Stoichiometry and assembly of mTOR complexesrevealed by single-molecule pulldownAnkur Jaina,b,1, Edwin Arauzc,1, Vasudha Aggarwala,b, Nikita Ikonc, Jie Chenc,2, and Taekjip Haa,b,d,e,2

aCenter for Biophysics and Computational Biology, bInstitute for Genomic Biology, cDepartment of Cell and Developmental Biology, dDepartment of Physics,and eHoward Hughes Medical Institute, University of Illinois at Urbana–Champaign, Urbana, IL 61801

Edited by Melanie H. Cobb, University of Texas Southwestern Medical Center, Dallas, TX, and approved November 11, 2014 (received for review October11, 2014)

The mammalian target of rapamycin (mTOR) kinase is a masterregulator of cellular, developmental, and metabolic processes. De-regulation of mTOR signaling is implicated in numerous humandiseases including cancer and diabetes. mTOR functions as part ofeither of the two multisubunit complexes, mTORC1 and mTORC2,but molecular details about the assembly and oligomerization ofmTORCs are currently lacking. We use the single-molecule pull-down (SiMPull) assay that combines principles of conventionalpulldown assays with single-molecule fluorescence microscopyto investigate the stoichiometry and assembly of mTORCs. Aftervalidating our approach with mTORC1, confirming a dimeric as-sembly as previously reported, we show that all major compo-nents of mTORC2 exist in two copies per complex, indicating thatmTORC2 assembles as a homodimer. Interestingly, each mTORCcomponent, when free from the complexes, is present as a mono-mer and no single subunit serves as the dimerizing component.Instead, our data suggest that dimerization of mTORCs is the resultof multiple subunits forming a composite surface. SiMPull alsoallowed us to distinguish complex disassembly from stoichiometrychanges. Physiological conditions that abrogate mTOR signalingsuch as nutrient deprivation or energy stress did not alter thestoichiometry of mTORCs. On the other hand, rapamycin treat-ment leads to transient appearance of monomeric mTORC1 beforecomplete disruption of the mTOR–raptor interaction, whereasmTORC2 stoichiometry is unaffected. These insights into assemblyof mTORCs may guide future mechanistic studies and explorationof therapeutic potential.

mTOR | mTORC | single molecule | stoichiometry | rapamycin

The mammalian target of rapamycin (mTOR) is a masterregulator of crucial cellular and developmental processes. As

a serine/threonine protein kinase belonging to the phosphatidy-linositol-3-kinase (PI3K)-related kinase family, mTOR integra-tes the sensing of nutrients, growth factors, oxygen, energy, anddifferent types of stress to regulate a myriad of biological pro-cesses such as cell growth, proliferation, differentiation, and me-tabolism (1). mTOR functions as part of at least two biochemicallyand functionally distinct complexes—mTORC1 and mTORC2(2). mTORC1, better characterized of the two complexes, is therapamycin-sensitive complex, composed of the proteins raptor andmLST8, and it is regulated by the inhibitory proteins PRAS40and DEPTOR (2, 3). mTORC1 is activated by nutrients (such asamino acids), growth factors, and cellular energy among otherstimuli (1, 2). mTORC2 contains rictor, mLST8, and mSin, as wellas the negative regulator DEPTOR (2, 3).PI3K-related kinases (PIKKs) such as ataxia telangiectasia

mutated (ATM), ATM and Rad3-related protein (ATR), andDNA-dependent protein kinase (DNA-PK) are known to oli-gomerize (4–6). Biochemical and genetic analyses have identi-fied self-association of mTOR and its orthologs in yeast andDrosophila (7–10). A cryoelectron microscopy (cryo-EM) studyrevealed that mTORC1 self-associates into a dimeric structure(11). Oligomerization of mTORC1 has been reported to be sensi-tive to nutrient status based on biochemical analyses of recombinantproteins (10, 12). Consensus is lacking on the oligomeric state of

mTORC2, which has been proposed to be monomeric, dimeric,or multimeric (7, 10, 13, 14). High-resolution structural analysisof mTORC2 has not been possible thus far, likely owing to itslarge size and multiplicity of interaction partners.Ensemble biochemical methods have inherent limitations in

analyzing multicomponent heterogeneous protein assemblies.These methods do not directly reveal the stoichiometry of in-teraction and offer low-resolution estimates of the sizes of pro-tein complexes. Additionally, the lengthy procedures oftenassociated with biochemical characterization may lead to loss oralteration of physiological protein complexes. We recently reporteda single-molecule pulldown (SiMPull) technology that combinesthe principles of conventional pulldown assays with single-moleculefluorescence microscopy (15). In SiMPull, protein complexes arepulled down from freshly lysed cells directly onto chambers forsingle-molecule fluorescence microscopy. When proteins are stoi-chiometrically labeled for example using fluorescent protein tags,SiMPull can reveal the stoichiometry of the protein complexes viasingle-molecule fluorescence photobleaching step analysis (15).We have used SiMPull to investigate the oligomeric assembly

of mTORCs. Upon validating our approach by demonstratingdimeric assembly of mTORC1, we find that mTORC2 is alsodimeric and contains two molecules of mTOR and rictor percomplex. Individual mTORC components are predominantlymonomeric, but under physiological conditions there is no evi-dence of monomeric interaction between mTOR and raptor orrictor. Multicolor imaging of individual complexes revealed thatalthough the two complexes are predominantly distinct, smallfractions of mTORC1 and mTORC2 components coexist in thesame complex. Physiological perturbations that abrogate mTORsignaling had no effect on the stoichiometry of mTOR com-plexes, indicating that inhibition of mTOR signaling can be

Significance

The mammalian target of rapamycin (mTOR) kinase is a centralregulator of cell growth, differentiation, and metabolism.mTOR is assembled into two distinct complexes, mTORC1 andmTORC2. Using single-molecule pulldown (SiMPull), we havedetermined the stoichiometric composition of mTORCs undergrowth and stress conditions. We find that both mTORC1 andmTORC2 form obligate dimers, in which major componentsexist in two copies per complex. Importantly, SiMPull allowedus to distinguish complex disassembly from stoichiometrychanges, providing insights into the effects of physiologicalconditions and the drug rapamycin on mTOR complexes.

Author contributions: A.J., E.A., J.C., and T.H. designed research; A.J., E.A., V.A., and N.I.performed research; A.J. and E.A. contributed new reagents/analytic tools; A.J., E.A., andV.A. analyzed data; and A.J., E.A., J.C., and T.H. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1A.J. and E.A. contributed equally to this work.2To whom correspondence may be addressed. Email: jiechen@illinois.edu or tjha@illinois.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1419425111/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1419425111 PNAS | December 16, 2014 | vol. 111 | no. 50 | 17833–17838

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

Oct

ober

16,

202

1

achieved without requiring disassembly of mTOR complexes orchanging their oligomeric state. On the other hand, treatment withrapamycin led to transient mTOR–raptor complexes containing onemTOR before complete disassembly of the interaction, whereasmTORC2 stoichiometry was unaffected.

ResultsAssay Validation and mTORC1 Stoichiometry. To study mTORcomplexes by SiMPull, we deemed it important to establisha system where a fluorescently tagged mTOR can incorporateinto endogenous complexes. To that end, we established a cellline stably expressing YFP–mTOR in which the endogenousmTOR was silenced by short hairpin RNA (Fig. 1A, henceforthcalled “YFP–mTOR stable cell line”). The YFP–mTOR proteinassociated with endogenous raptor and rictor (Fig. 1A). Moreimportantly, the cell line faithfully recapitulated known regu-lations of mTOR signaling, such as insulin- and serum-stimu-lated phosphorylation of mTORC1 targets S6K1 and 4E-BP1,and mTORC2 target Akt, as well as amino acid dependence ofS6K1 and 4E-BP1 phosphorylation (Fig. 1B and Fig. S1A). Whencell lysates from this line were applied to single-molecule im-aging chambers coated with an antibody against raptor, YFP–mTOR fluorescence spots were observed, well above the back-ground level of fluorescence seen in the control channel withoutthe antibody (Fig. 1 C and D), illustrating specific pulldown ofmTOR in complex with raptor, or mTORC1.To assess the sensitivity of SiMPull, we compared detection of

YFP–mTOR by Western blotting and by SiMPull with celllysates at the same concentrations. Remarkably, even at 1,000-fold dilution of the lysates SiMPull detected YFP–mTOR spe-cifically (Fig. S2A), whereas at 100-fold dilution of the samelysates, there was no longer any signal onWestern blots (Fig. S2B).Furthermore, in SiMPull, we were able to detect YFP–mTORpulled down via endogenous raptor using a 100-fold dilution of thelysate (Fig. S2 C and D), whereas no signal was detected in con-ventional coimmunoprecipitation using the same dilution of lysate

and raptor antibody (Fig. S2E); SiMPull required only 50 μL ofthe extract as opposed to 500 μL used for the correspondingcoimmunoprecipitation. Hence, the SiMPull method is highlysensitive compared with conventional biochemical methods.We then analyzed the fluorescence time trajectories of YFP–

mTOR pulled down with raptor. Most molecules (96%) bleachedin either one or two steps, indicating that mTORC1 contains oneor two molecules of fluorescently active YFP–mTOR (Fig. 1E).Nearly 60% of the molecules exhibited two-step bleaching,whereas 36% bleached in a single step. The molecules bleachingin two steps were nearly twice as bright as one-step bleacherson average, indicating a reliable classification based on photo-bleaching steps (Fig. 1E). Previous studies have determined thatfluorescent proteins may not all mature to completion and thefraction of fluorescently active YFP is ∼75% (15, 16). In a cali-bration experiment, photobleaching step distribution of mono-meric and dimeric YFP or a mixture of the two proteins wasmeasured, which revealed that the fraction of two-step photo-bleaching spots is linearly proportional to the fraction of dimericYFP included (Fig. 1 H and I). A comparison of the observedphotobleaching step distribution of YFP–mTOR in mTORC1 tothe calibration data suggested that nearly all mTORC1 complexes(>95%) contain two copies of YFP–mTOR, assuming the mat-uration level of YFP for all experiments is the same.We also transiently coexpressed YFP–mTOR and HA–raptor,

which assembled into functional mTORC1 complexes (Fig. S1B).Single-molecule fluorescence photobleaching analysis for YFP–mTOR pulled down with HA–raptor revealed that the complexeseach contained two copies of YFP–mTOR (Fig. S3 A and B).Similarly, when YFP–raptor and Flag–mTOR were coexpressed(Fig. S1 C and D), two copies of YFP–raptor were found in eachmTOR–raptor complex (Fig. 1 F and G). Additionally, themTORC1 inhibitors PRAS40 and DEPTOR were also present intwo copies per mTORC1 (Figs. S1 E–G and S3 C–F).To confirm that the complexes captured via SiMPull represent

physiological state of mTORC1 and are not assembled/dis-assembled upon cell lysis, we followed the strategy of Riley et al.

0 4 80

150

300

1 2 3 40

50

100 tdYFP only

Raptor

YFP-mTOR

Anti-RabbitNeutrAvidin PEG

Quartz slide

Biotin

Anti-Raptor

IP Raptor -0

300

Nf

IP Raptor -

YFP

RaptorYFP-mTOR

Anti-Rabbit

Anti-Raptor

A

D

FG

E

HA-YFP-Raptor

Flag-mTOR

Anti-Flag

0

200

Nf

IP Flag T7IP

YFP

Flag T71 2 3 4

0

20

40

60

%N

f

Steps 0 400 8000

50

100

Nf

Intensity (a.u.)

1 2 3 40

20

40

60

%N

f

Steps 0 200 400 6000

40

80

Nf

Intensity (a.u.)

pS6K1 T389

FBS

pAKT S473

S6K1

AA +-

+ -++

+ Insulin-

- -

-+

+-

-

AKT

p4EBP1 T37/46

4EBP1Raptor

Rictor

Lysate mTOR IP

mTORYFP-mTOR

293 YM 293 YM

293

mTORYFP-mTOR

YFP-mTOR

shRNA: Scr mTOR -

B C

H

% N

f

Steps 1 2 3 40

50

100 55% mYFP45% tdYFP

1 2 3 40

50

100 29% mYFP71% tdYFP

79% mYFP21% tdYFP

1 2 3 40

50

100mYFP only

1 2 3 40

50

100

0 4 80

100

200

0 4 80

150

300

0 4 80

125

250

0 4 80

125

250

1-step2-stepDiscarded

Int.

Nf

15 30 45 600

50

100

%td

YFP

% 2-step

y = 2.15x - 32R2= 0.997

I

N = 2939

N = 2840

Fig. 1. mTORC1 is dimeric. (A) Expression of mTORin the YFP–mTOR stable cells (YM) (Upper) and in-teraction with raptor and rictor assessed by co-IP(Lower) are shown. Scr, scramble. (B) YM cells werestimulated with insulin, serum, or amino acids (AA).(C) Schematic depiction of mTORC1 SiMPull. (D)Endogenous raptor was pulled down from YM cells,followed by SiMPull analysis. Shown are schematicdiagram (Left), representative YFP fluorescenceimages (Center), and average number of moleculesper imaging area (Nf). (E) Distribution of fluorescencephotobleaching steps (Left) and corresponding in-tensity (a.u., arbitrary units) distribution (Right) forsamples described in D. N, total number of moleculesanalyzed. (F) Flag–mTOR and HA–YFP–raptor werecoexpressed and Flag–mTOR was pulled down, fol-lowed by analyses similar to those in D. (G) Sameanalyses as in E for samples described in F. (Scale bar,5 μm.) (H) Monomeric YFP (mYFP) and tandem di-meric YFP (tdYFP) were mixed in the ratios as in-dicated and single-molecule fluorescence timetrajectories were analyzed. (Top) Number of fluo-rescence photobleaching steps. (Bottom) Intensity(Int., arbitrary units) of molecules exhibiting one ortwo photobleaching steps. (I) Observed fraction oftwo-step bleaching events changes linearly with thefraction of tdYFP.

17834 | www.pnas.org/cgi/doi/10.1073/pnas.1419425111 Jain et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

16,

202

1

(17) and mixed lysates from cells expressing YFP–raptor andFlag–mTOR separately, and compared them to lysates of cellscoexpressing the same proteins at similar expression levels (Fig.S3G). At least 10-fold more mTORC1 complexes were detectedin the lysates from cotransfected cells than the mixed lysates(Fig. S3H). Similar results were obtained when intact cellswere mixed before lysis (Fig. S3I). Thus, postlysis associationof mTORC1 components does not significantly contribute tothe complexes detected here via SiMPull.SiMPull requires dilution of cell lysates to achieve low pull-

down density suitable for single-molecule analysis, which couldlead to loss of weakly associated physiological complexes. Toaddress this issue, we treated YFP–mTOR-expressing cells withthe cross-linker dithiobis(succinimidylpropionate) (DSP) beforecell lysis. Lysis of cells using a Triton X-100–containing bufferdisrupted mTORC1 complexes as expected (18), whereas DSPcross-linking before cell lysis preserved the complex under the samelysis condition (Fig. S4 A and B), validating our cross-linking con-ditions. Importantly, cross-linked mTORC1 complexes exhibiteda photobleaching step distribution corresponding to that of a di-mer (Fig. S4C), suggesting that the physiological complexes areintact during SiMPull analysis without cross-linking. In summary,we find that mTORC1 is dimeric, containing two copies of eachcomponent. These findings are consistent with the previous cryo-EM data (11), thus validating our experimental system for anal-ysis of mTOR complexes by SiMPull.

mTORC2 Is Dimeric.mTORC2 assembly requires mSin and mLST8(14, 19) in addition to mTOR and rictor, but the oligomeric stateof mTORC2 is under debate (7, 10, 13, 14). When all four corecomponents of mTORC2 (mTOR, rictor, mSin, and mLST8) werecoexpressed, each component self-associated as indicated bycoimmunoprecipitation of the recombinant protein with two dif-ferent tags (Fig. 2A and Fig. S5A), suggesting that assembledmTORC2 is oligomeric. To determine the oligomeric state ofmTOR in mTORC2, we captured mTORC2 complexes from theYFP–mTOR stable cell line on SiMPull surfaces using an antibodyagainst endogenous rictor (Fig. 2B). Intriguingly, once again weobserved a photobleaching pattern characteristic for dimericYFP–mTOR (Fig. 2C). Similar results were obtained when YFP–mTOR was coexpressed with recombinant mTORC2 compo-nents (Fig. S6 A and B), which also assembled into functionalmTORC2 (Fig. S5 B and C).Next, we probed the stoichiometry of rictor in mTORC2. YFP–

rictor was coexpressed with Flag–mTOR, mSin–HA and HA–mLST8, and the assembly of mTORC2 was verified by ensemblepulldown (Fig. S5 D and E). Upon single-molecule pulldown ofmTORC2 via Flag–mTOR (Fig. 2D), we observed YFP–rictorphotobleaching consistent with two copies of rictor per mTORC2

(Fig. 2E). Furthermore, YFP–DEPTOR in mTORC2 was foundto display photobleaching distribution that corresponded to a mix-ture of monomers and dimers (∼60% dimer) (Fig. S6 C and D),indicating that each mTORC2 can harbor up to two copiesof DEPTOR.Once again, cell mixing (Fig. S6 E and F) and DSP cross-

linking (Fig. S6 G and H) experiments were performed, theresults of which confirmed that the dimeric stoichiometry mostlikely reflects physiological assembly of mTORC2 in live cells. Inconclusion, our results unequivocally reveal mTORC2 as a di-mer, in which each subunit is present in two copies.

mTORC1 and mTORC2 Are Mostly Distinct but Cocomplexes Exist.Biochemical characterizations so far suggest that mTORC1and mTORC2 are mutually exclusive, but functional cross-talkbetween these two complexes at multiple levels is increasinglyapparent (20–22). Even a small fraction of mTORC1/mTORC2cocomplex could be functionally significant, but may have es-caped detection by conventional biochemical methods. Multi-color SiMPull should allow direct visualization of such hybridcomplexes, if any. As a positive control, we coexpressed Flag–mTOR, mCherry–raptor, and YFP–PRAS40. Upon capturingFlag–mTOR, we observed both mCherry–raptor and YFP–PRAS40 fluorescence spots (Fig. 3A); ∼49% of YFP–PRAS40spots colocalized with mCherry–raptor, indicating their coexistencein the same complexes. Incomplete colocalization between the twolikely arises from incomplete chromophore maturation (∼40% formCherry and ∼75% for YFP) (23, 24) and the participation ofendogenous untagged proteins.Next, we coexpressed mCherry–raptor and YFP–rictor to-

gether with Flag–mTOR, mLST8, and mSin. An anti-Flag anti-body captured both mTORC1 and mTORC2 as visualized bymCherry and YFP fluorescent spots, respectively, and about 7%of mCherry–raptor spots reproducibly colocalized with YFP–rictor (Fig. 3B). Taken into consideration the incomplete chro-mophore maturation mentioned above, it is likely that the truefraction of the mTORC1/mTORC2 cocomplex is higher than7%. Under these experimental conditions, the colocalization bychance was ∼1% (Fig. S7A). Additionally, when YFP instead ofYFP–rictor was expressed, only a background level of YFP fluo-rescent spots was observed upon Flag–mTOR pulldown, and thesespots did not colocalize with mCherry–raptor (Fig. S7B). Impor-tantly, raptor coexpression did not alter mTORC2 stoichiometry,as rictor was still present in two copies (Fig. S7C).To verify that the hybrid complex was not an artifact of cell

lysis, we performed cell mixing experiments. mTORC1 compo-nents (Flag–mTOR and mCherry–raptor) and mTORC2 com-ponents (Flag–mTOR, YFP–rictor, mSin–HA, and HA–mLST8)were expressed in two separate pools of cells, which were mixed

Flag-mTOR

Anti-Flag

YFP-Rictor

mLST8

mSin

0

250

500

Nf

IPmTOR

Flag Flag-+ + -

IPmTOR

YFP

Flag+

-+

Flag-

Anti-Rabbit

Anti-Rictor

YFP-mTOR

Rictor

IP

YFP

Rictor - IP Rictor -0

200

400

Nf

D

B

E

mSin-Flag

+Flag-Rictor +-

HA-mTOR

+Flag-mLST8

++

Flag-mTOR+

+--

Lysate HA IP

++-+

++ +

+--

HA-mTOR

Flag-mTOR

A C

1 2 3 40

20

40

60

%N

f

Steps

1 2 3 40

20

40

60

%N

f

Steps 0 200 400 600 8000

50

100

Nf

Intensity (a.u.)

0 250 5000

40

80

Nf

Intensity (a.u.)

N = 2290

= 2588N

Fig. 2. mTORC2 is dimeric. (A) mTOR oligomerizes in mTORC2 shown by co-IP. (B and C) SiMPull for endogenous rictor from YM cells and analysis similar toFig. 1 D and E. (D and E) YFP–rictor was coexpressed with Flag–mTOR, mSin, and mLST8, and Flag–mTOR was pulled down. SiMPull data are presented as in Band C. (Scale bar, 5 μm.)

Jain et al. PNAS | December 16, 2014 | vol. 111 | no. 50 | 17835

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

Oct

ober

16,

202

1

during lysis (Fig. S7 D–F). After accounting for false colocalizationby chance, only 2% colocalization between mCherry–raptor andYFP–rictor was observed from mixed lysates. Therefore, althoughthe two mTOR complexes are predominantly distinct, hybrid mTORcomplexes containing both raptor and rictor or higher order as-semblies of mTORC1/mTORC2 exist, albeit at a low level.

mTORC1 and mTORC2 Components Are Monomeric. Because bothmTORC1 and mTORC2 are dimeric, we asked if mTOR orother core components could self-dimerize. To that end, eachcomponent tagged with YFP was individually expressed. WhenYFP–mTOR was captured using an anti-mTOR antibody, nearly75% of the molecules bleached in a single step, whereas 20%bleached in two steps, indicating that a majority of overexpressedmTOR was monomeric (Fig. 4A). Furthermore, when YFP–mTORand HA–mLST8 were coexpressed and the mTOR–mLST8 sub-complexes were captured through the HA tag, a photobleachingstep distribution characteristic of monomers was observed for YFP–mTOR (Fig. 4B). Similar analysis for HA–YFP–raptor and HA–YFP–rictor, pulled down through the HA tag, also revealed mo-nomeric distributions (Fig. 4 C and D). In addition, YFP–PRAS40bound to raptor was monomeric (Fig. S8 A and B). The observedsmall fractions of dimer may arise due to incorporation of YFP-tagged proteins in endogenous mTOR complexes. Taken to-gether, our results show that although both mTORC1 andmTORC2 are exclusively dimeric, individual mTORC compo-nents and subcomplexes are predominantly monomeric. Thus,no single mTORC subunit serves as a dimerizing component.

mTORC Stoichiometry Is Unchanged Under Various PhysiologicalConditions. Next we asked if the oligomerization of mTORcomplexes was affected by upstream signals or physiologicalconditions known to regulate mTOR signaling, including growthfactors, nutrient availability, and cellular energy levels. To ex-amine the effect of energy sufficiency, we starved the YFP–mTOR stable cells of glucose and glutamine, which leads toenergy depletion and inhibition of mTORC1 signaling (12), andbriefly (1 h) restimulated them with growth medium containingglucose and glutamine. As shown in Fig. 5A, a nearly equal numberof mTORC1 complexes pulled down through endogenous raptorwere detected by SiMPull, with similar photobleaching step dis-tribution, under starvation and stimulation conditions. Similarly,neither amino acids (Fig. 5B) nor leucine (Fig. S8 C and D) stim-ulation had any effect on the number or stoichiometry ofmTORC1. In addition, insulin stimulation did not affect mTORC1(Fig. 5C) or mTORC2 assembly (Fig. 5D). The starvation andstimulation conditions impacted mTOR signaling as expected (Fig.1B and Fig. S1A). These observations directly establish that in-hibition of mTOR activity by energy stress or nutrient- or growth-factor depletion can be achieved without disassembly of themTOR complexes.

Effect of Rapamycin on mTOR Complexes. Several models have beenproposed for the mechanism by which rapamycin inhibits mTOR.For example, rapamycin may limit substrate access to the kinasedomain (25, 26), may induce raptor dissociation from mTOR(11, 18, 27), or displace a key regulator (28). We investigated theeffect of rapamycin on the stoichiometry of mTOR complexes. Asreported (29), short-term treatment of cells with rapamycin dis-sociated raptor from mTOR and inhibited mTORC1 signaling,whereas on prolonged treatment, mTORC2 assembly was alsoaffected (Fig. S9 A and B). When YFP–mTOR stable cells weretreated with increasing concentrations of rapamycin (2–100 nM)for 30 min and mTORC1 complexes were captured from celllysates through endogenous raptor, the number of YFP–mTORpulled down decreased with increasing rapamycin dose (Fig. 6A):treatment with 2 nM of rapamycin led to reduction by 56%, anda maximal reduction of ∼90% was reached by 10 nM and 100 nMrapamycin (Fig. 6C). The residual mTORC1 appeared to beresistant to rapamycin, consistent with previously reportedobservations (18). Thus, acute rapamycin treatment at low dosedisrupted the interaction between mTOR and raptor.Additionally, we found that the fraction of dimers decreased

as the rapamycin dose was increased. At 2 nM of rapamycin,about 47% of the molecules bleached in two steps, whereas at100 nM, only 29% of the molecules exhibited two-step bleaching(Fig. 6B). The photobleaching analysis was performed at similarsurface densities of the complexes (∼0.1 molecules·μm−2), pre-cluding any artifacts due to differences in the immobilizationdensity. This observation of transient monomeric mTOR–raptorcomplexes indicates that rapamycin disrupts mTORC1 in at leasttwo steps, displacing one mTOR (or mTOR–raptor monomericsubcomplex) at a time. Consistent with these results, mTORbound to FKBP12–rapamycin is monomeric, as indicated by thesingle-step photobleaching exhibited by YFP–mTOR pulleddown via surface immobilized FKBP12–rapamycin (Fig. S8 Eand F). The reduction in mTORC1 signaling activity (pS6K1)corresponded with the loss of intact dimeric mTORC1 com-plexes (Fig. 6C), implying that disruption of mTORC1 dimericarchitecture contributes significantly to rapamycin-induced ab-rogation of mTORC1 signaling.Next, we investigated the effect of prolonged rapamycin treat-

ment on mTORC2 assembly by capturing endogenous rictor fromYFP–mTOR cell lysates. The number of mTORC2 complexesdecreased by 40% upon 6-h treatment with 100 nM of rapamycinand was reduced by 79% after 24 h (Fig. 6D). Interestingly, incontrast to the transient mTORC1 monomers, mTORC2remained a dimer at all time points (Fig. 6E). Rapamycin treat-ment did not affect mTOR expression levels (Fig. S9 C and D),whereas the amount of mTORC1 decreased (Fig. S9E). Thus, our

mCherry-RaptorFlag-mTORAnti-Flag

YFP-PRAS40

mCherry-Raptor

Flag-mTOR

Anti-Flag

A

B

YFP mCherry Overlay 49 ± 2%

7 ± 3%YFP-Rictor

mLST8

mSin

Fig. 3. mTORC1 and mTORC2 colocalization. (A) Flag–mTOR cocapturesYFP–PRAS40 (Left) and mCherry–raptor (Center). Overlay of the two images(Right) showed 49 ± 2% colocalization. (B) Flag–mTOR, mCherry–raptor,YFP–rictor, HA–mLST8, and mSin–HA were coexpressed. Flag–mTOR waspulled down, and YFP (rictor, Left) and mCherry (raptor, Center) were im-aged. Overlay of the two images (Right) showed 7 ± 3% colocalization acrosssix independent experiments. (Scale bar, 10 μm.)

YFP-mTOR

Anti-IgG

Anti-mTOR

A

HA-YFP-Raptor

Anti-HAHA

C

0

500

Nf

IP HA Flag 1 2 3 40

100

%N

f

N = 2232HA-YFP-Rictor

Anti-HA

D

0

200

Nf

IP HA Flag 1 2 3 40

100

%N

f

Steps

N = 1921

Anti-HA

YFP-mTOR

HA-mLST8

IP0

300

HA IgG

Nf

1 2 3 40

100

%N

f

N = 1073B

IP mTOR0

300

Nf

V5 1 2 3 40

100

%N

f

N = 2723

Steps

Fig. 4. mTORC components are monomeric. (A) YFP–mTOR, (B) YFP–mTORand HA–mLST8, (C) HA–YFP-raptor, or (D) HA–YFP–rictor were expressed inHEK293 cells and captured on SiMPull surfaces as depicted (diagrams onLeft). The number of molecules observed per imaging area (Center) and thedistribution of fluorescence photobleaching steps (Right) are shown.

17836 | www.pnas.org/cgi/doi/10.1073/pnas.1419425111 Jain et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

16,

202

1

results suggest that rapamycin does not directly affect mTORC2stoichiometry. Instead, rapamycin may sequester free mTOR andsubsequently impair mTORC2 assembly over time.

DiscussionUsing SiMPull we have determined the stoichiometry and as-sembly of mTOR complexes captured from whole cell lysates.In addition to confirming the dimeric structure of mTORC1as previously revealed by cryo-EM studies, we find thatmTORC2 assembles into a dimer, with two copies of each

subunit. Interestingly, all individual subunits of mTORCs in-cluding mTOR, when expressed alone, are monomeric. Thisexcludes a commonly presumed role of the HEAT repeats inmediating mTOR self-association. Interaction between mTORand raptor alone is sufficient to form mTORC1 dimers, butmTORC2 assembly and dimerization require coexpression ofmSin and mLST8 in addition to mTOR and rictor. Underphysiological conditions, there is no evidence of monomeric in-teraction between mTOR and raptor or mTOR and rictor.Hence, we propose a model for mTORC assembly where theinteraction between monomeric subunits is unstable; assemblyrequires multiple subunits to accumulate at high local concen-trations (Fig. 6F), which may be facilitated by membrane local-ization, subcellular compartmentalization, or scaffoldingproteins. The small but significant fraction of mTORC1–mTORC2 cocomplex revealed by SiMPull suggests a potentialphysical cross-talk between the two complexes that may haveevaded detection by conventional biochemical methods. Futureinvestigation examining the biological function of this cocomplexis warranted.Of significance, our SiMPull assays have captured the exis-

tence of monomeric mTORC1 (mTOR–raptor complex) uponacute rapamycin treatment, before complete disruption of themTOR–raptor interaction (Fig. 6F). The capture of this in-termediate state of mTORC1, which was not detected by cryo-EM analysis (11), further attests to the exquisite sensitivity of theSiMPull method. We also provide direct evidence that cellularstress conditions that abrogate mTORC1- or mTORC2-medi-ated signaling do not alter the number or oligomerization stateof mTOR complexes, indicating that effective inhibition ofmTOR signaling can be achieved without disassembling mTORcomplexes. Of note, Kim et al. recently reported that glucose andglutamine deprivation results in monomeric mTORC1 (12), which

A

0

125

250

Nf

α-RaptorAA

+-

++

--

% N

f

Steps1 2 3 4

0

60

1 2 3 4

N = 437- AA

N = 411+ AA

% N

f

Steps1 2 3 4

0

70

1 2 3 4

N = 1238- Insulin

N = 561+ Insulin

0

150

300

450

Nf

α-RictorInsulin

+-

++

--

D

0

125

250

Nf

α-RaptorGluc/gln

+-

++

--

% N

f

Steps1 2 3 4

0

60

1 2 3 4

N = 357- Gluc/gln

N = 414+ Gluc/gln

C 300

0

150Nf

α-RaptorInsulin

+-

++

--

% N

f

Steps1 2 3 4

0

60

1 2 3 4

N = 480- Insulin

N = 604+ Insulin

B

Fig. 5. Effect of physiological stimulations on mTORCs. YFP–mTOR stablecells were starved (−) of (A) glucose and glutamine (Gluc/gln) for 12 h, (B)amino acids (AA) for 2 h, or (C and D) serum for 24 h, followed by restim-ulation (+) with glucose/glutamine for 1 h, amino acids for 30 min, or 100 nMinsulin for 30 min, respectively. mTORC1 (A–C) or mTORC2 (D) was pulled downvia endogenous raptor or rictor, respectively, followed by SiMPull analysis.

IP: R

icto

r

Rap TPT (h) 0 6 12 24D

E

Rap (nM) 0 2 10 100

IP: R

apto

r

0

350

700

Nf

Rap (nM) 0 2 10 100 Con

A

BC

1 2 3 40

75

1 2 3 40

25

50

75

% N

f

Steps 1 2 3 40

75

1 2 3 40

75

0 4 80

80

0 4 80

70

0 4 80

60

1 2 3 40

80

1 2 3 40

80

1 2 3 40

80

0 4 80

150

0 4 80

150

0 4 80

75

0

175

Nf

0 4 8Int.

1 2 3 40

40

80

% N

f

Steps

0 4 80

50

Int.

Nf

N = 2971 N = 2805 N = 2274 N = 1367

N = 1348 N = 1850 N = 1303 N = 1553

Con0

300

600

Nf

Rap (h) 0 6 12 24

mTOR

mTORC1

mTORC2

Rictor

mSin

mLST8

RaptorRapamycin-FKBP12

mTORmTOR-raptor (total)

mTOR-raptor (dimer)pS6K1 T389

0 5 10 1000

50

100

150

%re

lativ

eto

0nM

Rap (nM)0 5 10 100

F

1-step2-step

Discarded

Fig. 6. Effect of rapamycin on mTORCs. YFP–mTOR stable cells were treated with (A–C) in-creasing doses of rapamycin (Rap) for 30 min or (Dand E) 100 nM rapamycin over the indicated timecourse. mTORC1 (A and B) and mTORC2 (D and E)were captured via endogenous raptor and rictor,respectively. Representative SiMPull images andnumber of molecules observed per imaging areaare shown in A and D. (Scale bar, 5 μm.) Dis-tributions of fluorescence photobleaching stepsare shown in B and E. N, total number of mole-cules analyzed. (C, Left) Relative levels of mTOR,mTOR–raptor (total), and mTOR–raptor (dimer)were obtained from SiMPull. (Right) Phosphory-lation of S6K1 (pT389) was measured by Westernblotting under similar conditions and quantifiedby densitometry. TPT, time posttreatment. (F) Amodel for the assembly of mTOR complexes. In-dividual mTORC components are monomericand assemble into homodimeric holocomplexes,mTORC1 or mTORC2. Rapamycin directly disruptsthe mTOR–raptor interaction leading to mono-meric mTORC1 and single proteins. Rapamycin–FKBP12-associated mTOR cannot be incorporatedinto mTORC1 or mTORC2, resulting in indirectdepletion of mTORC2 over time.

Jain et al. PNAS | December 16, 2014 | vol. 111 | no. 50 | 17837

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

Oct

ober

16,

202

1

is not observed in our SiMPull assays. Again, direct experimenta-tion with fresh cell lysates containing near-endogenous proteins,without any additional manipulation such as those required forconventional coimmunoprecipitation, may be advantageous incapturing physiologically relevant protein complexes. Dimerizationof other PIKKs such as ATM and DNA-PK play important roles inthe regulation of their activities (4, 6). Our data reveals differencesin assembly mode and regulation of mTOR compared with thesemembers of the PIKK family.Over 75% of the proteins in a cell can oligomerize (30).

SiMPull is a powerful method to investigate oligomeric proteinassemblies. This method provides direct and quantitative read-out of the assembly state of the proteins, expressed at endoge-nous levels directly in their native context. Proteins are capturedfrom freshly lysed cells and probed at single-molecule resolutionwithout requiring removal of other proteins, minimizing loss ofinteractions due to stringent wash steps associated with con-ventional immunoprecipitation. SiMPull worked wherever en-semble immunoprecipitation worked across the variety ofconstructs and samples tested in this study, highlighting theversatility of the assay. By performing SiMPull and biochemicalanalysis on the same samples, we were able to correlate thecomplex architecture with its functional activity. The advent ofgenetic engineering at endogenous loci and developments inshort genetically encoded fluorescent tags should enable pow-erful applications of this technology to near-endogenous systems.

Materials and MethodsAntibodies, Plasmids, and Other Reagents. These are all described in SI Materialsand Methods.

Cell Culture. The maintenance and transfection of HEK293 cells, and the gen-eration of YFP–mTOR stable cells are described in SI Materials and Methods.

Single-Molecule Analyses. Single-molecule imaging and spot counting, pho-tobleaching analysis, and single-molecule colocalization were performed aspreviously reported (15, 31). Detailed procedures are described in SI Mate-rials and Methods.

Western Blotting, Immunoprecipitation, and in Vitro mTOR Kinase Assays.These assays followed previously published procedures (32) and are de-scribed in detail in SI Materials and Methods.

Statistical Analysis. All data are presented as mean ± SD, or representativeimages, of at least three sets of independent experiments. Whenever nec-essary, statistical significance of the data was analyzed by performing one-sample or paired t tests.

ACKNOWLEDGMENTS. We thank Kaushik Ragunathan, Benjamin Leslie, KyuYoung Han, Kyung Suk Lee, Reza Vafabakhsh, and the members of the J.C.laboratory for helpful discussions. This work was supported by NationalScience Foundation Grant PHY 1430124 (to T.H.) and National Institutes ofHealth Grants AI083025 (to T.H.), GM089771 (to J.C.), AR048914 (to J.C.), andAG042332 (to J.C. and T.H.). T.H. is an Investigator of the Howard HughesMedical Institute.

1. Laplante M, Sabatini DM (2012) mTOR signaling in growth control and disease. Cell149(2):274–293.

2. Laplante M, Sabatini DM (2009) mTOR signaling at a glance. J Cell Sci 122(Pt 20):3589–3594.

3. Peterson TR, et al. (2009) DEPTOR is an mTOR inhibitor frequently overexpressed inmultiple myeloma cells and required for their survival. Cell 137(5):873–886.

4. Spagnolo L, Rivera-Calzada A, Pearl LH, Llorca O (2006) Three-dimensional structureof the human DNA-PKcs/Ku70/Ku80 complex assembled on DNA and its implicationsfor DNA DSB repair. Mol Cell 22(4):511–519.

5. Ball HL, Cortez D (2005) ATRIP oligomerization is required for ATR-dependentcheckpoint signaling. J Biol Chem 280(36):31390–31396.

6. Bakkenist CJ, Kastan MB (2003) DNA damage activates ATM through intermolecularautophosphorylation and dimer dissociation. Nature 421(6922):499–506.

7. Wullschleger S, Loewith R, Oppliger W, Hall MN (2005) Molecular organization oftarget of rapamycin complex 2. J Biol Chem 280(35):30697–30704.

8. Zhang Y, Billington CJ, Jr, Pan D, Neufeld TP (2006) Drosophila target of rapamycinkinase functions as a multimer. Genetics 172(1):355–362.

9. Wang L, Rhodes CJ, Lawrence JC, Jr (2006) Activation of mammalian target of rapa-mycin (mTOR) by insulin is associated with stimulation of 4EBP1 binding to dimericmTOR complex 1. J Biol Chem 281(34):24293–24303.

10. Takahara T, Hara K, Yonezawa K, Sorimachi H, Maeda T (2006) Nutrient-dependentmultimerization of the mammalian target of rapamycin through the N-terminal HEATrepeat region. J Biol Chem 281(39):28605–28614.

11. Yip CK, Murata K, Walz T, Sabatini DM, Kang SA (2010) Structure of the human mTORcomplex I and its implications for rapamycin inhibition. Mol Cell 38(5):768–774.

12. Kim SG, et al. (2013) Metabolic stress controls mTORC1 lysosomal localization anddimerization by regulating the TTT-RUVBL1/2 complex. Mol Cell 49(1):172–185.

13. Tao Z, Barker J, Shi SD, Gehring M, Sun S (2010) Steady-state kinetic and inhibitionstudies of the mammalian target of rapamycin (mTOR) kinase domain and mTORcomplexes. Biochemistry 49(39):8488–8498.

14. Frias MA, et al. (2006) mSin1 is necessary for Akt/PKB phosphorylation, and its iso-forms define three distinct mTORC2s. Curr Biol 16(18):1865–1870.

15. Jain A, et al. (2011) Probing cellular protein complexes using single-molecule pull-down. Nature 473(7348):484–488.

16. Ulbrich MH, Isacoff EY (2007) Subunit counting in membrane-bound proteins. NatMethods 4(4):319–321.

17. Riley KJ, Yario TA, Steitz JA (2012) Association of Argonaute proteins and microRNAscan occur after cell lysis. RNA 18(9):1581–1585.

18. Kim DH, et al. (2002) mTOR interacts with raptor to form a nutrient-sensitive complexthat signals to the cell growth machinery. Cell 110(2):163–175.

19. Guertin DA, et al. (2006) Ablation in mice of the mTORC components raptor, rictor, ormLST8 reveals that mTORC2 is required for signaling to Akt-FOXO and PKCalpha, butnot S6K1. Dev Cell 11(6):859–871.

20. Huang J, Manning BD (2009) A complex interplay between Akt, TSC2 and the twomTOR complexes. Biochem Soc Trans 37(Pt 1):217–222.

21. Dibble CC, Asara JM, Manning BD (2009) Characterization of Rictor phosphoryla-tion sites reveals direct regulation of mTOR complex 2 by S6K1.Mol Cell Biol 29(21):5657–5670.

22. Julien LA, Carriere A, Moreau J, Roux PP (2010) mTORC1-activated S6K1 phosphor-ylates Rictor on threonine 1135 and regulates mTORC2 signaling. Mol Cell Biol 30(4):908–921.

23. Ulbrich MH, Isacoff EY (2008) Rules of engagement for NMDA receptor subunits. ProcNatl Acad Sci USA 105(37):14163–14168.

24. Dunne PD, et al. (2009) DySCo: Quantitating associations of membrane proteins usingtwo-color single-molecule tracking. Biophys J 97(4):L5–L7.

25. McMahon LP, Choi KM, Lin TA, Abraham RT, Lawrence JC, Jr (2002) The rapamycin-binding domain governs substrate selectivity by the mammalian target of rapamycin.Mol Cell Biol 22(21):7428–7438.

26. Yang H, et al. (2013) mTOR kinase structure, mechanism and regulation. Nature497(7448):217–223.

27. Oshiro N, et al. (2004) Dissociation of raptor frommTOR is a mechanism of rapamycin-induced inhibition of mTOR function. Genes Cells 9(4):359–366.

28. Fang Y, Vilella-Bach M, Bachmann R, Flanigan A, Chen J (2001) Phosphatidic acid-mediated mitogenic activation of mTOR signaling. Science 294(5548):1942–1945.

29. Sarbassov DD, et al. (2006) Prolonged rapamycin treatment inhibits mTORC2 assemblyand Akt/PKB. Mol Cell 22(2):159–168.

30. Goodsell DS, Olson AJ (2000) Structural symmetry and protein function. Annu RevBiophys Biomol Struct 29:105–153.

31. Jain A, Liu R, Xiang YK, Ha T (2012) Single-molecule pull-down for studying proteininteractions. Nat Protoc 7(3):445–452.

32. Yoon MS, Sun Y, Arauz E, Jiang Y, Chen J (2011) Phosphatidic acid activates mammaliantarget of rapamycin complex 1 (mTORC1) kinase by displacing FK506 binding protein 38(FKBP38) and exerting an allosteric effect. J Biol Chem 286(34):29568–29574.

17838 | www.pnas.org/cgi/doi/10.1073/pnas.1419425111 Jain et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

16,

202

1