DOI: 10.1126/science.1250127, 688 (2014);345 Science

et al.Nikolai A. Naryshkinmice with spinal muscular atrophy

splicing modifiers improve motor function and longevity inSMN2

This copy is for your personal, non-commercial use only.

clicking here.colleagues, clients, or customers by , you can order high-quality copies for yourIf you wish to distribute this article to others

here.following the guidelines

can be obtained byPermission to republish or repurpose articles or portions of articles

): August 16, 2014 www.sciencemag.org (this information is current as of

The following resources related to this article are available online at

http://www.sciencemag.org/content/345/6197/688.full.htmlversion of this article at:

including high-resolution figures, can be found in the onlineUpdated information and services,

http://www.sciencemag.org/content/suppl/2014/08/06/345.6197.688.DC1.html can be found at: Supporting Online Material

http://www.sciencemag.org/content/345/6197/688.full.html#relatedfound at:

can berelated to this article A list of selected additional articles on the Science Web sites

http://www.sciencemag.org/content/345/6197/688.full.html#ref-list-1, 24 of which can be accessed free:cites 56 articlesThis article

http://www.sciencemag.org/content/345/6197/688.full.html#related-urls1 articles hosted by HighWire Press; see:cited by This article has been

http://www.sciencemag.org/cgi/collection/medicineMedicine, Diseases

subject collections:This article appears in the following

registered trademark of AAAS. is aScience2014 by the American Association for the Advancement of Science; all rights reserved. The title

CopyrightAmerican Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005. (print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by theScience

on

Aug

ust 1

6, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

on

Aug

ust 1

6, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

on

Aug

ust 1

6, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

on

Aug

ust 1

6, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

on

Aug

ust 1

6, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

on

Aug

ust 1

6, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

on

Aug

ust 1

6, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

MOTOR NEURON DISEASE

SMN2 splicing modifiers improvemotor function and longevity in micewith spinal muscular atrophyNikolai A. Naryshkin,1 Marla Weetall,1 Amal Dakka,1 Jana Narasimhan,1 Xin Zhao,1

Zhihua Feng,2 Karen K. Y. Ling,2 Gary M. Karp,1 Hongyan Qi,1 Matthew G. Woll,1

Guangming Chen,1 Nanjing Zhang,1 Vijayalakshmi Gabbeta,1 Priya Vazirani,1

Anuradha Bhattacharyya,1 Bansri Furia,1 Nicole Risher,1 Josephine Sheedy,1

Ronald Kong,1 Jiyuan Ma,1 Anthony Turpoff,1 Chang-Sun Lee,1 Xiaoyan Zhang,1

Young-Choon Moon,1 Panayiota Trifillis,1 Ellen M. Welch,1 Joseph M. Colacino,1

John Babiak,1 Neil G. Almstead,1 Stuart W. Peltz,1* Loren A. Eng,3 Karen S. Chen,3

Jesse L. Mull,4 Maureen S. Lynes,4 Lee L. Rubin,4 Paulo Fontoura,5 Luca Santarelli,5

Daniel Haehnke,5 Kathleen D. McCarthy,3 Roland Schmucki,5 Martin Ebeling,5

Manaswini Sivaramakrishnan,5 Chien-Ping Ko,2 Sergey V. Paushkin,3 Hasane Ratni,5

Irene Gerlach,5 Anirvan Ghosh,5 Friedrich Metzger5*

Spinal muscular atrophy (SMA) is a genetic disease caused by mutation or deletion

of the survival of motor neuron 1 (SMN1) gene. A paralogous gene in humans, SMN2,

produces low, insufficient levels of functional SMN protein due to alternative splicing

that truncates the transcript. The decreased levels of SMN protein lead to progressive

neuromuscular degeneration and high rates of mortality. Through chemical screening

and optimization, we identified orally available small molecules that shift the balance

of SMN2 splicing toward the production of full-length SMN2 messenger RNA with high

selectivity. Administration of these compounds to D7 mice, a model of severe SMA,

led to an increase in SMN protein levels, improvement of motor function, and

protection of the neuromuscular circuit. These compounds also extended the life

span of the mice. Selective SMN2 splicing modifiers may have therapeutic potential

for patients with SMA.

Spinal muscular atrophy (SMA) is the lead-

ing genetic cause of infant mortality with

an incidence of 1 in 11,000 live births (1–3).

SMA is characterized by progressive degen-

eration of anterior horn a motoneurons,

as well as muscle weakness and atrophy that

affect mainly proximal muscles. The disorder is

associated with a high rate of mortality due to

respiratory complications (4). Clinically, SMA

disease phenotypes range from severe type I to

mild types III and IV. Patients are classified

based on the age of onset, clinical severity, and

the ability to achieve milestones of motor co-

ordination development (5). There is currently

no treatment for the disease except supportive

care to improve quality of life by minimizing

disability and discomfort (6, 7).

The disease SMA is caused by deletion or mu-

tation of the survival of motor neuron 1 (SMN1)

gene, which encodes the SMN protein (8, 9).

Humans carry two paralogous SMN1 and SMN2

genes that are ubiquitously expressed, with the

SMN protein principally produced from SMN1

full-length (FL) mRNA. The coding regions of

SMN1 and SMN2 differ only in a translationally

silent C-to-T transition at nucleotide 840, lead-

ing to alternative splicing and exclusion of exon 7

frommost SMN2 transcripts. The resultingmRNA,

referred to as D7, encodes an unstable SMND7

protein that is rapidly degraded (10, 11). How-

ever, the low levels of SMN2 transcripts that

include exon 7 produce small amounts of fully

functional SMN protein. SMN protein is a ubi-

quitously expressed 38-kD protein that is involved

in multiple cellular functions including pre-mRNA

splicing, small nuclear ribonucleoprotein (snRNP)

biogenesis, transcription, stress response, apo-

ptosis, axonal transport, and cytoskeletal dynam-

ics (12–14). The exact mechanism by which SMN

protein deficiency leads to motoneuron loss is

unclear (15).

Though the lack of the SMN1 gene is the un-

derlying cause of the disease, SMN2 is the pri-

mary disease-modifying gene in SMA (16). In SMA

patients, the number of SMN2 genes varies be-

tween two and six copies in a somatic cell (17), and

SMNprotein levels generally correlatewith SMN2

copy number (16, 18). Similarly, the severity of

disease inversely correlates with SMN2 copy num-

ber and the amount of functional SMN protein

(16, 19, 20). Relative to the levels of SMN in

healthy individuals, SMN protein levels are re-

duced by ~70, ~50, and ~30% in type I, II and

III patients, respectively (17, 21). This suggests

that moderate differences in SMN protein lev-

els can modify disease severity.

To determine whether pharmacologically in-

creasing SMN levels would ameliorate pheno-

types associated with the loss of the SMN1 gene,

we sought to identify compounds that selectively

modulate SMN2 pre-mRNA splicing to include

exon 7 (illustrated schematically in fig. S1). By

using a HEK293H human embryonic kidney cell

line harboring an SMN2 minigene to screen a

library of small molecules (fig. S2), we identified

several chemical classes of compounds that pro-

mote the inclusion of exon 7 in the SMN2mRNA

(see supplementary materials and methods). Af-

ter further optimization of these splicing modi-

fiers, we identified three series of compounds

that were orally available (SMN-C1, SMN-C2,

and SMN-C3) (Fig. 1A and fig. S2). With nano-

molar potency, these compounds increased FL

SMN2mRNA levels and concomitantly reducedD7

mRNA levels in SMA type I patient fibroblasts after

24 hours of treatment (Fig. 1B). The abundance of

SMN protein was also increased in these fibro-

blasts, as determined byWestern blot analysis (Fig.

1C). Fibroblasts fromSMAtype I, II, and III patients

and from an unaffected control exhibited different

baseline levels of SMNprotein (DF), but SMN levels

increased in a dose-dependent manner when the

cells were treated with SMN-C1 or SMN-C2, as de-

termined by homogeneous time-resolved fluores-

cence (HTRF) (Fig. 1D).

To ensure that the increase in SMN protein

levels also occurred in disease-relevant cells, we

quantified SMN protein in Islet-1 positive moto-

neurons and Islet-1 negative cells (mainly glia

and neuronal stem cells) in cell cultures gen-

erated from patient-derived induced pluripotent

stem cells (iPSCs). Treatment with SMN-C3 in-

creased SMN protein levels in cells from SMA

type I (Fig. 1E) and II patients (fig. S3). Quan-

titatively, the increase was similar in Islet-1–

positive motoneurons and Islet-1–negative cells

(Fig. 1E and fig. S3). Thus, these small-molecule

splicing modifiers can modulate SMN2 splicing

toward exon 7 inclusion and thereby increase the

production of SMN protein in different cell types

and across SMA disease severities.

We next characterized the selectivity of the

SMN splicing modifiers in SMA type I fibro-

blasts by performing RNA sequence analysis

on cells treated with SMN-C3 (500 nM) or with

control solvent [0.5% dimethyl sulfoxide (DMSO)].

SMN-C3 treatment altered the expression of

very few of the 11,714 genes that were expressed

under either control or treatment conditions

(Fig. 2A). Specifically, six genes each were up- or

down-regulated by a factor > 2 (log2 > 1 or < –1)

(fig. S4) after SMN-C3 treatment [confirmed by

quantitative real-time polymerase chain reac-

tion (RT-qPCR); see fig S4], indicating that this

compound did not induce widespread changes

in gene expression. Notably, SMN-C3 did not

cause deregulated expression of gene families

involved in DNA or RNA metabolism or cell

688 8 AUGUST 2014 • VOL 345 ISSUE 6197 sciencemag.org SCIENCE

1PTC Therapeutics, 100 Corporate Court, South Plainfield, NJ07080, USA. 2Section of Neurobiology, Department ofBiological Sciences, University of Southern California, LosAngeles, CA 90089, USA. 3SMA Foundation, 888 SeventhAvenue, Suite 400, New York, NY 10019, USA. 4Departmentof Stem Cell and Regenerative Biology and the Harvard StemCell Institute, Harvard University, Cambridge, MA 02138,USA. 5Roche Pharmaceutical Research and EarlyDevelopment, Roche Innovation Center Basel, F. Hoffmann-LaRoche, Grenzacherstrasse 124, 4070 Basel, Switzerland.*Corresponding author. E-mail: friedrich.metzger@roche.com

(F.M.); speltz@ptcbio.com (S.W.P.)

RESEARCH | REPORTS

survival (fig. S4). Treatment of cells with SMN-C1

(100 nM) produced a similar gene expression

profile as that observed with SMN-C3 (fig. S5).

Among the genes whose expression was altered

by SMN-C3, there are several candidates that

could conceivably play a role in the regulation of

SMN expression or protein function and thus

contribute to the compound’s therapeutic effect

(see below). For example, DNA polymerase N

(POLN), a low-fidelity DNA polymerase (22) that

is up-regulated by SMN-C3, is involved in DNA

repair and homologous recombination (23, 24).

PAP-associated domain containing 4 protein

(PAPD4), which is down-regulated by SMN-C3,

is a noncanonical poly(A) polymerase that reg-

ulates the generation and stability of micro-

RNAs as well as adenylation of microRNAs after

cleavage byDicer (25–28). However, to date, none

of these genes has been directly linked to SMN

expression and/or function or to SMA patho-

genesis. It is also possible that indirect pathways

(e.g., involving microRNA regulation) may play a

role in SMN regulation.

A separate analysis of annotated splice junc-

tions within the transcripts identified a few sin-

gle splice junctions that were affected by SMN-C3

treatment without changes in the overall abun-

dance of the corresponding mRNAs (Fig. 2B).

Among the affected splice junctions were the

SMN2 exon 7 junctions, confirming the RT-PCR

results. SMN-C3 treatment also induced changes

in multiple splice junctions of PDXDC1 (Fig. 2B),

a gene of unknown function that is the most

strongly down-regulated gene on a transcrip-

tional level. Similar results were seen when SMA

type I patient fibroblasts were treatedwith SMN-

C1 (100 nM) (fig. S5). These results suggest that

the SMN splicingmodifiers are highly specific for

processing of SMN2 pre-mRNA.

We also assessed the role of several cis-acting

splicing regulatory elements in the activity of our

splicing modifiers. We evaluated the Exinct

SCIENCE sciencemag.org 8 AUGUST 2014 • VOL 345 ISSUE 6197 689

Fig. 1. Small molecules modulate SMN2 alternative splicing and increase

SMN protein levels in cells from SMA patients. (A) Chemical structures of

SMN-C1, SMN-C2, and SMN-C3 small molecules. (B) RT-PCR analysis of

SMN2 mRNAs. (Top) SMN-C1 treatment for 24 hours increases the level of FL

mRNA and reduces the level of mRNA lacking exon 7 (D7) in SMA type I patient

fibroblasts. (Bottom) The effects of SMN-C1, SMN-C2, and SMN-C3 on FL and

D7 mRNA levels in SMA type I patient fibroblasts are concentration-dependent.

(C) (Left) Western blot of SMN protein in SMA type I patient fibroblasts after

48 hours of continuous treatment with SMN-C3. Glyceraldehyde-3-phosphate

dehydrogenase (GAPDH) was used as a loading control. (Right) The effects of

SMN-C1, SMN-C2, and SMN-C3 on SMN protein abundance are concentration-

dependent. GAPDH and actin were used as loading controls. (D) HTRFanalysis

(48) of the SMN protein after 48 hours of continuous treatment with SMN-C1

(left) and SMN-C2 (right) in fibroblasts from SMA type I, II, and III patients or

from an unaffected control (carrier). (E) (Left) Motoneuron cultures gener-

ated from iPSCs derived from a SMA type I patient. Cells were treated with

DMSO (control) or SMN-C3 (300 nM) and immunostained for SMN and

Islet-1. Scale bar, 100 mm. (Right) The effect of SMN-3 on SMN protein levels

in Islet-1–positive motoneurons and Islet-1–negative cells is concentration-

dependent. (B to E) Data represent mean T SEM (error bars) of three or four

independent samples per data point. Concentration-dependence data were

fitted to a Hill equation.

RESEARCH | REPORTS

(extended inhibitory context)/ESE1 (exonic splic-

ing enhancer 1), ESE2, and ESE3/TSL2 (terminal

stem-loop 2) in exon 7 and the intronic splicing

silencer (ISS)–N1 (29–33) in intron 7 in the

context of an SMN2 minigene. Several point

mutations in these elements affected splicing

of the SMN2 exon 7, consistent with published

evidence (fig. S6). However, in all cases, SMN-C3

retained the ability to correct splicing toward

exon 7 inclusion, indicating that these mutations

did not disrupt structures critical for compound

activity. These results point to a more complex

mechanism of action involving additional cis-

and trans-acting factors.

We next investigated whether the SMN splic-

ingmodifiers are active in animalmodels of SMA.

To assess compound activity on SMN2 splicing

and SMN protein in blood and tissues, we used

the C/C-allele SMAmouse model. These animals

have a mild form of SMA and live a normal life

span but show muscle weakness, peripheral

necrosis, and reduced body weight gain (34). In

pharmacokinetic studies, all three compounds

showed excellent brain penetration in adultmice

(fig. S7). The SMN2 pre-mRNA splicing in

whole blood from untreated adult C/C-allele

mice generated ~40% FL and 60% D7 transcripts.

Adult C/C-allele mice were given either 1 dose

(single dosing) or 10 consecutive daily doses (mul-

tiple dosing) of 10mg/kg SMN-C3.Within 4 hours

after single or multiple dosing, SMN2 FL mRNA

increased to 90% of total mRNA, and D7 mRNA

was reduced to 10%. The levels of FL and D7

mRNA returned to baseline within 32 hours in

both dosing regimens (Fig. 3A), suggesting that

the mRNA response kinetics were similar after

single or multiple dosing. The kinetics of the

FL mRNA increase correlated well with that of

SMN-C3 plasma levels (Fig. 3B). SMN protein

levels were increased by ~50 to 70% over base-

line in brain and quadricepsmuscle after a single

oral dose (peaking at 24 hours) and bymore than

200% above baseline in brain and quadriceps

muscle after 10 days of repeated daily oral dosing

(Fig. 3C). We observed similar effects with an-

other compound. SMN-C2 treatment at 20mg/kg

per day shifted alternative SMN2 splicing toward

FL mRNA (fig. S8) in brain and quadriceps mus-

cle and resulted in marked elevation of SMN

protein levels in brain, spinal cord and quadri-

ceps muscle to levels above those measured in

control mice (heterozygous littermates) (fig. S8).

In another experiment in which we treated C/C-

allele mice for 10 days orally with SMN-C3 at

10 mg/kg per day, SMN protein levels became

elevated in all tissues that we evaluated. In mus-

cle tissues, these levels were similar to those mea-

sured in control mice (Fig. 3D). Thus, repeated

oral dosing of adult C/C-allele mice with SMN2

splicingmodifiers can shift SMN2 splicing toward

the production of FLmRNA and thereby increase

SMN protein in multiple relevant tissues in vivo.

We next tested the SMN splicing modifier

compounds in the D7 model of SMA. These mice

have a severe form of SMA and die within 3

weeks after birth (35). Inmice treatedwith SMN-

C3 by intraperitoneal injection once daily from

postnatal day 3 (P3) through P9, SMN protein

levels in the brain and quadriceps muscle were

elevated in a dose-dependent manner by up to

150% (Fig. 3E) and 90% (Fig. 3F), respectively.

Similarly, SMN-C2 treatment induced a dose-

dependent increase in SMN protein levels in

brain and spinal cord (fig. S8).

We next assessed the effect of treatment with

SMN2 splicing modifiers on motor function and

life span of D7 mice. The animals were treated

with SMN-C3 at doses of 0.3, 1, and 3 mg/kg per

day by intraperitoneal injections from P3 through

P23 and thereafter at doses of 1, 3, and 10 mg/kg

per day, respectively, by oral gavage. At P16, vehicle-

treated D7 mice were much smaller than heterozy-

gous littermate controls and appeared moribund.

In contrast, D7 mice treated with the high dose of

SMN-C3 showed a phenotype similar to that of

heterozygous controls (Fig. 3G). SMN-C3 treat-

ment induced a dose-dependent bodyweight gain

in the D7 mice, with some animals showing a

body weight that was ~80% that of heterozygous

controls (Fig. 3H). SMN-C3 normalized the mo-

tor behavior of D7 mice, illustrated by the ability

of the mice to right themselves as quickly as het-

erozygous controls (Fig. 3I) and by their level of

locomotor activity (see movies S1 to S3). Most

importantly, whereas vehicle-treated mice died

within 3 weeks after birthwith amedian survival

of 18 days, SMN-C3 treatment increased survival

in a dose-dependentmanner to amedian survival

time of 28 days in the low-dose (0.3 mg/kg per

day) group. In the two higher-dose groups (1 and

3 mg/kg per day), ~90% of animals survived be-

yond P65 when the study was completed (Fig.

3J). Similar effects were observed with SMN-C2

treatment at 1 mg/kg per day intraperitoneally

(P3 through P23) and 6 mg/kg per day by oral

gavage thereafter with marked body weight gain

and survival beyond 150 days after birthwhen the

study was completed (fig. S8).

690 8 AUGUST 2014 • VOL 345 ISSUE 6197 sciencemag.org SCIENCE

Fig. 2. SMN splicing modifiers demonstrate high specificity as assessed

by RNA sequencing. (A) Difference in total transcript expression of SMN-

C3 (500 nM) versus DMSO-treated SMA type I patient fibroblasts (set to 1 for

each transcript) for 11,714 human genes with an RPKM (reads per kilobase

per million reads) ≥ 1 in either DMSO or SMN-C3 treatment conditions. Abun-

dance of mRNA is shown as log2 fold change (Log2FC) values (0 = no change,

+1 = doubling, –1 = reduction by half). SMN2 is highlighted by the arrow,

showing no significant change in total mRNA abundance. (B) Differential

effects of treatment on individual splice junctions in human transcripts. For

each splice junction, spanning reads were counted in both treated and con-

trol conditions. Affected splice junctions are characterized by either ab-

solute difference in counts (D) or relative changes (Log2FC). The product

p = D x Log2FC was used to rank splice junctions (up-regulated in blue, down-

regulated in red). The top 114 splice junctions with p > 100 are shown

(~300,000 splice junctions analyzed in total).Twenty splice junctions showed

p > 300. Note that two SMN2 transcript variants [NM_022875 (var a) and

NM_022877 (var c)] share the critical target splice junction 5′ of intron 7. For

STRN3, a switch between variants NM_014574 and NM_001083893 is

observed. Fifteen splice junctions belong to PDXDC1 with all splice junctions

similarly affected. PDXDC2P is a pseudogene, probably identified due to the

strong similarity between the PDXDC1 and PDXDC2P transcripts around

that junction (not resolved by the mapping algorithm). Data represent

analysis of RNA sequencing data from three different experimental samples

per group.

RESEARCH | REPORTS

Finally, we demonstrated that this effect on D7

mouse survival and body weight gain was as-

sociated with changes in the hallmark neuro-

muscular pathology of SMA. We evaluated the

number of motoneurons in the spinal cord, neu-

romuscular junction (NMJ) phenotype, and

muscle atrophy in D7 mice treated from P3 to

P14 with SMN-C3 at different doses. In a dose-

dependent manner, SMN-C3 prevented spinal

cord motoneuron loss (Fig. 4, A and B), splenius

capitis NMJ denervation (Fig. 4, C and D), and

extensor digitorum longus (EDL) muscle atro-

phy, with complete prevention at the highest

dose (Fig. 4, E and F). SMN-C3 treatment also

prevented phenotypes associated with SMN

deficiency in the mouse, including tail and eye

necrosis. Evaluation of adult mice at P65 dem-

onstrated that the weight of tibialis anterior

muscle and the innervation of splenius capitis

muscle NMJs were near normal, indicating

that the effect was preserved into adulthood

(fig. S9).

We have shown that treatment with SMN2

splicing modifiers starting early after birth pre-

vents muscle atrophy in amousemodel of severe

SMA. The compounds also increase life span and

body weight gain and preventmotor dysfunction

andneuromuscular deficits into adulthood. In con-

trast to our findings, other small-molecule com-

pounds have exhibited only modest effects on life

span in D7 mice [reviewed in (36)]. Adult D7 mice

treated with SMN-C3 from P3 were near normal,

suggesting that SMN2 splicing modifiers have the

SCIENCE sciencemag.org 8 AUGUST 2014 • VOL 345 ISSUE 6197 691

Fig. 3. SMN splicing modifiers increase SMN protein expression and

provide therapeutic benefit to D7 mice. (A) Relative expression of SMN2

FL and D7 mRNA in whole blood after once daily oral doses of SMN-C3

(10 mg/kg per day) on day 1 (blue) and day 10 (red) in adult C/C-allele mice.

(B) Levels of SMN-C3 in plasma of C/C-allele mice after SMN-C3 treatment

(10 mg/kg per day) on days 1 and 10. (C) SMN protein levels in quadriceps

muscle (open circles) and brain (solid circles) of C/C-allele mice after once

daily oral doses of SMN-C3 (10 mg/kg) on days 1 and 10. (D) SMN protein in

the brain, spinal cord, quadriceps, and longissimus muscle of C/C-allele mice

and heterozygous littermates after 10 daily oral doses of vehicle or SMN-C3

at 10 mg/kg. (E and F) SMN protein levels in the brain (E) and quadriceps

muscle (F) of D7 mice after seven daily intraperitoneal doses (P3 through

P9) of vehicle or SMN-C3 (0.1, 0.3, or 1 mg/kg). (G to J) Mice were treated

from P3 to P23 once daily with vehicle or SMN-C3 by intraperitoneal injection at

0.3, 1, or 3 mg/kg, and thereafter once daily by oral gavage with 1, 3, or 10 mg/kg.

(G) Appearance of a vehicle-treated D7 mouse (D7 Veh), a SMN-C3–treated D7

mouse (D7 SMN-C3), and a vehicle-treated heterozygous mouse (HET Veh). (H)

Body weight from P3 through P60. Numbers at right indicate survivors at P60

among 10 (HET) or 16 (D7)mice per group. (I) Righting reflex of D7mice. Shown is

the mean time (from three trials) for a mouse to right itself after being put onto its

back, assessed at P9 and P16 (also seemovie S1). (J) Kaplan-Meier survival curves

from P3 to P65. (A to F and H to J) RT-PCR and protein data represent means T

SEM (error bars) of four to six animals per data point; data in (G) to (J) from 16 D7

mice and 10 HETmice per group. **P < 0.01 and ***P < 0.001, as assessed by

one-way analysis of variance (ANOVA) followed by Bonferroni test (D) or

Dunnett’s test using vehicle-treated animals (E and F) or HETmice (I) as control.

RESEARCH | REPORTS

potential to provide sustained protection when

administered throughout the life span of amouse.

Splicing modulators have been described that

target various steps of spliceosome assembly

[e.g., modulators that affect splicing at the pre-

spliceosomal complex A, U2 snRNP function, SR

protein activity, and progression of complex A to

complex B (37–43)], thus producing widespread

changes in pre-mRNA splicing. For example,

chlorhexidine, clotrimazole, and flunarizine were

found to affect alternative splicing of 1444, 874,

and 326 transcripts, respectively (43). In contrast,

the splicing modifiers described here are more

selective, affecting the alternative splicing of only

a few transcripts. We have shown that these com-

pounds retain the ability to modify splicing of the

SMN2 pre-mRNA in exon or intron 7 after intro-

ducing a series of singlemutations in these regions.

We speculate that these molecules interact with

specific primary or secondary RNA structures in

the SMN2 pre-mRNA that are distinct and com-

plex (44), or with specific protein-RNA complexes.

Future identification and characterization of the

binding targets of the SMN2 splicing modifiers

will help to elucidate the molecular basis of

their selectivity.

The splicingmodifiers reported here are orally

bioavailable compounds that penetrate into all

of the tissues we tested—including brain, spinal

cord, and muscle—and consequently exert their

action on SMN2 splicing in all cells of the body.

The global correction of SMN2 splicing to in-

crease SMN protein in all tissues is likely to have

advantages relative to modalities that target only

the central nervous system (45–47). The prom-

ising preclinical activity of the SMN2 splicing

modifiers suggests that this strategy may merit

investigation as a possible therapy for patients

with SMA.

REFERENCES AND NOTES

1. J. Pearn, Lancet 315, 919–922 (1980).

2. E. A. Sugarman et al., Eur. J. Hum. Genet. 20, 27–32 (2012).

3. T. O. Crawford, C. A. Pardo, Neurobiol. Dis. 3, 97–110 (1996).

4. J. A. Markowitz, P. Singh, B. T. Darras, Pediatr. Neurol. 46, 1–12

(2012).

5. S. J. Kolb, J. T. Kissel, Arch. Neurol. 68, 979–984 (2011).

6. A. Lewelt, T. M. Newcomb, K. J. Swoboda, Curr. Neurol.

Neurosci. Rep. 12, 42–53 (2012).

7. C. H. Wang et al., J. Child Neurol. 22, 1027–1049 (2007).

8. S. Lefebvre et al., Cell 80, 155–165 (1995).

9. C. L. Lorson, E. Hahnen, E. J. Androphy, B. Wirth, Proc. Natl.

Acad. Sci. U.S.A. 96, 6307–6311 (1999).

10. S. Cho, G. Dreyfuss, Genes Dev. 24, 438–442 (2010).

11. B. G. Burnett et al., Mol. Cell. Biol. 29, 1107–1115 (2009).

12. C. Fallini, G. J. Bassell, W. Rossoll, Brain Res. 1462, 81–92

(2012).

692 8 AUGUST 2014 • VOL 345 ISSUE 6197 sciencemag.org SCIENCE

Fig. 4. SMN splicing modifiers prevent neuromuscular pathology in D7 mice. Animals were treated from P3

through P14 once daily with vehicle or SMN-C3 by intraperitoneal injection at 0.3 or 3 mg/kg.The number of ventral

horn motoneurons in lumbar segments three to five (L3 to L5), the innervation of splenius capitis NMJs, and EDL muscle cross-sectional area were assessed at

P14 by immunohistochemical methods. (A) Immunomicrographs showing L3-to-L5 motoneurons labeled with choline acetyltransferase antibodies. Scale

bar, 100 mm. (B) Quantification of the number of motoneurons in L3-to-L5 spinal cord segments. (C) Immunomicrographs showing the innervation pattern

of the splenius capitis muscle. NMJs are labeled with a-bungarotoxin for acetylcholine receptor clusters (red) and anti-synaptophysin and neurofilament

antibodies for presynaptic nerve terminals (green). Scale bars, 40 mm (upper panel); 20 mm (lower panel). Dashed boxes indicate NMJs shown at a higher

magnification in the lower panel. (D) Quantification of NMJ innervation. (E) Cross sections of EDL muscles. Scale bar, 100 mm. (F) Quantification of the total

cross-sectional area of EDL muscles. (B, D, and F) Data represent means T SEM (error bars) of five to seven mice per group. **P < 0.01 and ***P < 0.001, as

assessed by one-way ANOVA followed by Dunnett’s test using vehicle-treated D7 mice as a control.

RESEARCH | REPORTS

13. B. Schrank et al., Proc. Natl. Acad. Sci. U.S.A. 94, 9920–9925

(1997).

14. S. Paushkin, A. K. Gubitz, S. Massenet, G. Dreyfuss, Curr. Opin.

Cell Biol. 14, 305–312 (2002).

15. A. H. Burghes, C. E. Beattie, Nat. Rev. Neurosci. 10, 597–609

(2009).

16. M. D. Mailman et al., Genet. Med. 4, 20–26 (2002).

17. T. O. Crawford et al., PLOS ONE 7, e33572 (2012).

18. B. Wirth et al., Hum. Genet. 119, 422–428 (2006).

19. M. Feldkötter, V. Schwarzer, R. Wirth, T. F. Wienker, B. Wirth,

Am. J. Hum. Genet. 70, 358–368 (2002).

20. E. Arkblad, M. Tulinius, A. K. Kroksmark, M. Henricsson,

N. Darin, Acta Paediatr. 98, 865–872 (2009).

21. S. J. Kolb et al., BMC Neurol. 6, 6 (2006).

22. K. Takata, T. Shimizu, S. Iwai, R. D. Wood, J. Biol. Chem. 281,

23445–23455 (2006).

23. G. L. Moldovan et al., Mol. Cell. Biol. 30, 1088–1096

(2010).

24. L. Zietlow, L. A. Smith, M. Bessho, T. Bessho, Biochemistry 48,

11817–11824 (2009).

25. A. M. Burroughs et al., Genome Res. 20, 1398–1410

(2010).

26. Z. S. Kai, A. E. Pasquinelli, Nat. Struct. Mol. Biol. 17, 5–10

(2010).

27. T. Katoh et al., Genes Dev. 23, 433–438 (2009).

28. E. R. Kinjo et al., Exp. Neurol. 248, 546–558 (2013).

29. N. N. Singh, R. N. Singh, E. J. Androphy, Nucleic Acids Res. 35,

371–389 (2007).

30. C. L. Lorson, E. J. Androphy, Hum. Mol. Genet. 9, 259–265

(2000).

31. L. Cartegni, M. L. Hastings, J. A. Calarco, E. de Stanchina,

A. R. Krainer, Am. J. Hum. Genet. 78, 63–77 (2006).

32. N. N. Singh, R. N. Singh, RNA Biol. 8, 600–606 (2011).

33. N. N. Singh, E. J. Androphy, R. N. Singh, Crit. Rev. Eukaryot.

Gene Expr. 14, 271–285 (2004).

34. M. Osborne et al., Hum. Mol. Genet. 21, 4431–4447

(2012).

35. T. T. Le et al., Hum. Mol. Genet. 14, 845–857 (2005).

36. J. Seo, M. D. Howell, N. N. Singh, R. N. Singh, Biochim. Biophys.

Acta 1832, 2180–2190 (2013).

37. K. O’Brien, A. J. Matlin, A. M. Lowell, M. J. Moore, J. Biol. Chem.

283, 33147–33154 (2008).

38. A. N. Kuhn, M. A. van Santen, A. Schwienhorst, H. Urlaub,

R. Lührmann, RNA 15, 153–175 (2009).

39. D. Kaida et al., Nat. Chem. Biol. 3, 576–583 (2007).

40. M. Hasegawa et al., ACS Chem. Biol. 6, 229–233

(2011).

41. M. G. Berg et al., Mol. Cell. Biol. 32, 1271–1283 (2012).

42. J. G. Chang et al., PLOS ONE 6, e18643 (2011).

43. I. Younis et al., Mol. Cell. Biol. 30, 1718–1728 (2010).

44. N. N. Singh et al., Nucleic Acids Res. 41, 8144–8165

(2013).

45. M. A. Passini et al., Sci. Transl. Med. 3, 72ra18 (2011).

46. Y. Hua et al., Nature 478, 123–126 (2011).

47. K. Sahashi et al., EMBO Mol. Med. 5, 1586–1601 (2013).

48. F. Degorce et al., Curr. Chem. Genomics 3, 22–32

(2009).

ACKNOWLEDGMENTS

We thank O. Khwaja, C. Czech, T. Kremer, L. Müller, W. Muster,

S. Kirchner, L. Green, E. Pinard, J. Kazenwadel, C. Horn, O. Spleiss,

C. Sarry, P. Kueng, F. Knoflach, L. Himmelein, F. Birzele,

K. von Herrmann, A. Mollin, J. Hedrick, M. Dumble, I. Huq,

P. Martin, D. Mankoff, S. Jung, J. Crona, M. Haley, T. Yang, S. Choi,

S. Hwang, M. Dali, W. Lennox, S. Yeh, J. Yang, J. Petruska,

J. Breslin, J. Baird, B. Scharf, T. Tripodi, G. Ryan, J. Tivade, J. Du,

D. Minn, C. Romfo, and C. Trotta for help and support with this

project. This work was supported by grants from the SMA

Foundation and the Harvard Stem Cell Institute. F. Hoffmann-La

Roche and PTC Therapeutics have filed three patent applications

entitled “Compounds for Treating Spinal Muscular Atrophy”

(WO2013/101974 A1, WO2013/112788 A1, and WO2013/119916 A2).

SUPPLEMENTARY MATERIALS

www.sciencemag.org/content/345/6197/688/suppl/DC1

Materials and Methods

Figs. S1 to S9

Tables S1 to S4

References (49–56)

Movies S1 to S3

23 December 2013; accepted 24 June 2014

10.1126/science.1250127

LIPID CELL BIOLOGY

Polyunsaturated phospholipidsfacilitate membrane deformation andfission by endocytic proteinsMathieu Pinot,1,2 Stefano Vanni,1 Sophie Pagnotta,3 Sandra Lacas-Gervais,3

Laurie-Anne Payet,4 Thierry Ferreira,4 Romain Gautier,1 Bruno Goud,2

Bruno Antonny,1* Hélène Barelli1

Phospholipids (PLs) with polyunsaturated acyl chains are extremely abundant in a few

specialized cellular organelles such as synaptic vesicles and photoreceptor discs, but their

effect on membrane properties is poorly understood. Here, we found that polyunsaturated

PLs increased the ability of dynamin and endophilin to deform and vesiculate synthetic

membranes. When cells incorporated polyunsaturated fatty acids into PLs, the plasma

membrane became more amenable to deformation by a pulling force and the rate of

endocytosis was accelerated, in particular, under conditions in which cholesterol was

limiting. Molecular dynamics simulations and biochemical measurements indicated that

polyunsaturated PLs adapted their conformation to membrane curvature. Thus, by

reducing the energetic cost of membrane bending and fission, polyunsaturated PLs may

help to support rapid endocytosis.

Most cellularmembranes contain phospho-

lipids (PLs) with saturated and mono-

unsaturated acyl chains. However, in a

few specialized organelles, such as syn-

aptic vesicles, up to 80% of PLs contain

at least one polyunsaturated acyl chain (1, 2).

Such high levels suggest that polyunsaturated

lipids might endow membranes with specific

physicochemical properties.

In photoreceptor discs, which are perfectly

flat, polyunsaturated PLs facilitate the conforma-

tional change of rhodopsin (3). The influence of

polyunsaturated PLs on protein machineries that

act on curvedmembranes is unclear. Nevertheless,

exogenous treatment of neurons with polyun-

saturated fatty acids facilitates SNARE (soluble

N-ethylmaleimide–sensitive factor attachment

protein receptor) assembly (4) and the recycling

of synaptic vesicles (5, 6). Furthermore, poly-

unsaturated PLs make pure lipid bilayers more

flexible (7).

We studied the effect of polyunsaturated PLs

on the activity of the guanosine triphosphatase

(GTPase) dynamin and the banana-shaped pro-

tein endophilin, which cooperate in membrane

fission by assembling into spirals around the neck

of membrane buds (8, 9). We used liposomes or

giant unilamellar vesicles (GUVs) with a fixed

composition in terms of PL polar head groups

and varied the ratio between mono and polyun-

saturated PLs (tables S1 and S2; also see sup-

plementary materials and methods). We chose

C16:0-C18:1 and C18:0-C22:6 PLs (Fig. 1A) because

they are the most abundant of each PL class.

Dynamin hydrolyzed guanosine triphosphate

(GTP) 7.5 times faster on large liposomes con-

tainingpolyunsaturatedPLs comparedwithmono-

unsaturated PLs (Fig. 1B and fig. S1A). Moreover,

polyunsaturated PLs eliminated the sharp re-

sponse of dynamin to membrane curvature (Fig.

1C and fig. S1B) (10). Because GTP hydrolysis oc-

curs through contacts between dynaminmolecules

within the assembled spiral (11), these results

suggest that polyunsaturated PLs facilitate dynamin

self-assembly on flat membranes.

Using electron microscopy, we analyzed our

liposomes incubated with dynamin, endophilin,

and GTP, amixture optimal formembrane fission

(12). Before incubation, the liposomes displayed

a similar size distribution (radiusR=30 to 200nm)

(Fig. 1, D and E). After incubation, polyunsatu-

rated liposomes were consumed into small vesi-

cles (R ≈ 20 nm), whereas monounsaturated

liposomes appeared unchanged (Fig. 1, D and E,

and fig. S2, A and D). When GTP hydrolysis was

blocked, characteristic endophilin-dynamin spi-

rals (8, 9) formed on polyunsaturated liposomes,

but not on monounsaturated liposomes (Fig. 1D

and fig. S2, B, C, and E). Thus, polyunsaturated

membranes are sensitized to the mechanical ac-

tivities of the endophilin-dynamin complex.

Adhesion of liposomes to the electron micros-

copy grid can lead to an overestimate of au-

thentic fission owing to membrane breakage on

the stiff support (13). To overcome this caveat, we

used fission assays based on visualization ofmod-

el membranes by fluorescence microscopy. First,

we incubated GUVs with dynamin, endophilin,

and GTP and monitored the GUV diameter over

time (12). GUVs containing polyunsaturated PLs

showed a 14 to 30% decrease in size within

1 hour, suggesting consumption by membrane

fission, whereas GUVsmade ofmonounsaturated

SCIENCE sciencemag.org 8 AUGUST 2014 • VOL 345 ISSUE 6197 693

1Institut de Pharmacologie Moléculaire et Cellulaire,Université Nice Sophia Antipolis and CNRS, 06560 Valbonne,France. 2Unité Mixte de Recherche 144, Institut Curie andCNRS, F-75248 Paris, France. 3Centre Commun deMicroscopie Appliquée, Université Nice Sophia Antipolis, ParcValrose, 06000 Nice, France. 4Signalisation et TransportsIoniques Membranaires, Université de Poitiers and CNRS,Poitiers, France.*Corresponding author. E-mail: antonny@ipmc.cnrs.fr

RESEARCH | REPORTS