Nanoparticle-based local translation reveals mRNA asa translation-coupled scaffold with anchoring functionShunnichi Kashidaa, Dan Ohtan Wangb,c, Hirohide Saitod, and Zoher Guerouia,1

aPASTEUR, Département de chimie, École normale supérieure, Paris Sciences et Lettres (PSL) University, Sorbonne Université, CNRS, 75005 Paris, France;bInstitute for Integrated Cell-Material Sciences, Kyoto University, 606-8501 Kyoto, Japan; cWuya College of Innovation, Shenyang Pharmaceutical University,Shenyang, Liaoning, 110016, People’s Republic of China; and dDepartment of Life Science Frontiers, Center for iPS Cell Research and Application, KyotoUniversity, 606-8507 Kyoto, Japan

Edited by Joseph Puglisi, Stanford University School of Medicine, Stanford, CA, and approved May 20, 2019 (received for review January 8, 2019)

The spatial regulation of messenger RNA (mRNA) translation is centralto cellular functions and relies on numerous complex processes. Biomi-metic approaches could bypass these endogenous complex processes,improve our comprehension of the regulation, and allow for controllinglocal translation regulations and functions. However, the causality be-tween local translation and nascent protein function remains elusive.Here, we developed a nanoparticle (NP)-based strategy to magneticallycontrol mRNA spatial patterns in mammalian cell extracts and investi-gate how local translation impacts nascent protein localization and func-tion. Bymonitoring the translation of themagnetically localizedmRNAs,we show that mRNA–NP complexes operate as a source for the contin-uous production of proteins from defined positions. By applying thisapproach to actin-binding proteins, we triggered the local formationof actin cytoskeletons and identified the minimal requirements for spa-tial control of the actin filament network. In addition, our bottom-upapproach identified a role formRNAas a translation-coupled scaffold forthe function of nascent N-terminal protein domains. Our approach willserve as a platform for regulating mRNA localization and investigatingthe function of nascent protein domains during translation.

local translation | magnetic nanoparticles | mRNA

One essential element often missing to explain cell fatecontrol concerns the spatiotemporal regulation of gene

expressions. Although DNA stays compacted in the nucleus,messenger RNA (mRNA) species are exported from the nucleusto precise locations in the cytosol before being translated.Translational control under mRNA spatial regulation signifi-cantly contributes to many important cellular processes, such asthe establishment of polarity, asymmetric division, synapticplasticity, and memory consolidation (1). For instance, the localtranslation of actin-associated proteins from subcellular-localized mRNAs has been increasingly recognized as an im-portant process for regulating the dynamic formation of the actincytoskeleton during cell polarization (2–7). Although bio-chemistry, genetics, cell imaging, biophysics, and “omics” ap-proaches have increased our understanding of the molecules andreaction mechanisms involved in the mRNA localization and localtranslation, none of these approaches can distinguish the func-tional contribution of the spatiotemporal dynamics of the mRNAfrom the molecular signaling. Thus, the causality between localtranslation and cell function remains elusive. Being able to designa bottom-up approach to bypass subcellular transport systems anduncouple endogenous biochemical regulation from physical con-straints could reveal the principles underlying complex biologicalsystems. In vitro synthetic biology and cell-free systems are pow-erful approaches for understanding the physical constraints thatshape gene expression networks, such as molecular crowding,compartmentalization, and space (diffusion and gradient) (8–12).However, understanding the spatiotemporal nature of translation

is still challenging and requires the development of novel orthog-onal methods for harnessing biomolecule functions. Optogenetics,magnetogenetics, and nanoparticle (NP)-based technologies couldprovide spatiotemporal control of biomolecule activities, but theseapproaches remain mainly limited to the control of signaling

pathways and transcription-based gene expressions (13–32) andhave not yet achieved spatiotemporal translation control.Here, we examined how magnetic control can be implemented

to spatially manipulate mRNA localization and its translation.As a first step toward this goal, we developed an in vitro localtranslation model based on synthetic mRNA-conjugated mag-netic NPs (mRNA-NPs) to control translation-based proteinsynthesis in space within confined HeLa cell extracts (Fig. 1).The mRNA-NPs were designed as movable mRNA-based pro-tein factories upon magnetic control. When supplied with suffi-cient translational ingredients, they could continuously synthesizespecific proteins with spatiotemporal control (Fig. 1). We achievedspatial patterns of mRNA-NPs in confined HeLa cell extractsusing magnetic forces to trigger asymmetric mRNA-NP string-likepatterns (Fig. 1). Using this platform, we simultaneously moni-tored the localized translation and the dynamic behavior of theprotein products. The results revealed that translation-coupledanchoring of the N-terminal nascent protein domains during thetranslation and mRNA localization can produce a local concen-tration of the functional protein domains. Its functional relevanceto cell function is demonstrated by the striking differences in actinfilament patterns that were triggered by simply rearranging anactin-binding protein domain toward the N terminus or C ter-minus of the protein translated on the mRNA-NPs.

ResultsTranslation of mRNA-NP Conjugates in HeLa Cell Extracts. As a firststep toward the spatial manipulation of mRNA molecules using

Significance

The spatial regulation of messenger RNA (mRNA) translation iscentral to cellular functions, yet no current approach hasachieved the spatial control of mRNAs to reveal the direct cau-sality between local translation and cellular processes. Here, wedevelop a biomimetic control system that bypasses endogenouscomplex processes to analyze this relationship. We demonstratethe magnetic manipulation of mRNA conjugated to nano-particles to spatially control translation-based protein synthesis.We examined how physical constraints can shape mRNA trans-lation and identify a mechanism of spatial control. We found thegeneral principles underlying the spatial regulation of actin cy-toskeleton patterns by local translation. From a biological per-spective, this phenomenon may represent a general mechanismof spatial control that cells use to produce functional diversity.

Author contributions: S.K., D.O.W., H.S., and Z.G. designed research; S.K. performed re-search; S.K., D.O.W., H.S., and Z.G. contributed new reagents/analytic tools; S.K. and Z.G.analyzed data; and S.K. and Z.G. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Published under the PNAS license.1To whom correspondence may be addressed. Email: zoher.gueroui@ens.fr.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1900310116/-/DCSupplemental.

Published online June 19, 2019.

13346–13351 | PNAS | July 2, 2019 | vol. 116 | no. 27 www.pnas.org/cgi/doi/10.1073/pnas.1900310116

Dow

nloa

ded

by g

uest

on

Mar

ch 1

8, 2

020

Dow

nloa

ded

by g

uest

on

Mar

ch 1

8, 2

020

Dow

nloa

ded

by g

uest

on

Mar

ch 1

8, 2

020

magnetic forces, we designed mRNA-NPs and assessed theirtranslation efficiency by monitoring green fluorescent protein(GFP) synthesis in HeLa cell extracts. mRNAs encoding GFPwere transcribed in vitro and biotinylated at their 3′ end. ThemRNAs were grafted onto the surface of streptavidin-conjugatedsuperparamagnetic NPs (with a size of 200 nm; Materials andMethods) by mixing in high-salt buffer at a stoichiometry of about800–1,600 mRNAs per NP (SI Appendix, Fig. S1A).We next monitored GFP protein synthesis from mRNA-NPs in

HeLa cell extracts, a model cytoplasm with a full complement oftranslational factors. To mimic the geometrical confinement bythe cell membranes, the extracts and mRNA-NPs were encapsulated

in spherical droplets dispersed in mineral oil with diameters rangingfrom 10–130 μm (Materials and Methods and SI Appendix, Fig. S1B).To assess the translational efficiency of the mRNA-NPs, we moni-tored the fluorescence of GFP translated from GFP-encodingmRNA-NPs in encapsulated extracts and found that GFP pro-teins were continuously produced from mRNA grafted on NPs asfunction of time (SI Appendix, Fig. S2). Thus, mRNA-NPs donot affect mRNA translatability and can be used for studyingtranslation and protein synthesis.Next, we assessed whether we could observe asymmetrically

localized mRNA-NPs and their translation efficiency in thedroplets. For the fluorescent detection of mRNA-NPs, Cy5-labeled biotin-mRNAs were supplemented with nonfluorescentbiotin-mRNA during the conjugation with NPs (Materials andMethods). Droplets of cell extracts were transferred into a slidechamber in the vicinity of a permanent neodymium (NdFeB)magnet that generates a field of about 0.15 T and a gradient of 50T·m−1 (Materials and Methods and Fig. 2A). Within a few secondsafter applying the magnetic field, the dispersed mRNA-NPs be-came organized into micrometric string-like assemblies that werealigned perpendicular to the magnet side of the droplet interface(Fig. 2B). The droplets were incubated for 4 h with or without thefield, and their fluorescence intensity showed that translated GFPproteins were homogeneously distributed within the droplets (Fig.2C and SI Appendix, Fig. S2B). This finding suggests that the newlysynthesized GFP proteins were successfully released from mRNA-NPs and rapidly diffused throughout the encapsulated cell ex-tracts. Relative GFP concentration showed that the efficiency ofGFP synthesis was independent of the magnetic field, indicatingthat neither the magnetic field nor the string-like mRNA-NPassemblies perturbed the translation process (Fig. 2C and SI

Fig. 1. Magnetic NP-based spatiotemporal regulation of mRNA translation.mRNAs conjugated to iron-oxide NPs (mRNA-NPs) are designed as an RNA-basedlocal source of protein synthesis, which combines the features of magnetic mo-bility and specific mRNA localization around the NP. The mRNA-NPs can contin-uously synthesize specific nascent proteins (e.g., GFP) and release them from theNPs. The spatial control of mRNA-NPs is mediated by a magnetic field to generatemRNA localization to assess the effect of mRNA-NPs on local translation.

mRNA-NP

mRNA(GFP)

Bright Field mRNA-NP GFP

Magnet

Bright Field mRNA-NP GFP

Magnet

Magnet

Magnet

8 min 68 min38 min 128 min

mC

herr

yG

FP

mRNA(GFP-mCherry)

NS

Magnet

Rel

ativ

e G

FPC

once

ntra

tion

(A.U

.)

55.47 54.18

0

20

40

60

80measurement after 4hr incubation.

Without magnet With magnet

n.s.

mRNA(GFP-375 a.a.-mCherry)

mRNA(GFP-122 a.a.)

10 min 70 min40 min 130 min

GFP

10 min 70 min40 min 130 min

GFP

GFP mRNA-NP areaGFP Counter edge areamCherry Center area

GFP-retention

diffused GFP-mCherry

Fluo

resc

ent i

ntei

nsty

(%)

min-20

0

20

40

60

80

100

120

0 20 40 60 80100120140-20

0

20

40

60

80

100

120

0 20 40 60 80 100 120 140

GFP

acc

umul

atio

n on

m

RN

A- N

P ar

ea (%

)

min

GFP-375a.a.-mCherry GFP-mCherry GFP-122a.a.

Magnet

Magnet

A

E

B

D

C

H

G

F I

Fig. 2. Translation of mRNA-NP conjugates in HeLa cell extract droplets. (A–C) GFP translation from localized mRNA-NPs. (A) Schematic of the translation fromasymmetrically localized mRNA-NPs performed in HeLa cell extract confined in droplets. The HeLa extract droplets were incubated in an observation chamber beside apermanent magnet. (B) Epifluorescence microscopy observation of Cy5-labeled mRNAs conjugated to NPs and asymmetrically aligned perpendicular to the magnet sideof the droplet interface. The translation of mRNA-NPs produces GFP that homogeneously diffuses within the droplet. (Scale bars: 20 μm.) (C) Relative GFP concentrationmeasured in the droplets after 4 h of incubation with (n = 52) or without (n = 26) the magnet. A.U., arbitrary unit; n.s., not significant. (D–F) GFP-mCherry translationfrom asymmetrically localized mRNA-NPs. (D, Top) Schematic of the translation process on localized mRNA-NPs. N-terminal nascent GFPs are anchored to mRNA-NPs viatranslating mCherry peptides and ribosomes. (D, Bottom) Representative time-lapse images of translated GFP and mCherry with epifluorescence microscopy. (Scale bars:20 μm.) (E) Cy5-labeled mRNA on NPs colocalizes with the translated GFP. (Scale bars: 20 μm.) (F) Time-dependent evolution of the normalized mean fluorescenceintensities of GFP andmCherry in selected regions of interest in the droplets (one frame every 5min). Error bars represent SD. (G andH) Representative time-lapse imagesof translated GFPs from localizedmRNA-NPs formRNA(GFP-122aa) andmRNA(GFP-375aa-mCherry), respectively. (Scale bars: 20 μm.) (I) Time-dependent evolution of themean fluorescence intensity of accumulated GFPs on mRNA-NPs for mRNA(GFP-375aa-mCherry), mRNA(GFP-122aa), and mRNA(GFP-mCherry).

Kashida et al. PNAS | July 2, 2019 | vol. 116 | no. 27 | 13347

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Mar

ch 1

8, 2

020

Appendix, Fig. S2B). To have information about the spatial lo-calization of ribosomes, we labeled the 28s ribosomal RNA ofthe ribosomes by using exciton-controlled hybridization-sensitivefluorescent oligonucleotide (ECHO) probes (33). In the absenceof magnetic localization of the mRNA-NPs, we observed a strongbackground noise and a few natural hotspots (SI Appendix, Fig. S3).In contrast, we observed bright fluorescent signals colocalizingwith mRNA-NPs in the presence of magnetic accumulation of themRNA-NPs, which supports an enrichment of the ribosomes on themRNA-NPs, and thus translation localization (SI Appendix, Fig. S3).

Monitoring of Nascent Proteins on Localized mRNA-NPs. To bettercharacterize the mRNA–NP complexes, we developed an assay tomonitor protein synthesis on localized mRNA-NPs and designed afusion protein permitting the use of nascent N-terminal GFP as areal-time reporter of local translation. We hypothesized thatthe translation duration of GFP-mCherry would be about twicethat of GFP and would allow nascent GFP to monitor the locationof its own mRNA until the release of GFP-mCherry proteins (Fig.2 D, schematic). To investigate this hypothesis, the translationof mRNA(GFP-mCherry)-NPs was analyzed using time-lapse mi-croscopy. About 10 min after the reaction started, we first detected aGFP signal that was persistent for hours and matched the localizedmRNA-NP pattern (Fig. 2 E and F). Furthermore, the GFP fluo-rescence within the droplets increased homogeneously, indicatingthat GFP-mCherry proteins were continuously translated andreleased from the mRNA-NPs (Fig. 2 D and F, SI Appendix, Fig.S4, and Movies S1 and S2). After 70–80 min, the GFP signals onthe mRNA-NPs reached a maximum and then slowly decayed,possibly because of the limited translational efficiency of theconfined cell extracts. Regarding mCherry, its fluorescencestarted to increase homogeneously within the droplets 50 minafter the reaction was initiated, in agreement with the slowermCherry maturation time (40 min) compared with GFP (34–36)(8 min; Fig. 2 F, red curve and Movies S1 and S2). In addition,we did not observe any colocalization between mCherry andmRNA-NPs, confirming that GFP-mCherry fusion proteins didnot stall or accumulate on NP surfaces.These observations identified several remarkable properties of

the local translation mediated by mRNA-NPs. One is that themRNA-NPs function to locally and continuously produce GFP-mCherry proteins that are released and dispersed in the solution.Therefore, as we hypothesized, the accumulation of GFP fluo-rescence on mRNA(GFP-mCherry)-NPs indicates that nascentN-terminal GFP was spatially retained on the mRNA via ribo-somes during mCherry synthesis (Fig. 2 D, schematic). Thus, wecould consider the N-terminal GFP as a transient fluorescentindicator of local translation from mRNA-NPs. To further in-vestigate the link between protein synthesis and the spatial re-tention of the nascent protein domain, we next monitored localGFP synthesis as a function of the length of the downstreamsequence after the GFP sequence. We designed two RNA con-structs whose downstream sequences were shorter and longerthan mRNA(GFP-mCherry), respectively: mRNA[GFP-122amino acids (aa)] and mRNA(GFP-375aa-mCherry) (Fig. 2 Gand H). We found a small and short-lived GFP protein accu-mulation on RNA-NPs during mRNA(GFP-122aa) translation(Fig. 2G, SI Appendix, Fig. S4, and Movie S3), but strong GFPprotein accumulation on the NPs during mRNA(GFP-375aa-mCherry)-NPs translation (Fig. 2H, SI Appendix, Fig. S4, andMovie S4). These data underlie that the spatial retention time ofGFP on mRNA-NPs is correlated with the peptide elongationlength during translation, as shown by the time delay differencebetween the GFP signal on mRNA-NPs and the GFP diffusionfraction in the droplet (Fig. 2I and SI Appendix, Fig. S5). Thetranslation speed in the droplets was estimated to be about 0.2–0.6aa per second (SI Appendix, Fig. S5 B and C), thus five- to 10-fold slower than that previously measured in mouse embryonicstem cells (37). We have attributed this difference to two factors(38, 39): First, our cell extract preparations were diluted, and,second, our experiments were performed at 30 °C, which is lower

than physiological temperature. From a biological perspective,this phenomenon may represent a simple translation-basedretaining mechanism of proteins at the site of translation for atime span defined by the mRNA translation duration.

Spatiotemporal Control of F-Actin Meshwork by Localized mRNATranslation. Next, we asked whether the spatial retention of theN-terminal protein domain on localized mRNA-NPs could pro-duce downstream functional diversity. We chose actin cytoskeletonpattern formation because of its relevance to important functionsof a living cell, its versatile patterns, and its high dynamic nature(40, 41). Many efforts have been made to decipher the molecularplayers involved in targeting and locally regulating actin-relatedmRNAs, yet current models are tightly dependent on the cellularcontext through multiple processes and regulators. We focused ona functional domain that is widely shared by the actin-bindingprotein superfamily, namely, the actin-binding domain of utrophin(ABD), which is shared with dystrophin and α-actinin (42–45).This 30-kDa domain contains two actin-binding sites to bind andstabilize actin filaments, and its GFP fusion protein can be used asan actin filament-promoting factor and label (46). We replaced themCherry protein of the GFP-mCherry fusion with ABD (Fig. 3).To investigate the significance of the N-terminal position of

ABD in the mRNA sequence, two mRNA fusions were developedin which the order of ABD and GFP was swapped (ABD-GFPand GFP-ABD; Fig. 3 A, Left). The mechanism of spatialretention permitted us to anticipate two situations: The trans-lation of mRNA(GFP-ABD)-NPs was designed to spatiallyretain N-terminal GFP on the mRNA-NPs area and will pro-duce GFP-ABD that diffuses immediately after synthesis, andthe translation of mRNA(ABD-GFP)-NPs was designed to spatiallyretain N-terminal ABD in the mRNA-NP area and then releaseABD-GFP that diffuses homogeneously in the droplets (Fig. 3A).First, we checked the translation of these two mRNAs and their

capacity to trigger the formation of actin filaments (F-actin). Thetwo mRNAs (without conjugation to NPs) were translated for 3 hin cell extract droplets supplemented with monomeric actin andfluorescent phalloidin, a probe for observing F-actin (SI Appendix,Fig. S6). We found that both mRNA constructs generate aninterconnected F-actin meshwork (SI Appendix, Fig. S6). Phal-loidin staining perfectly matched the translated ABD-GFP andGFP-ABD localization, which indicated that the respectiveABDs have the capacity to both promote and bind F-actin in thecell extracts (SI Appendix, Fig. S6).To assess actin filament formation from the localized mRNAs, we

next conjugated each mRNA construct with NPs (800:1 ratio) andassembled the mRNA-NPs into string-like patterns in confined ex-tracts. Interestingly, when examining the translation of asymmetricallydistributed mRNA-NPs, we found that the pattern of the F-actinmeshwork was highly dependent on the N-terminal or C-terminalposition of the ABD sequence on the mRNAs (Fig. 3 B and C).The translation of localized mRNA(GFP-ABD)-NPs led to the

formation of a homogenous F-actin meshwork without any localenrichment of the filaments in the vicinity of the mRNA-NPs (Fig.3 B, magnified views), as indicated by the phalloidin signal (Cy5-mRNA-NPs and phalloidin in Fig. 3B). In addition, we observedthe GFP signal accumulate at the position of Cy5-labeled mRNA-NPs and confirmed the spatial retention of N-terminal GFPduring downstream ABD local translation as in the case of GFP-mCherry (Cy5-mRNA-NP and GFP-ABD in Fig. 3B and SI Ap-pendix, Fig. S7). In contrast, the local translation of mRNA(ABD-GFP)-NPs showed a phalloidin signal that densely accumulatedaround Cy5-labeled mRNA-NPs, which suggested that the F-actinwas locally nucleating by the spatially retained N-terminal ABDon mRNA-NPs (Cy5-mRNA-NP and phalloidin in Fig. 3C, mag-nified views). Moreover, string-like assemblies of mRNA-NPsdisplayed a more compact and denser pattern, as if the forma-tion of an F-actin meshwork had modified the mRNA(ABD-GFP)-NP spatial organization (Cy5-mRNA-NP and phalloidin inFig. 3 B and C, magnified views). Although some diffused ABD-GFP molecules randomly promoted F-actin formation in the rest

13348 | www.pnas.org/cgi/doi/10.1073/pnas.1900310116 Kashida et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 1

8, 2

020

of the droplet, the locally translated ABD-GFP mainly corecruitedaround the mRNA-NPs with the accumulating F-actin cytoskeleton(Cy5-mRNA-NP, phalloidin, and GFP-ABD in Fig. 3C and SIAppendix, Fig. S7). The formation of localized F-actin mediated bythe spatially retained N-terminal ABD was also confirmed usingtime-lapse microscopy (Fig. 3D and Movie S5). F-actin accumula-tion on mRNA-NPs formed before nonlocalized actin formation, asrevealed by the local enrichment in phalloidin staining at the earlystage of translation (after about 10 min; Fig. 3D andMovie S5). Theobservation of an ABD-GFP–stained F-actin meshwork was de-tected later (after about 40 min), consistent with the timescale ofthe steady-state production of full-length ABD-GFP proteins (Fig.3D). Time-lapse observation also showed that the compacting ofstring-like assemblies of mRNA-NPs was correlated with the for-mation of localized F-actin meshwork (Fig. 3D and Movie S5).Altogether, these data showed that the N-terminal position of

ABD in the mRNA sequence and mRNA localization in thedroplet collaboratively dictated the micrometric local formationof the F-actin cytoskeleton.

Spatial Retention of Actin-Binding Proteins on Hotspots of TranslationalActivity Promotes Reorganization of the Actin Filament Meshwork.Ourobservations strongly suggest that mRNA(ABD-GFP)-NPs behaveas hotspots of active ABD-covered NPs to promote local F-actinformation. In addition, we found that mRNA(ABD-GFP)-NPstring-like patterns were changed upon local F-actin formation andeventually densely organized. Accordingly, we next examined howthe spatial organization of mRNA–NP complexes can be regu-lated by their own protein products in the absence of externalmagnetic forces. We first examined how individual mRNA(GFP-ABD)-NPs and mRNA(ABD-GFP)-NPs could trigger different

F-actin morphologies. As expected, the translation of mRNA(GFP-ABD)-NPs led to the formation of a homogenous F-actin mesh-work with a homogeneous spatial distribution of NPs within thedroplets (Fig. 4 A and C), as observed in Fig. 3B. Time-dependentconfocal observations distinguished the local translation activity onmRNA(GFP-ABD)-NPs, as indicated by the colocalization of Cy5-labeled mRNAs with GFP fluorescence (Fig. 4 A and C, SI Ap-pendix, Fig. S8, and Movie S6, Top). In contrast, the translation ofmRNA(ABD-GFP)-NPs led to a major spatial reorganization ofmRNA-NPs as a function of time (Fig. 4 B andD, SI Appendix, Fig.S8, and Movie S6, Bottom). After 15 min, the initially homoge-neous distribution of mRNA-NPs in the droplets started to formrandom micrometric clusters, and eventually evolved into largeassemblies within 45 min (Fig. 4D). Interestingly, this spatial re-organization of mRNA-NPs was correlated with the spatialremodeling of the F-actin meshwork (Fig. 4D). The formation of ahybrid meshwork composed of F-actin connected to NPs may ex-plain the dynamic clustering of mRNA-NPs observed in Fig. 4Dand may account for the compaction of mRNA-NP string-likestructures into dense organizations (Fig. 3 C and D). To test thishypothesis, we monitored the translation of nonbiotinylatedmRNA(ABD-GFP) in the presence of unconjugated NPs (SI Ap-pendix, Fig. S9 and Movie S7). In this case, F-actin formed a ho-mogeneous meshwork coexisting with randomly distributed NPs,thus phenocopying the translation of mRNA(GFP-ABD)-NPs (SIAppendix, Fig. S9 and Movie S7). These findings suggest that themultivalency provided by the multiple mRNA-NPs functions aslocal hotspots for translational activity, which is required for thelarge-scale reorganization of the F-actin meshwork in our system.In conclusion, these data suggest that the spatial retention of ABD

on hotspots for translational activity impacts the spatiotemporal

GFP spatial retention

actin cytoskeletonmorphology

i )

ii ) ABD GFPProtein Domain position

in mRNA sequence

DomainSwapping

ABDGFP

no accumulation

i ) ABDGFP

Magnet

Magnet

F-Actin labelmRNA-NP GFP-ABD Merge F-Actin labelF-

Act

inla

bel

mRNA-NP GFP-ABD Merge

actin filament accumulation

ii ) ABD GFP

Temporal order of translation

ABD spatial retention

GFP-ABD diffusion

ABD-GFPdiffusion

Magnet

Magnet

Magnet

Brig

htFi

eld

10 min 20 min 30 min 40 min 60 min 90 min

AB

D-G

FP

-100

100

300

500

700

900

1100

1300

0 30 60 90 120 150 180Fluo

resc

ent I

nten

isity

(A.U

.)

min

Phalloidin 568 (mRNA-NP Area)

GFP (mRNA-NP Area)

A

B

D

C

Fig. 3. Spatial control of the F-actin meshwork by localized mRNA translation highlights the key role of spatial retention of actin-binding protein. (A) To assess theeffect of the spatial retention of the N-terminal domain of actin-binding protein (ABD) during local translation on the actin filament assembly, twomRNA sequenceswere designed by swapping the order of ABD and GFP: ABD-GFP and GFP-ABD. These mRNAs were translated from localized mRNA-NPs in HeLa extract droplets.The translation of mRNA(GFP-ABD)-NPswas designed to spatially retain N-terminal GFP on the localizedmRNA-NPs, whereas the translation of mRNA(ABD-GFP)-NPswas designed to spatially retain N-terminal ABD on the localized mRNA-NPs. Representative confocal fluorescence microscopy images of Cy5-mRNA-NPs, phalloidin-568 (F-actin marker), and GFP-ABD (B) or ABD-GFP (C) are shown after 3 h of translation. (B, Magnified views) Formation of a homogenous F-actin meshworkwithout any local enrichment of filaments in the vicinity of the mRNA-NPs. (C, Magnified views) Formation of a local and dense F-actin meshwork colocalizing withmRNA-NPs. (D, Left) Representative time-lapse images of F-actin formation upon the translation of localized mRNA(ABD-GFP)-NPs. (D, Right) Quantification of themean fluorescence intensity of phalloidin-568 and GFP signal as a function of time in the mRNA-NP area. A.U., arbitrary unit. (Scale bars: B–D, 20 μm.)

Kashida et al. PNAS | July 2, 2019 | vol. 116 | no. 27 | 13349

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Mar

ch 1

8, 2

020

dynamics of F-actin assemblies, which can, in turn, regulate the lo-calization of mRNA-NPs. Thus, the temporary retention of ABDrenders the mRNA complexes the ability to interact with their owncoding proteins and downstream functional products.

DiscussionOur bottom-up approach is a first step toward the spatial controlof mRNA translation using magnetic NPs, thus expanding theapplications of magnetogenetics and NP-based technologies toRNA-based gene expression and spatial control. The designedmRNA-based complexes provide a platform for the spatial ma-nipulation of local translation. This feature relies on two proper-ties of the mRNA-NPs. First, by concentrating numerous mRNAson an iron-oxide NP with defined stoichiometry, the mRNA-NPsbehave as hotspots of local translation to continuously producenascent proteins. Second, the mRNA-NPs can be manipulated bymagnetic forces to generate localized spatial patterns. An excitingperspective of this work concerns the implementation of mRNA-NPs in cells and organisms where the spatiotemporal control ofRNA translation is key to determining cell fate.Over the past years, several studies have identified how mo-

lecular players such as cis-acting RNA localization elements andRNA-binding proteins, along with the transport of RNA gran-ules by molecular motors along cytoskeletal fibers, can act col-lectively to specifically regulate subcellular protein synthesis (1).However, the mechanisms regulating the dynamic behavior andspatial positioning of the synthetized proteins upon translationtermination and protein release remain elusive. To remain lo-calized in a subcellular area, specific mechanisms are used bycells to counterbalance the diffusion and spreading in space ofthe newly synthetized proteins. For instance, protein anchoringto a localized scaffold at the site of translation and reducedmobility due to geometrical constraints or molecular crowdingmay confine proteins at the site of translation. Interestingly, inour study, by monitoring both local translation on mRNA-NPsand nascent fluorescent protein synthesis, we found that mRNA

can act as an anchoring molecule that can spatially retain N-terminal folded protein domains during translation. By examiningthe local translation of magnetic assemblies of mRNA-NPs codingfor ABD, we found that the spatial retention of ABD in themRNA-NP area was required to localize F-actin patterns (Fig. 3and SI Appendix, Fig. S6). These localized F-actin assemblies can, inturn, modify the localization of the mRNA-NP position, providing afeed-forward mechanism between localized mRNA translation ac-tivity and cytoskeleton self-organization (Fig. 4). In this picture, thetemporary spatial retention of ABD renders the mRNA-NP theability to interact with its own synthesized protein and functionalpartners (Fig. 5). Experiments performed in bacteria and yeast havedemonstrated that protein complexes can assemble while subunits

mRNA(GFP-ABD)-NP Without Magnet

GFP-ABD MergemRNA-NP ABD-GFP MergemRNA-NP

AB

D-G

FPM

erge

mR

NA

-NP

mRNA(ABD-GFP)-NP Without Magnet

mRNA(ABD-GFP)-NP Without MagnetmRNA(GFP-ABD)-NP Without Magnet

GFP

-AB

DM

erge

mR

NA

-NP 2 hr15 m 45 m 1 hr

?

15 m 45 m 1 hr 2 hr

F-actin and NP disconnection

GFP retention on NP

F-actin and NP Network

ABD retention on NP

?

A

C D

B

Fig. 4. Spatial retention of actin-binding proteins on hotspots of translational activity promotes reorganization of the actin filament meshwork. Confocal obser-vations of translation from mRNA(GFP-ABD)-NPs (A) and mRNA(ABD-GFP)-NPs (B) assess how the spatial organization of mRNA–NP complexes can be regulated bytheir own protein products. Observations were performed after 2 h of translation. The F-actin meshwork is labeled by GFP-ABD andmRNA-NPs by Cy5 labeled-mRNAs.(C and D) Time-lapse observations of the dynamics of mRNA translation and mRNA–NP complexes. (C) N-terminal GFP signals (green) are localized on mRNA(GFP-ABD)-NPs (red, Cy5 labeled-mRNAs) after 45 min of translation. mRNA-NPs are spatially distributed in the droplet, and GFP-ABD–bound actin filaments form a ho-mogenous meshwork. (D) Translation of mRNA(ABD-GFP)-NPs leads to a spatial reorganization of mRNA-NPs as a function of time: mRNA(ABD-GFP)-NPs (red) andABD-GFP (green) bound to actin filaments spontaneously accumulate together to form a dynamic meshwork that remodels spatially. (Scale bars: 20 μm.).

Fig. 5. Proposed model of mRNA as a translation-coupled scaffold withanchoring function to spatially constrain protein interactions. Spatiallyconstrained protein interactions could be cotranslationally mediated by thespatial retention of nascent protein domains on localized mRNA. For in-stance, the N-terminal position of the actin-binding domain sequence onlocalized mRNA leads to the spatial retention of the protein domain duringtranslation and eventually promotes actin filament localization.

13350 | www.pnas.org/cgi/doi/10.1073/pnas.1900310116 Kashida et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 1

8, 2

020

are synthetized by the ribosome machinery cotranslationally in cells(47–49). In addition, recent studies have discovered novel functionsof RNA as scaffolds that promote protein–protein interactions, withthe 3′ untranslated region of the mRNAs recruiting proteins to thesite of translation to determine the subcellular protein localization(50), or with long noncoding RNA acting as a platform for post-translational functions (51). Here, we show that the coding se-quence of mRNA, in addition to providing genetic information, canact as a scaffold molecule to localize protein–protein interactionsbetween N-terminal protein domains that fold during translationand their binding targets (Fig. 5).The mechanism for spatially retaining protein domains during local

translation may have relevant physiological implications. Given themodular organization of protein domain structures, a temporary localconcentration of functional modules that are produced early intranslation can be generated. Numerous actin-binding proteins haveABDs at their N-terminal position (42–45), suggesting the possibilityof the spatial and temporal retention of the ABDs on mRNA tem-plates for a duration defined by the extent of the downstreammRNAtranslation process. For instance, in humans, utrophin mRNAs andproteins are both known to be localized at neuromuscular junctions inmouse muscle cells, and the long utrophin mRNA sequence down-stream of the ABD (9,537 bases which is 13-fold longer than that ofGFP) may thus serve as a scaffold to spatially retain ABD (42, 43).

To conclude, our in vitro local translation model allowed us toidentify possible anchoring functions of mRNA occurring cotrans-lationally and mediated by the spatial retention of the N-terminaldomain (Fig. 5). Further investigations are needed to determine ifthis anchoring property occurs in vivo to enable localized protein–protein interactions by creating an enriched concentration of na-scent functional protein domains that regulate local translation.

Materials and MethodsIn vitro-transcribed andbiotin end-labeledmRNAwas conjugated to streptavidin-coated magnetic NPs and encapsulated with HeLa cell extract in droplets. Thedroplets in a slide chamber beside theNdFeBmagnetwere incubated at 30 °C andmonitored under epifluorescence or a confocal microscope. Phalloidin-568 wassupplemented with the extract to visualize actin cytoskeleton formation. De-tailed materials and methods are available in SI Appendix.

ACKNOWLEDGMENTS. We acknowledge the members of the BiophysicalChemistry group of the École normale supérieure (ENS) for fruitful discussions.We thank Dr. K. Endo and Dr. P. Karagiannis for carefully reading the manu-script. We also thank Drs. I. Oomoto, C. Parr, A. Hubstenberger, and Y. Fujita.S.K. was supported by a Fellowship, “bourse du gouvernement français.” Thiswork was supported by Human Frontier Science Program, Program GrantRGP0050/2014 (to D.O.W., H.S., and Z.G.), by the CNRS (Z.G.), and by theENS (Z.G.).

1. K. C. Martin, A. Ephrussi, mRNA localization: Gene expression in the spatial di-mension. Cell 136, 719–730 (2009).

2. D. St Johnston, Moving messages: The intracellular localization of mRNAs. Nat. Rev.Mol. Cell Biol. 6, 363–375 (2005).

3. J. Condeelis, R. H. Singer, How and why does β-actin mRNA target? Biol. Cell 97, 97–110 (2005).

4. S. Hüttelmaier et al., Spatial regulation of β-actin translation by Src-dependentphosphorylation of ZBP1. Nature 438, 512–515 (2005).

5. L. A. Mingle et al., Localization of all seven messenger RNAs for the actin-polymerization nucleator Arp2/3 complex in the protrusions of fibroblasts. J. CellSci. 118, 2425–2433 (2005).

6. S. Mili, K. Moissoglu, I. G. Macara, Genome-wide screen reveals APC-associated RNAsenriched in cell protrusions. Nature 453, 115–119 (2008).

7. F. K. Mardakheh et al., Global analysis of mRNA, translation, and protein localization:Local translation is a key regulator of cell protrusions. Dev. Cell 35, 344–357 (2015).

8. C. Tan, S. Saurabh, M. P. Bruchez, R. Schwartz, P. Leduc, Molecular crowding shapes geneexpression in synthetic cellular nanosystems. Nat. Nanotechnol. 8, 602–608 (2013).

9. M. M. K. Hansen et al., Macromolecular crowding creates heterogeneous environ-ments of gene expression in picolitre droplets. Nat. Nanotechnol. 11, 191–197 (2016).

10. E. Karzbrun, A. M. Tayar, V. Noireaux, R. H. Bar-Ziv, Programmable on-chip DNAcompartments as artificial cells. Science 345, 829–832 (2014).

11. M. Weitz et al., Diversity in the dynamical behaviour of a compartmentalized pro-grammable biochemical oscillator. Nat. Chem. 6, 295–302 (2014).

12. P. Torre, C. D. Keating, S. S. Mansy, Multiphase water-in-oil emulsion droplets for cell-free transcription-translation. Langmuir 30, 5695–5699 (2014).

13. J. E. Toettcher, C. A. Voigt, O. D. Weiner, W. A. Lim, The promise of optogenetics in cellbiology: Interrogating molecular circuits in space and time. Nat. Methods 8, 35–38 (2011).

14. P. Del Pino, Tailoring the interplay between electromagnetic fields and nanomaterialstoward applications in life sciences: A review. J. Biomed. Opt. 19, 101507 (2014).

15. P. del Pino et al., Gene silencing mediated by magnetic lipospheres tagged with smallinterfering RNA. Nano Lett. 10, 3914–3921 (2010).

16. R. Chen, A. Canales, P. Anikeeva, Neural recording and modulation technologies. Nat.Rev. Mater. 2, 16093 (2017).

17. F. Etoc et al., Subcellular control of Rac-GTPase signalling by magnetogenetic ma-nipulation inside living cells. Nat. Nanotechnol. 8, 193–198 (2013).

18. M. Meister, Physical limits to magnetogenetics. eLife 5, 589–599 (2016).19. A. Tay, A. Kunze, C. Murray, D. Di Carlo, Induction of calcium influx in cortical neural

networks by nanomagnetic forces. ACS Nano 10, 2331–2341 (2016).20. M. A. Wheeler et al., Genetically targeted magnetic control of the nervous system.

Nat. Neurosci. 19, 756–761 (2016).21. F. Scherer et al., Magnetofection: Enhancing and targeting gene delivery by magnetic

force in vitro and in vivo. Gene Ther. 9, 102–109 (2002).22. Q. A. Pankhurst, J. Connolly, S. K. Jones, J. Dobson, Applications of magnetic nano-

particles in biomedicine. J. Phys. D Appl. Phys. 36, R167–R181 (2003).23. M. Creixell, A. C. Bohórquez, M. Torres-Lugo, C. Rinaldi, EGFR-targeted magnetic

nanoparticle heaters kill cancer cells without a perceptible temperature rise. ACSNano 5, 7124–7129 (2011).

24. M. Cruz-Acuña, J. R. Halman, K. A. Afonin, J. Dobson, C. Rinaldi, Magnetic nanoparticlesloaded with functional RNA nanoparticles. Nanoscale 10, 17761–17770 (2018).

25. R. J. Mannix et al., Nanomagnetic actuation of receptor-mediated signal transduction.Nat. Nanotechnol. 3, 36–40 (2008).

26. M. H. Cho et al., A magnetic switch for the control of cell death signalling in in vitroand in vivo systems. Nat. Mater. 11, 1038–1043 (2012).

27. H. Huang, S. Delikanli, H. Zeng, D. M. Ferkey, A. Pralle, Remote control of ion channels andneurons throughmagnetic-field heating of nanoparticles. Nat. Nanotechnol. 5, 602–606 (2010).

28. S. A. Stanley et al., Radio-wave heating of iron oxide nanoparticles can regulateplasma glucose in mice. Science 336, 604–608 (2012).

29. M. Yamaguchi, A. Ito, A. Ono, Y. Kawabe, M. Kamihira, Heat-inducible gene ex-pression system by applying alternating magnetic field to magnetic nanoparticles.ACS Synth. Biol. 3, 273–279 (2014).

30. C. Hoffmann et al., Spatiotemporal control of microtubule nucleation and assemblyusing magnetic nanoparticles. Nat. Nanotechnol. 8, 199–205 (2013).

31. C. C. Chen et al., DNA-gold nanorod conjugates for remote control of localized geneexpression by near infrared irradiation. J. Am. Chem. Soc. 128, 3709–3715 (2006).

32. D. Seo et al., A mechanogenetic toolkit for interrogating cell signaling in space andtime. Cell 165, 1507–1518 (2016).

33. I. Oomoto et al., ECHO-liveFISH: In vivo RNA labeling reveals dynamic regulation ofnuclear RNA foci in living tissues. Nucleic Acids Res. 43, e126 (2015).

34. J. Shin, V. Noireaux, Study of messenger RNA inactivation and protein degradation inan Escherichia coli cell-free expression system. J. Biol. Eng. 4, 9 (2010).

35. Y. V. Svitkin et al., N1-methyl-pseudouridine in mRNA enhances translation througheIF2α-dependent and independent mechanisms by increasing ribosome density. Nu-cleic Acids Res. 45, 6023–6036 (2017).

36. P. J. Macdonald, Y. Chen, J. D. Mueller, Chromophore maturation and fluorescencefluctuation spectroscopy of fluorescent proteins in a cell-free expression system. Anal.Biochem. 421, 291–298 (2012).

37. N. T. Ingolia, L. F. Lareau, J. S. Weissman, Ribosome profiling ofmouse embryonic stem cellsreveals the complexity and dynamics of mammalian proteomes. Cell 147, 789–802 (2011).

38. S. Mikami, T. Kobayashi, M. Masutani, S. Yokoyama, H. Imataka, A human cell-derivedin vitro coupled transcription/translation system optimized for production of re-combinant proteins. Protein Expr. Purif. 62, 190–198 (2008).

39. S. Mikami, M. Masutani, N. Sonenberg, S. Yokoyama, H. Imataka, An efficientmammalian cell-free translation system supplemented with translation factors. Pro-tein Expr. Purif. 46, 348–357 (2006).

40. G. Letort, H. Ennomani, L. Gressin, M. Théry, L. Blanchoin, Dynamic reorganization ofthe actin cytoskeleton. F1000 Res. 4, F1000 Faculty Rev-940 (2015).

41. T. D. Pollard, J. A. Cooper, Actin, a central player in cell shape and movement. Science326, 1208–1212 (2009).

42. R. Vater et al., Utrophin mRNA expression in muscle is not restricted to the neuro-muscular junction. Mol. Cell. Neurosci. 10, 229–242 (1998).

43. N. H. Keep et al., Crystal structure of the actin-binding region of utrophin reveals ahead-to-tail dimer. Structure 7, 1539–1546 (1999).

44. E. Borrego-Diaz et al., Crystal structure of the actin-binding domain of α-actinin 1:Evaluating two competing actin-binding models. J. Struct. Biol. 155, 230–238 (2006).

45. K. Röper, S. L. Gregory, N. H. Brown, The ‘spectraplakins’: Cytoskeletal giants withcharacteristics of both spectrin and plakin families. J. Cell Sci. 115, 4215–4225 (2002).

46. S. J. Winder et al., Utrophin actin binding domain: Analysis of actin binding andcellular targeting. J. Cell Sci. 108, 63–71 (1995).

47. Y. W. Shieh et al., Operon structure and cotranslational subunit association directprotein assembly in bacteria. Science 350, 678–680 (2015).

48. A. Shiber et al., Cotranslational assembly of protein complexes in eukaryotes revealedby ribosome profiling. Nature 561, 268–272 (2018).

49. C. Mayr, Protein complexes assemble as they are being made. Nature 561, 186–187 (2018).50. B. D. Berkovits, C. Mayr, Alternative 3′ UTRs act as scaffolds to regulate membrane

protein localization. Nature 522, 363–367 (2015).51. J. H. Yoon et al., Scaffold function of long non-coding RNA HOTAIR in protein

ubiquitination. Nat. Commun. 4, 2939 (2013).

Kashida et al. PNAS | July 2, 2019 | vol. 116 | no. 27 | 13351

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Mar

ch 1

8, 2

020

Correction

BIOPHYSICS AND COMPUTATIONAL BIOLOGYCorrection for “Nanoparticle-based local translation revealsmRNA as a translation-coupled scaffold with anchoring function,”by Shunnichi Kashida, Dan Ohtan Wang, Hirohide Saito, andZoher Gueroui, which was first published June 19, 2019; 10.1073/pnas.1900310116 (Proc. Natl. Acad. Sci. U.S.A. 116, 13346–13351).

The authors note that, in Fig. 3C, the label above the secondimage from the right incorrectly reads “GFP-ABD.” It shouldinstead read “ABD-GFP.” The corrected figure and its legendappear below.

Published under the PNAS license.

Published online August 12, 2019.

www.pnas.org/cgi/doi/10.1073/pnas.1912167116

GFP spatial retention

actin cytoskeletonmorphology

A

i )

ii ) ABD GFPProtein Domain position

in mRNA sequence

DomainSwapping

ABDGFP

B

Dno accumulation

i ) ABDGFP

Magnet

C

Magnet

F-Actin labelmRNA-NP GFP-ABD Merge F-Actin label

F-A

ctin

labe

l

mRNA-NP ABD-GFP Merge

actin filament accumulation

ii ) ABD GFP

Temporal order of translation

ABD spatial retention

GFP-ABD diffusion

ABD-GFPdiffusion

Magnet

Magnet

Magnet

Brig

htFi

eld

10 min 20 min 30 min 40 min 60 min 90 min

AB

D-G

FP

-100

100

300

500

700

900

1100

1300

0 30 60 90 120 150 180Fluo

resc

ent I

nten

isity

(A.U

.)

min

Phalloidin 568 (mRNA-NP Area)

GFP (mRNA-NP Area)

Fig. 3. Spatial control of the F-actin meshwork by localized mRNA translation highlights the key role of spatial retention of actin-binding protein. (A) Toassess the effect of the spatial retention of the N-terminal domain of actin-binding protein (ABD) during local translation on the actin filament assembly, twomRNA sequences were designed by swapping the order of ABD and GFP: ABD-GFP and GFP-ABD. These mRNAs were translated from localized mRNA-NPs inHeLa extract droplets. The translation of mRNA(GFP-ABD)-NPs was designed to spatially retain N-terminal GFP on the localized mRNA-NPs, whereas thetranslation of mRNA(ABD-GFP)-NPs was designed to spatially retain N-terminal ABD on the localized mRNA-NPs. Representative confocal fluorescence mi-croscopy images of Cy5-mRNA-NPs, phalloidin-568 (F-actin marker), and GFP-ABD (B) or ABD-GFP (C) are shown after 3 h of translation. (B, Magnified views)Formation of a homogenous F-actin meshwork without any local enrichment of filaments in the vicinity of the mRNA-NPs. (C,Magnified views) Formation of alocal and dense F-actin meshwork colocalizing with mRNA-NPs. (D, Left) Representative time-lapse images of F-actin formation upon the translation of lo-calized mRNA(ABD-GFP)-NPs. (D, Right) Quantification of the mean fluorescence intensity of phalloidin-568 and GFP signal as a function of time in the mRNA-NP area. A.U., arbitrary unit. (Scale bars: B–D, 20 μm.)

www.pnas.org PNAS | August 13, 2019 | vol. 116 | no. 33 | 16655

CORR

ECTION