2 activates ERK independentlyof MEK1/2 (e.g., by inhibiting the expression of an ERK phospha-tase). To understand how E2 treatment generates the biphasic ERKactivation pattern, we investigated whether both peaks were tran-scription dependent by inhibiting transcription with Actinomycin D(ActD) (Fig. 1D). Treatment of cells with ActD 1 h before or afterE2 treatment did not affect the first ERK activation peak but re-duced the second peak to basal levels. Therefore, the first ERKactivation peak is transcription independent, whereas the secondpeak is transcription dependent.

ERK2 Is Required for Mediating E2-Regulated EMT in LAM patient-derived cells. We previously had shown a role for ERK2 in cellmigration and survival in mammary epithelial cells (22, 25).Using lentiviral-based shRNAi constructs to knock down ERK1or ERK2, we found that E2 promoted migration in both controlGFP and ERK1-KD cells to almost the same level, whereasERK2 knockdown significantly reduced cell migration to a level aslow as that observed when ERK signaling was inhibited by the

A B C

D

0

20

40

60

80

100

Ctrl AZD

E2

Migration

0

10

20

30

40

50

60

Ctrl AZD

E2

Invasion

Cells p

er area

Cells

per a

re

a

E2

Time 0 6Hr 12Hr 24Hr

E-Cad

N-Cad

Tubulin

AZ

DA

ZD

Fig. 1. E2-stimulated cell migration and invasion correlate with robust andbiphasic ERK activation in LAM patient-derived cells. (A and B) LAM cells, 1 hafter attachment to the upper Transwell membrane (30,000 cells per well),were treated with or without 10 nM E2 (lower well) in the presence or ab-sence of the MEK1/2 inhibitor AZD (5 μM) (upper well) in phenol red-, EGF-,and FBS-free IIA complete medium and were studied for migration and in-vasion. Cells that migrated/invaded to the other side of the Transwell insertmembrane were fixed, stained, and quantified. (C) LAM cells were starvedovernight in phenol red-, EGF-, and FBS-free IIA complete medium and werestimulated with 10 nM E2 with or without pretreatment with the MEK1/2AZD (5 μM) for 1 h, Levels of EMT-associated markers E-cadherin (E-Cad) andN-cadherin (N-Cad) were analyzed at different time points by immunoblot.Tubulin was used as an internal control. (D) LAM cells were starved as de-scribed above. The transcription inhibitor ActD (5 μg/mL) was added 1 hbefore or after 10 nM E2 stimulation, and the MEK1/2 inhibitor AZD (5 μM)was added 1 h before E2 stimulation. Levels of phosphorylated ERK1/2were analyzed at different time points by immunoblot and were quan-tified by ImageJ. Total ERK1/2 was used as an internal control. Each experi-ment was performed in triplicate. Graphs are representative of multipleexperiments (n = 5).

0

20

40

60

80

100

120

Ctrl

E2

E2 +AZD

12Hr 24Hr #1 #2 #1 #2

GFP ERK1 KD ERK2 KD

Ce

lls

pe

r a

re

a

Migration

0

30

60

90

120

150

Ce

lls

pe

r a

re

a

12Hr 24Hr

Ctrl ER

K1

WT

ER

K2

WT

Migration

0

10

20

30

40

50

60

Cells

per a

re

a

12Hr 24Hr

Ctrl

Invasion

ER

K1

WT

ER

K2 W

T

Tubulin

ERK1

ERK2

GFP ERK2 KD

T7-ERK1

HA-ERK2

E-Cad

N-Cad

A B

C

D E

F

GFP ERK1KD ERK2KD

#1 #1

#2 #2

ERK1 shRNA

ERK2 shRNA

E-Cad

N-Cad

Tubulin

Fig. 2. ERK2 is required for mediating EMT in LAM patient-derived cells.(A) Control or ERK isoform KD cells, after 1-h attachment to the upperTranswell membrane (30,000 cells per well), were treated with 10 nM E2in the presence or absence of AZD (5 μM) and were studied for migration/invasion. Cells that migrated/invaded to the other side of the Transwell in-sert membrane were fixed, stained, and quantified. (B) Morphology of ERKisoform KD in LAM cells. (Magnification: 20×; scale bar: 50 μm.) (C) Expres-sion of EMT-associated markers E-cadherin and N-cadherin in ERK isoformKD cells as shown in A. ERK knockdowns were completed with two distinctshRNA constructs in A, B, and C (SI Materials and Methods). (D and E) Themigration and invasion of LAM cells expressing ERK WT were studied usingthe method described in A. For rescue experiments in ERK KD cells (ERK2shRNA #2), T7-tagged ERK1 WT or HA-tagged ERK2 WT was expressed. Thelevels of ERK1/2 and EMT markers are shown in F. Tubulin was used as aninternal control. Each experiment was performed in triplicate. Graphs arerepresentative of multiple experiments (n = 3).

2 of 6 | www.pnas.org/cgi/doi/10.1073/pnas.1309110110 Gu et al.

Dow

nloa

ded

by g

uest

on

July

30,

202

0

MEK1/2 inhibitor AZD (Fig. 2A). Furthermore, ERK2 knock-down was associated with a transition from a mesenchymal spin-dle-like phenotype to an epithelial cuboidal-like morphology (Fig.2B). A mesenchymal-to-epithelial transition (MET) was con-firmed by decreased expression of the mesenchymal marker N-cadherin and increased expression of the epithelial marker E-cadherin (Fig. 2C).In single-isoform ERK-KD cells, the ability of the remaining

ERK isoform to respond to E2 stimulation was unaffected (Fig.S1A). To analyze further the specific influence of ERK2 on theEMT-like phenotype of LAM patient-derived cells, we overex-pressed T7-tagged WT ERK1 or HA-tagged WT ERK2 in LAMcells. LAM cells overexpressing ERK2 WT but not ERK1 WTpromoted migration and invasion (Fig. 2 D and E). In addition,overexpression of functional ERK1 WT at levels greater thanendogenous ERK2 did not rescue the effects of ERK2 knock-down, whereas rescue with ERK2 WT did reverse the phenotype(Fig. 2F and Fig. S1B). Together, these data indicate ERK2,rather than ERK1, is the main contributor to the E2-inducedmigration and invasion in LAM patient-derived cells.

E2 Induces Fra1 Transcription to Cell Migration and Invasion in LAMpatient-derived cells. We recently reported that the regulationof Fra1 and its effector ZEB1/2 downstream of ERK2 signalingis important in promoting EMT in mammary epithelial cells (22,25). Because the second peak of the biphasic ERK activationafter E2 stimulation for 6–8 h was transcription-dependent (Fig.1D), we hypothesized that this later or sustained ERK activa-tion is associated with transcription-dependent changes in gene

expression linked to EMT. Supporting this notion, we detectedan increase in the expression of the LRGs Fra1 and ZEB1 inLAM patient-derived cells stimulated with E2 (Fig. 3A).We next knocked down or overexpressed Fra1 to demonstrate

further a role for this gene in the E2-stimulated invasive phenotypeof LAM patient-derived cells. As shown in Fig. 3 B–F, ZEB1 ex-pression was associated with Fra1 expression upon E2 stimula-tion. Fra1 knockdown led to a significant decrease in migrationand invasion, and this change in behavior was tightly linked to aswitch from a mesenchymal-like to an epithelium-like pheno-type, with decreased expression of N-cadherin and increased ex-pression of E-cadherin. Upon Fra1 overexpression, these resultswere reversed (Fig. 3F). Importantly, knockdown of Fra1 did notaffect E2-stimulated ERK activation peaks (Fig. S2A). Combined,our results suggest that in LAM cells cultured under the con-ditions described, E2 regulates migration and invasion throughERK2–Fra1–ZEB1/2 signaling.

E2Stimulates Fra1 Transcription Downstream of the ERK2 PathwayIndependent of mTORC1 Inhibition in LAM patient-derived cells. Wehave shown that E2–ERK2 signaling regulates the migrationand invasion of LAM cells through the induction of the EMT-associated gene Fra1. We have found that rapamycin also canpartially suppress migration/invasion in these cells without pre-venting the induction of Fra1 mRNA (Fig. 4A and Fig. S2 B andC). We therefore investigated whether Fra1 protein expressionis regulated by mTORC1 and whether Fra1 is a point of con-vergence for the E2–ERK2 and activated mTORC1 pathways inthe LAM cells. Because Fra1 mRNA has a structured 5′ UTR

0 15’ 30’ 1Hr 2Hr 4Hr 8Hr 12Hr

E2

pERK1

pERK2

ERK1

ERK2

Fra1

ZEB1

Tubulin

GFP #1 #2 Fra1 WT

Fra1

E-Cad

N-

Tubulin

Fra1 KD

Invasion

GFP Fra1 WT

Fra1 KD #1 Fra1 KD #2

A

E

F

Migration

Cells p

er area

Cells p

er area

D

GFP #1 #2 Fra1

WTFra1 KD

GFP #1 #2 Fra1

WTFra1 KD

0

0.2

0.4

0.6

0.8

1

1.2

Fra1

ZEB1

Ctrl Fra1 KD #1 Fra1 KD #2

E2 8 Hour

Fo

ld

In

crease

B

C

0

20

40

60

80

0

30

60

90

120

Cad

Fig. 3. E2-induced Fra1 expression contributes to migration and invasion of LAM patient-derived cells. (A) Time-course E2 stimulation in LAM cells. Cells werestarved as previously described, stimulated with 10 nM E2, and lysed at different time points. Levels of phospho-ERK1/2, Fra1, and ZEB1 were analyzed byimmunoblot. Total ERK1/2 and tubulin were used as internal controls. (B) ZEB1 expression correlated with Fra1 expression upon 8-h E2 stimulation in controlor Fra1 KD cells. (Two distinct shRNA constructs were used. See SI Materials and Methods.) (C) Morphology of Fra1 KD and Fra1 WT in LAM cells. (Magni-fication: 20×; scale bar: 50 μm.) (D and E) After 1-h attachment to the upper Transwell membrane (30,000 cells per well), control cells, Fra1 KD cells, or LAMcells expressing Fra1 WT were treated with 10 nM E2 and ertr studied for migration/invasion. Cells that migrated/invaded to the other side of the Transwellinsert membrane were fixed, stained, and quantified. (F) Expression of EMT-associated markers E-cadherin and N-cadherin in Fra1 KD cells or LAM cellsexpressing Fra1 WT. Tubulin was used as an internal control. Each experiment was performed in triplicate. Graphs are representative of multiple experiments(n = 3).

Gu et al. PNAS Early Edition | 3 of 6

CELL

BIOLO

GY

Dow

nloa

ded

by g

uest

on

July

30,

202

0

(Fig. S3) and because mTORC1 enhances the translation effi-ciency of mRNAs with a highly structured 5′ UTR, we hypoth-esized that E2–ERK2 signaling regulates Fra1 transcription andFra1 stability (26), whereas mTORC1 might regulate Fra1 trans-lation efficiency.We first measured the mRNA levels of Fra1 and its down-

stream effector ZEB1/2 in E2-treated LAM cells using real-timePCR. Because Fra1 is an LRG, and based on our data shown inFig. 1D, the LAM patient-derived cells were stimulated with E2for 8 h and were treated with ActD or rapamycin for 1 h beforeor after E2 stimulation. Treatment of these cells with ActD orrapamycin for 1 h after E2 allows the induction of IEGs butwill reveal if Fra1 expression depends on its transcription and/ortranslation, respectively. E2 promoted an increase in mRNAlevels of both Fra1 and ZEB1/2 in a transcription-dependentmanner, because pretreatment with the transcription inhibitorActD (5 μg/mL) reduced their expression (Fig. 4 A–C). The mRNAlevels of Fra1, in contrast to ZEB1/2, were not affected dramaticallywhen rapamycin was added 1 h before E2 stimulation (Fig. 4D),consistent with the role of rapidly induced IEGs, such as c-Fosand c-Jun, in the subsequent expression of LGRs such as Fra1 andthe contribution of Fra1 to the induction of ZEB1/2. Combinedwith the observation that addition of the MEK inhibitor after 1 halso prevented Fra1 induction, these data indicate the impor-tance of the later ERK activation peak at around 6–8 h in me-diating events linked to Fra1 and ZEB1/2 expression and EMT-likephenotypes.Interestingly, although the mRNA expression of Fra1 remained

elevated with rapamycin treatment at 8 h after E2 treatment, we

found that the expression of its downstream effector ZEB1/2 wasreduced significantly (Fig. 4 E and F). This result suggested thatFra1 protein expression might be suppressed by rapamycin, thuspreventing it from inducing ZEB1/2 expression. We next investi-gated if the mTORC1 effector S6K1 regulates the expression ofFra1 and ZEB1/2 mRNA. To do so, we used the S6K1-specificATP-competitive inhibitor PF4708671. As shown in Fig. 4 D–F,inhibiting either mTORC1 or S6K1 did not affect Fra1 but didsuppress ZEB1/2 mRNA levels induced by E2 at 8 h, suggestingthat mTORC1/S6K1 might mediate the translation of Fra1 down-stream of E2–ERK2 pathway.

mTORC1 and S6K1 Enhance the Translation Efficiency of Fra1 TranscribedDownstream of the E2–ERK2 Pathway Through eIF4B Phosphorylation.To test the possibility that constitutively activated mTORC1might contribute to enhanced translation of Fra1, we performedthe same experiment described in Fig. 4, but this time we examinedthe protein levels of Fra1 and ZEB1. Rapamycin inhibition wasconfirmed by reduced phosphorylation of its downstream regulatorS6K1 at T389, whereas blocking the kinase activity of S6K1 withPF4708671 did not affect the phosphorylation of S6K1 in theseTSC2-null LAM patient-derived cells but did antagonize down-stream signaling (Fig. 5A). Both Fra1 and ZEB1/2 protein levelswere reduced significantly upon either mTORC1 or S6K1 inhibition(Fig. 5 A and B). Given that Fra1 mRNA expression was notinhibited by rapamycin, but ZEB1/2 expression was, that Fra1 isan upstream regulator of ZEB1/2 expression (22), and that theexpression of Fra1 and ZEB1 protein was suppressed significantlyby rapamycin treatment, we conclude that E2-activated ERK2regulates the expression of the Fra1 gene and the stability of Fra1protein (19, 20) and that mTORC1-S6K1 enhances the efficiencyof Fra1 translation. It is worth noting that neither mTORC1 norS6K1 inhibition affected the activation of ERK by E2 at 8 h,indicating that hyperactive mTORC1 signaling is independent ofE2–ERK2 signaling.Because of the loss of TSC2, LAM cells have a high basal level

of eIF4B phosphorylation. Reduced eIF4B phosphorylation byupstream inhibition of mTORC1 or S6K1 inhibition was asso-ciated with significantly reduced expression of Fra1 and ZEB1(Fig. 5 A and B). To support our hypothesis that mTORC1–S6K1 signaling could regulate the translation efficiency of Fra1mRNA, we predicted that Fra1 is highly structured within its 5′UTR (Fig. S3). The translation efficiency of mRNAs with highlystructured 5′ UTRs is enhanced by the RNA helicase activity ofeIF4A, which is proposed to be regulated by S6K1-mediatedphosphorylation and recruitment of the eIF4B regulatory sub-unit into the translation preinitiation complex (23). To confirmfurther the crucial role of eIF4B in regulating Fra1 translation inLAM cells, we generated LAM cell lines that stably expressedeIF4B WT or a S422D mutant that was constitutively phospho-mimetic at S422. E2 promoted Fra1 expression at 8 h in eIF4BWT and S422D mutant cells. Inhibiting the mTORC1 pathwayresulted in a significant drop in Fra1 in eIF4BWT cells, whereas theexpression of Fra1 and ZEB1 was rescued in cells expressing thephosphomimetic eIF4B S422D upon S6K1 inhibition (Fig. 5 C andD). Our data suggest that the hyperactive mTORC1–S6K1 pathwayworks downstream of E2-ERK2–dependent Fra1 transcription toregulate its translation efficiency through phosphorylation of eIF4B,further confirming the crucial role of eIF4B in regulating Fra1translation in TSC2-null LAM cells.

DiscussionThe development of LAM appears to involve a multistep pro-cess, with initiation caused by mutations resulting in loss offunction of the tumor-suppressor protein complex and a hy-pothesized role for E2 in promoting disease progression (5, 9).Because of its constitutive activation in LAM cells, mTORC1 hasbeen considered the primary therapeutic target for LAM (5, 12).

A

FC

B

0

1

2

3

4

5Fra1

8Hr

mRNA

Ctrl Rapa Rapa

PF4708671

E2

0

5

10

15

20

25 ZEB1

8Hr

mRNA

Ctrl Rapa Rapa

PF4708671

E2

0

5

10

15

20

25

30

ZEB2

8Hr

mRNA

Ctrl Rapa Rapa

PF4708671

E2

0

1

2

3

4

5

6

1 Hr Pre-block before E2

Block after 1Hr E2

Ctrl ActD Rapa ActD+Rapa

E2

Fo

ld

in

crease

Fra1

8Hr

mRNA

0

5

10

15

20

25

1 Hr Pre-block before E2

Block after 1 Hr E2

Fo

ld

in

cre

as

e

Ctrl ActD Rapa ActD+Rapa

E2

ZEB2

8Hr

mRNA

0

5

10

15

20

25

1 Hr Pre-block before E2

Block after 1 Hr E2

Ctrl ActD Rapa ActD+Rapa

E2

ZEB1

8Hr

mRNA

Fo

ld

in

crease

Fo

ld

in

crease

Fo

ld

in

crease

Fo

ld

in

crease

E

D

Fig. 4. E2-ERK2 regulates Fra1 mRNA expression. (A–C) SYBR Green-basedreal-time PCR analysis of Fra1 and ZEB1/2 mRNA expression at 8 h in LAMpatient-derived cells starved overnight and treated with 10 nM E2 in thepresence or absence of the transcription inhibitor ActD (5 μg/mL) or rapa-mycin (rapa; 20 nM) or their combination 1 h before or after E2 stimulation.β-Actin was used as an internal control. (D–F) After overnight starvation,LAM cells were treated with 20 nM rapamycin and/or the S6K1 inhibitorPF4708671 (30 μM) for 1 h before E2 stimulation. mRNA expression of Fra1(D) and ZEB1/2 (E and F) were analyzed at 8 h. Each experiment was performedin triplicate. Graphs are representative of multiple experiments (n = 3).

4 of 6 | www.pnas.org/cgi/doi/10.1073/pnas.1309110110 Gu et al.

Dow

nloa

ded

by g

uest

on

July

30,

202

0

Indeed, sirolimus, a rapamycin analog, recently was shown tostabilize pulmonary function in LAM patients (12). However, theclinical efficacy of rapamycin is not permanent, and cessation oftreatment results in reacquisition of most disease symptoms (13,27). The long-term risks and benefits of continuous treatmentwith mTORC1 inhibitors in this patient population, and whetherdrug resistance will emerge, are unknown but are predictedbased on several studies. We have focused our attention on otherpathways that work upstream of or in parallel with the mTORC1pathway and which may represent additional targets for LAMtherapy. Because E2 has long been thought to contribute to thedevelopment of LAM (6, 10, 15–18), we have investigated thepossible connections between these signaling processes. Here, wehave identified a mechanism for the coordination and integrationof E2-stimulated ERK activation, Fra1 induction, and mTORC1/S6K1-mediated Fra1 translation in contributing to an EMT-likephenotype in LAM patient-derived cells.We have found that E2 causes a biphasic activation of ERK1/2

in LAM patient-derived cells (621-101 cells). The early-phaseERK activation (0.5–1 h) is transcription independent, whereasthe later-phase ERK activation (6–8 h) is transcription dependent.We also demonstrate that sustained ERK2 signaling is requiredfor the invasive features of the LAM patient-derived cells. Sus-tained ERK signaling is important for the LRG and protein-expression profiles and for dictating specific cell fates (19, 20, 28).For example, Fra1 is an LRG that increases cell migration andinvasion associated with EMT (22, 25).We therefore examined the possible connection between the

E2–ERK pathway and the constitutively active mTORC1 path-way at the level of the transcription factor Fra1 in E2-stimulatedLAM patient-derived cells based on (i) our observations thatthe later transcription-dependent ERK activation mediates in-duction of the EMT-associated LRG Fra1 in LAM cells; (ii) ourprevious observations that ERK2 signaling is critical for the

acquisition of a mesenchymal-like phenotype in mammary epithelialcells expressing Ras-V12 during EMT; and (iii) the demonstrationthat TSC2-null LAM cells have constitutively high mTORC1

T0Ctrl ActD Rapa

PF4708671E2 8 Hour

ERK1/2p-S6KT389Fra1

ZEB1

Tubulin

p-eIF4B

eIF4B

ActD Rapa

Fold

incr

ease

Ctrl ActD Rapa ActD RapaPF4708671

E2

00.20.40.60.8

11.21.41.61.8

Fra1

Flag-eIF4B

Tubulin

p-eIF4B

Fra1

E2 8 HourPF4708671

eIF4B

AZD

ZEB1

00.20.40.60.8

11.21.4

Empty vectorEmpty vectorWTS422D

PF4708671E2

AZD

8Hr Fra1

Fold

incr

ease

A

D

C

B

Fig. 5. mTORC1 regulates Fra1 mRNA translation efficiency. (A and B) After overnight starvation, LAM cells were treated with 5 μg/mL ActD, 20 nMrapamycin, and/or the S6K1 inhibitor PF4708671 (30 μM) for 1 h before E2 stimulation. Cells were lysed at 8 h for protein expression analyses. Levels ofphosphorylated ERK1/2, phospho-S6K1 T389, phospho-eIF4B (pSer422), eIF4B, Fra1, and ZEB1 were analyzed by immunoblot. Tubulin was used as an internalcontrol. Fra1 expression in A was normalized to the internal control and was quantified in B by ImageJ. (C and D) LAM cells stably expressing eIF4B WT orS422D were starved as described and were treated for 1 h with AZD (5 μM) or PF4708671 (30 μM) in the presence of 10 nM E2. Levels of phospho-eIF4B, eIF4B,Fra1, and ZEB1 were analyzed by immunoblot. Tubulin was used as an internal control. Fra1 expression in C was normalized to the internal control andquantified in D by ImageJ. Graphs are representative of multiple experiments (n = 3).

MEK 1/2

ERK 2

LAM patient-derivedTSC2-null

angiomyolipoma cells TuberinHamartin

Rheb

mTORC1constitutively active

Estrogen

Fra1 mRNA

Fra1 Protein

MigrationInvasion

ZEB1/2

Transcription-independentand-dependent ERK2 activation

Loss of TSC2

AZD6244

Rapamycin

S6K

p-eIF4B

PF4708671Translationefficiency

eIF4A

P

P

Phosphorylationstabilization

Fig. 6. Summary of the convergence of the E2–ERK and mTORC1 signalingpathways in regulating migration and invasion in LAM patient-derived cells.These have TSC2 mutations and therefore are hyperactive in mTORC1 activity(10). E2 activates the ERK pathway, uses biphasic and sustained ERK2 activation,and stimulates the transcriptional induction of the LRG, Fra1. Concurrently,constitutively activation of mTORC1 in LAM cells enhances the translationof accumulating Fra1 mRNA. Blocking mTORC1/S6K1 with either an mTORC1inhibitor or an S6K1 inhibitor inhibits the phosphorylation of eIF4B andtherefore decreases the translation efficiency of Fra1 and downstream cel-lular processes such as ZEB1/2 induction, EMT markers, and cell migrationand invasion. Inhibitors used in this study are the transcription inhibitor ActD(5 μg/mL), the MEK1/2 inhibitor AZD6244 (AZD; 5 μM), the mTORC1 inhibitorrapamycin (rapa; 20 nM), and the S6K1 inhibitor PF4708671 (30 μM).

Gu et al. PNAS Early Edition | 5 of 6

CELL

BIOLO

GY

Dow

nloa

ded

by g

uest

on

July

30,

202

0

activity regardless of any stimulation. Given that rapamycintreatment of these LAM cells also affects their migratory andinvasive properties, we asked how mTORC1 contributes to thesephenotypes: Does it act upstream or downstream of the E2–

ERK2 pathway, and does it act in parallel with another pathwaythat also regulates migration and invasion (29) or in a separatepathway that converges to modulate a key regulator of the EMT-like phenotype observed in E2-treated LAM cells? Our datareveal a coordination between these pathways and thus supportthe latter possibility. We have shown that E2 acts mainly throughthe ERK2 isoform and promotes the transcriptional up-regula-tion of the Fra1 gene. The constitutively activated mTORC1translational machinery enhances the translation efficiency ofFra1 mRNA, which is predicted to have a secondary structure atits 5′ UTR. The presence of highly phosphorylated eIF4B, whichis sensitive to mTORC1 and S6K1 inhibitors in these cells,appears to increase the efficiency of Fra1 mRNA translation.Our findings also demonstrate that the combination of bothmTORC1 signaling and E2–ERK2 signaling is more effectivethan targeting either pathway alone (Fig. S2 B and C).In conclusion, the E2-regulated ERK2 pathway and the con-

stitutive mTORC1 pathway converge on the Fra1–ZEB1/2 tran-scriptional network to promote migration and invasion in LAMpatient-derived cells, as summarized in Fig. 6. These observationsreveal the importance of examining the effects of targeting E2signaling in combination with mTORC1 pathway in treating LAM,because this combination potently inhibits distinct pathways thatconverge to control cell growth, proliferation, migration/invasion,and survival. These findings have immediate significance and clini-cal impact. Both the ER antagonist fulvestrant and rapamycin are

FDA approved, and thus preclinical trials now can be designedwhich we anticipate will reveal the effectiveness of this com-bination. In addition, further defining these converging pathwaysand the downstream common pathways will reveal new ther-apeutic targets and new biomarkers for LAM and other ER-sensitive cancers.

Materials and MethodsCells derived from renal angiomyolipomas from LAM patients (621-101 TSC2-null cells) were used and maintained in IIA complete medium with 10% (vol/vol) FBS as previously described (30). Before stimulant treatment, LAM cellswere starved overnight in phenol red-, EGF-, and FBS-free IIA completemedium. For inhibition, cells were treated for 1 h with the MEK1/2 inhibitorAZD 6244 at 5 μM, the S6K1 inhibitor PF4708671 at 30 μM, or the mTORC1inhibitor rapamycin at 20 nM before stimulation with either EGF at 10–20ng/mL or 17 β-Estradiol (E2) at 10 nM. For infection, LAM cells were infectedwith either retroviruses or lentiviruses to overexpress or knock down specificgenes of interest using 2 μg/mL puromycin as selection. LAM cells were lysedat different time points for Western blot. To measure migration and in-vasion, Transwell-based assays (Life Technologies-Invitrogen) were used.Cells that migrated to or invaded the other side of the insert membraneswere fixed and quantified. SYBR Green-based RT-PCR was used to detectmRNA levels in LAM cells under different conditions. Full methods areavailable in SI Materials and Methods.

ACKNOWLEDGMENTS. We thank members of the J.B. laboratory for helpfulcomments and discussions. This research was funded by awards from LAMFoundation (to J.B., E.P.H., J.J.Y., and X.G.), the Adler Foundation (E.P.H. andJ.J.Y.), the LAM Treatment Alliance (E.P.H.), the National Institute of Diabe-tes and Digestive and Kidney Diseases (E.P.H.), National Institutes of HealthGrants CA046595 and GM51405 (to J.B.), and by National Heart Lung andBlood Institute Grant HL098216 (to J.J.Y.).

1. Bittmann I, Rolf B, Amann G, Löhrs U (2003) Recurrence of lymphangioleiomyoma-tosis after single lung transplantation: New insights into pathogenesis. Hum Pathol34(1):95–98.

2. Karbowniczek M, et al. (2003) Recurrent lymphangiomyomatosis after transplantation:Genetic analyses reveal a metastatic mechanism. Am J Respir Crit Care Med 167(7):976–982.

3. Sato T, et al. (2002) Mutation analysis of the TSC1 and TSC2 genes in Japanese pa-tients with pulmonary lymphangioleiomyomatosis. J Hum Genet 47(1):20–28.

4. Yu J, Astrinidis A, Henske EP (2001) Chromosome 16 loss of heterozygosity in tuberoussclerosis and sporadic lymphangiomyomatosis. Am J Respir Crit Care Med 164(8 Pt 1):1537–1540.

5. Yu J, Henske EP (2010) mTOR activation, lymphangiogenesis, and estrogen-mediatedcell survival: The “perfect storm” of pro-metastatic factors in LAM pathogenesis.Lymphat Res Biol 8(1):43–49.

6. Henske EP, McCormack FX (2012) Lymphangioleiomyomatosis - a wolf in sheep’sclothing. J Clin Invest 122(11):3807–3816.

7. Harari S, Torre O, Moss J (2011) Lymphangioleiomyomatosis: What do we know andwhat are we looking for? European Respiratory Review: An Official Journal of theEuropean Respiratory Society 20(119):34–44.

8. Matsui K, et al. (2000) Hyperplasia of type II pneumocytes in pulmonary lym-phangioleiomyomatosis. Arch Pathol Lab Med 124(11):1642–1648.

9. Yu J, Astrinidis A, Howard S, Henske EP (2004) Estradiol and tamoxifen stimulateLAM-associated angiomyolipoma cell growth and activate both genomic and non-genomic signaling pathways. Am J Physiol Lung Cell Mol Physiol 286(4):L694–L700.

10. Carsillo T, Astrinidis A, Henske EP (2000) Mutations in the tuberous sclerosis complexgene TSC2 are a cause of sporadic pulmonary lymphangioleiomyomatosis. Proc NatlAcad Sci USA 97(11):6085–6090.

11. Strizheva GD, et al. (2001) The spectrum of mutations in TSC1 and TSC2 in womenwith tuberous sclerosis and lymphangiomyomatosis. Am J Respir Crit Care Med 163(1):253–258.

12. McCormack FX, et al.; National Institutes of Health Rare Lung Diseases Consortium;MILES Trial Group (2011) Efficacy and safety of sirolimus in lymphangioleiomyoma-tosis. N Engl J Med 364(17):1595–1606.

13. Taillé C, Borie R, Crestani B (2011) Current management of lymphangioleiomyoma-tosis. Curr Opin Pulm Med 17(5):374–378.

14. Safe S, Kim K (2008) Non-classical genomic estrogen receptor (ER)/specificity proteinand ER/activating protein-1 signaling pathways. J Mol Endocrinol 41(5):263–275.

15. CohenMM, Freyer AM, Johnson SR (2009) Pregnancy experiences among women withlymphangioleiomyomatosis. Respir Med 103(5):766–772.

16. De Pauw RA, et al. (2003) Renal angiomyolipoma in association with pulmonarylymphangioleiomyomatosis. Am J Kidney Dis 41(4):877–883.

17. Karbowniczek M, Robertson GP, Henske EP (2006) Rheb inhibits C-raf activity andB-raf/C-raf heterodimerization. J Biol Chem 281(35):25447–25456.

18. Yu JJ, et al. (2009) Estrogen promotes the survival and pulmonary metastasis of tu-berin-null cells. Proc Natl Acad Sci USA 106(8):2635–2640.

19. Murphy LO, MacKeigan JP, Blenis J (2004) A network of immediate early geneproducts propagates subtle differences in mitogen-activated protein kinase signalamplitude and duration. Mol Cell Biol 24(1):144–153.

20. Murphy LO, Smith S, Chen RH, Fingar DC, Blenis J (2002) Molecular interpretation ofERK signal duration by immediate early gene products. Nat Cell Biol 4(8):556–564.

21. Murphy LO, Blenis J (2006) MAPK signal specificity: The right place at the right time.Trends Biochem Sci 31(5):268–275.

22. Shin S, Dimitri CA, Yoon SO, Dowdle W, Blenis J (2010) ERK2 but not ERK1 inducesepithelial-to-mesenchymal transformation via DEF motif-dependent signaling events.Mol Cell 38(1):114–127.

23. Ma XM, Blenis J (2009) Molecular mechanisms of mTOR-mediated translational con-trol. Nat Rev Mol Cell Biol 10(5):307–318.

24. Holz MK, Ballif BA, Gygi SP, Blenis J (2005) mTOR and S6K1 mediate assembly of thetranslation preinitiation complex through dynamic protein interchange and orderedphosphorylation events. Cell 123(4):569–580.

25. Shin S, Blenis J (2010) ERK2/Fra1/ZEB pathway induces epithelial-to-mesenchymaltransition. Cell Cycle 9(13):2483–2484.

26. Vial E, Marshall CJ (2003) Elevated ERK-MAP kinase activity protects the FOS familymember FRA-1 against proteasomal degradation in colon carcinoma cells. J Cell Sci116(Pt 24):4957–4963.

27. Casanova A, Ancochea J (2011) [Lymphangioleiomyomatosis: New therapeutic ap-proaches]. Arch Bronconeumol 47(12):579–580.

28. Mackeigan JP, Murphy LO, Dimitri CA, Blenis J (2005) Graded mitogen-activatedprotein kinase activity precedes switch-like c-Fos induction in mammalian cells. MolCell Biol 25(11):4676–4682.

29. Mendoza MC, Er EE, Blenis J (2011) The Ras-ERK and PI3K-mTOR pathways: Cross-talkand compensation. Trends Biochem Sci 36(6):320–328.

30. Arbiser JL, et al. (2001) The generation and characterization of a cell line derived froma sporadic renal angiomyolipoma: Use of telomerase to obtain stable populations ofcells from benign neoplasms. Am J Pathol 159(2):483–491.

6 of 6 | www.pnas.org/cgi/doi/10.1073/pnas.1309110110 Gu et al.

Dow

nloa

ded

by g

uest

on

July

30,

202

0