Gentamicin B1 is a minor gentamicin component withmajor nonsense mutation suppression activityAlireza Baradaran-Heravia, Jürgen Niessera, Aruna D. Balgia, Kunho Choia, Carla Zimmermana, Andrew P. Southb,Hilary J. Andersona, Natalie C. Strynadkaa, Marcel B. Ballyc, and Michel Robergea,1

aDepartment of Biochemistry and Molecular Biology, University of British Columbia, Vancouver, BC, Canada V6T 1Z3; bDepartment of Dermatology andCutaneous Biology, Thomas Jefferson University, Philadelphia, PA 19107; and cDepartment of Experimental Therapeutics, British Columbia Cancer Agency,Vancouver, BC, Canada V5Z 1L3

Edited by C. Thomas Caskey, Baylor College of Medicine, Houston, TX, and approved February 23, 2017 (received for review December 20, 2016)

Nonsense mutations underlie about 10% of rare genetic disease cases.They introduce a premature termination codon (PTC) and prevent theformation of full-length protein. Pharmaceutical gentamicin, a mixtureof several related aminoglycosides, is a frequently used antibiotic inhumans that can induce PTC readthrough and suppress nonsensemutations at high concentrations. However, testing of gentamicin inclinical trials has shown that safe doses of this drug produce weak andvariable readthrough activity that is insufficient for use as therapy. Inthis study we show that the major components of pharmaceuticalgentamicin lack PTC readthrough activity but the minor componentgentamicin B1 (B1) is a potent readthrough inducer. Molecular dynam-ics simulations reveal the importance of ring I of B1 in establishing aribosome configuration that permits pairing of a near-cognate complexat a PTC. B1 induced readthrough at all three nonsense codons incultured cancer cells with TP53 (tumor protein p53) mutations, in cellsfrom patients with nonsense mutations in the TPP1 (tripeptidyl pepti-dase 1), DMD (dystrophin), SMARCAL1 (SWI/SNF-related, matrix-associated, actin-dependent regulator of chromatin, subfamily a-like1), and COL7A1 (collagen type VII alpha 1 chain) genes, and in anin vivo tumor xenograft model. The B1 content of pharmaceuticalgentamicin is highly variable and major gentamicins suppress thePTC readthrough activity of B1. Purified B1 provides a consistentand effective source of PTC readthrough activity to study the poten-tial of nonsense suppression for treatment of rare genetic disorders.

gentamicin B1 | nonsense mutation | premature stop codon readthrough |rare genetic diseases | cancer

Nonsense mutations, which change an amino acid codon to apremature termination codon (PTC) (TGA, TAG, or TAA),

cause about 10% of rare genetic disease cases (1). Compoundsthat permit insertion of an amino acid at the PTC can enableformation of full-length protein and increased protein function.This strategy, termed nonsense suppression or PTC readthrough,has the potential to treat large numbers of patients across multiplerare genetic disorders. However, drugs that can induce therapeuticlevels of readthrough at safe doses are not yet available.Gentamicin is an approved aminoglycoside antibiotic that can

induce PTC readthrough in cultured cells but only at concentra-tions that are orders of magnitude higher than the 10 μg/mLthreshold in blood above which gentamicin is toxic in humans (2).Clinical trials in patients with Duchenne muscular dystrophy andcystic fibrosis have indicated that therapeutically relevant levelsof readthrough cannot be achieved at subtoxic gentamicin doses(3–5). Moreover, variable response to gentamicin treatment hasbeen observed in animal models and humans (5). Gentamicin isnot a pure compound but a mixture of related aminoglycosidesisolated from Micromonospora purpurea fermentation. It is com-posed principally of major gentamicins C1, C1a, C2, C2a, and C2bas well as minor related aminoglycosides (6).In this study, using a variety of in vitro and in vivo models we

report that the major gentamicins lack PTC readthrough activity butthat gentamicin B1, a minor component of pharmaceutical genta-micin, is potently active.

ResultsGentamicin B1 Is the Major PTC Readthrough Component in PharmaceuticalGentamicin. Gentamicin has been tested for PTC readthrough inmany different cell and animal models and in clinical trials but hasshown a lack of potency and unexplained variability in its effects(5). In our study of PTC readthrough of the tumor suppressor p53,we also observed variation in the effects of gentamicin. Usingfluorescence imaging to measure readthrough at a PTC in theTP53 (tumor protein p53) gene (c.637C > T; p.R213X) in HDQ-P1human cancer cells (7), we examined three different phar-maceutical gentamicin batches. Two showed low activity at1 mg/mL and one showed no activity (Fig. 1 A and B and Fig.S1). Gentamicin is not a pure compound. The US Pharma-copeial Convention (USP) specifies 25–50% gentamicin C1;10–35% gentamicin C1a; and 25–55% gentamicin C2 + C2a,whereas the European Pharmacopoeia requires 20–40% gen-tamicin C1; 10–30% gentamicin C1a; 40–60% gentamicin C2 +C2a + C2b; and no more than 3% sisomycin (6). Sisomycin is oneof several minor components, including gentamicin A, B, B1,garamine, and ring C.To determine whether variability in PTC readthrough could result

from differing proportions of components with varying readthroughpotency, we tested >98% pure samples of the major gentamicins(C1, C1a, C2, C2a, and C2b) for TP53 PTC readthrough (Fig. 1C).All major gentamicins were inactive, even at a high concentration of1 mg/mL (Fig. 1D and Fig. S1). The pharmacopeial conventionsallow small amounts of minor aminoglycosides in pharmaceutical

Significance

The small number of patients for each of the >5,000 rare geneticdiseases restricts allocation of resources for developing disease-specific therapeutics. However, for all these diseases about 10%of patients share a common mutation type, nonsense mutations.They introduce a premature termination codon (PTC) that formstruncated proteins. Pharmaceutical gentamicin, a mixture of sev-eral related aminoglycosides, is an antibiotic frequently used inhumans that shows weak and variable PTC readthrough activity.Using a variety of in vitro and in vivo assays we report that themajor gentamicin components lack PTC readthrough activity butthat a minor component, gentamicin B1, is responsible for most ofthe PTC readthrough activity of this drug and has potential totreat patients with nonsense mutations.

Author contributions: A.B.-H., J.N., M.B.B., and M.R. designed research; A.B.-H., J.N., A.D.B.,K.C., and C.Z. performed research; A.P.S. contributed new reagents/analytic tools; A.B.-H.,J.N., A.D.B., K.C., C.Z., N.C.S., M.B.B., and M.R. analyzed data; and A.B.-H., J.N., H.J.A., andM.R. wrote the paper.

Conflict of interest statement: A.B.-H. and M.R. have an ownership interest in Codon-XTherapeutics. A patent application pertaining to the results presented in the paper hasbeen filed. The authors declare no additional competing financial interests.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.1To whom correspondence should be addressed. Email: michelr@mail.ubc.ca.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1620982114/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1620982114 PNAS | March 28, 2017 | vol. 114 | no. 13 | 3479–3484

GEN

ETICS

See Retraction Published November 26, 2018

Dow

nloa

ded

by g

uest

on

Apr

il 18

, 202

0 D

ownl

oade

d by

gue

st o

n A

pril

18, 2

020

Dow

nloa

ded

by g

uest

on

Apr

il 18

, 202

0

gentamicin (6). Testing >98% pure samples of the minoraminoglycosides found in the gentamicin batches showed thatgentamicin B1 (Fig. 1E and Fig. S1) was potently active, in-ducing much higher levels of PTC readthrough and at muchlower concentrations than active gentamicin batches (Fig. 1 Dand F and Fig. S1). Automated capillary Western analysisdemonstrated that B1 induced the production of large amountsof full-length p53, the readthrough product, as well as a smallerincrease in truncated p53 (Fig. 1G). We also examined theactivity of the purified gentamicin components in a HeLa cell-free translation assay using TP53 mRNA bearing the sameR213X (UGA) nonsense mutation. B1 induced PTC read-through in vitro but equimolar concentrations of the othergentamicins did not (Fig. 1H). These data show that the majorgentamicins present in pharmaceutical gentamicin have noPTC readthrough activity and that any activity displayed bypharmaceutical gentamicin batches is due to the action of aminor component like B1 on the translational machinery.

Molecular Dynamics Simulations Predict Elongation-Like Conformationof A1824 and A1825 rRNA Residues in the Presence of Gentamicin B1.Competition between binding of a near-cognate aminoacyl–tRNAcomplex and the eRF1–eRF3–GTP eukaryotic release factorcomplex to the ribosome A site determines whether readthroughor termination will occur at a PTC (8). Critical to accurate codonrecognition by both complexes is the conformation of A1824 andA1825, two conserved rRNA residues at helix 44 of the ribosomethat can dynamically flip in and out of the helix. Structural studieshave shown that binding of eRF1 flips A1825 out to facilitatestacking of termination codon bases and enable accurate recog-nition of the termination codon, whereas A1824 remains insidethe helix (9). By contrast, A1825 and A1824 adopt a flipped-outconformation upon binding of a cognate tRNA complex, enablingthem to interact with and stabilize the codon–anticodon duplexand ensuring fidelity of tRNA selection (10–12). This conforma-tion is also produced when G418, another aminoglycoside that caninduce PTC readthrough, binds rRNA helix 44 (13).

0

20

40

60

80

100

1 10 100 1000g/ml

Gentamicin B1

Gentamicin A Gentamicin BGentamicin C1Gentamicin C1aGentamicin C2 Gentamicin C2aGentamicin C2b

Garamine

Ring C

Sisomicin

0

20

40

60

80

100

1 10 100 1000

% p

53-p

ositi

ve n

ucle

i

g/ml

Gentamicin-Batch 1 Gentamicin-Batch 2 Gentamicin-Batch 3

Gentamicin (1000 μg/ml)Batch1 Batch2 Batch3

p53

Mer

ged

with

Hoe

chst

A B C

G H

D

p53

Mer

ged

with

Hoe

chst

Gentamicin B1 (μg/ml)100 300

66

40

Gentamicin B1100 300 1000 100 300 1000 100 300 1000 1 3 10 30 50 100 300 (μg/ml)un

treat

ed

Batch 1 Batch 2 Batch 3Gentamicin

116

TR-p53

FL-p53

Vinculin

66

40 TR-p53

FL-p53

untre

ated

Gentam

icin B

1

Garamin

Sisomici

n

Gentam

icin A

Gentam

icin B

Gentam

icin C

1a

Gentam

icin C

1

Gentam

icin C

2b

Gentam

icin C

2a

Gentam

icin C

2

Ring C

O

O

O

O

R1

NH2

HONH2

NH2

OH

CH3

NHH3C

HO

HO

HO

B1B

R1CH3

H

Gentamicin B (μg/ml)1000

C1C1aC2/C2a§

C2b

R1CH3

HCH3

H

R2NHCH3

NH2NH2

NHCH3

O

O

O

O

R1

R2

H2NNH2

NH2

OH

CH3

NHH3C

HO

E F

*

* *

**

*

FL-p53 1 2.5 1.2 1.1 1.2 1.2 1.3 1 1 1.2 0.8 1.1FL-p53 0 0 2.4 24 0 0 4.1 0 6.7 39 0.8 0.7 2.6 25 60 158 514

% p

53-p

ositi

ve n

ucle

i

*

**

Fig. 1. PTC readthrough by gentamicin components. (A, B, D, and F) The proportion of HDQ-P1 cells showing nuclear p53 immunofluorescence, a measure ofreadthrough, after a 72-h exposure to different concentrations of three gentamicin batches or the indicated compounds was determined by automatedfluorescence microscopy (7). A and F show representative images, with p53 immunofluorescence in green and nuclei in blue. (Scale bar, 100 μm.) Completesets of images are shown in Fig. S1. B and D show quantitative measurements in ∼2,000 cells per replicate (mean ± SD, n = 3). *P < 0.01 relative to untreatedsamples. (C and E) Structures of the major gentamicins and of gentamicin B1 and B ( in C, §C2 and C2a are epimers at the ring I 6′ position). (G) The productionof full-length p53 (FL-p53) and truncated p53 (TR-p53) in HDQ-P1 cells exposed for 72 h to different concentrations of three gentamicin batches or B1 wasdetermined by automated capillary electrophoresis Western analysis and the results are displayed as pseudoblots (7). Vinculin was used as a protein loadingcontrol. The numbers indicate the amounts of normalized FL-p53 against vinculin relative to the amount of TR-p53 detected in untreated cells. (H) R213X TP53mRNA was subjected to in vitro translation in the presence of the indicated compounds at 0.5 μM for 20 min and p53 was detected as in G.

3480 | www.pnas.org/cgi/doi/10.1073/pnas.1620982114 Baradaran-Heravi et al.

See Retraction Published November 26, 2018

Dow

nloa

ded

by g

uest

on

Apr

il 18

, 202

0

In ribosome-bound G418, a methyl and a hydroxy substituentat position 6′ of ring I prevent A1824 and A1825 from movingback into the helix (13) (Fig. 2 A and D). Gentamicin B1 alsocontains two substituents at position 6′, a methyl and an aminemoiety (Fig. 1E). To investigate whether gentamicin B1 mightalso affect the movement of these rRNA residues we used mo-lecular dynamics (MD) simulations using as a reference the G418-bound RNA fragment of the yeast ribosome-G418 crystal struc-ture [Protein Data Bank (pdb): 4U4O] (13). This RNA fragmentis identical in yeast and humans. Control simulations using as in-put G418 and its bound rRNA fragment from the crystal structureshowed that the 6′ substituents of G418 sterically prevented bothA1824 and A1825 from moving into the helix (Fig. S2A), consis-tent with the structure (13) (Fig. 2 A and D). The rRNA fragmentresembled the tRNA binding-like conformation throughout thesimulation (Fig. 2G and Fig. S2A). When an additional controlsimulation was carried out without G418, both A1824 andA1825 moved inwards (Fig. 2G and Fig. S2B), indicating that thesimulation is able to distinguish between the different states.We then replaced G418 with B1 and performed a MD simula-

tion. The two 6′ substituents of B1 also prevented A1824 andA1825 from moving back into the helix (Fig. 2 B, E, and G).Moreover, the 6′ amine and the 3′ and 4′ hydroxyl substituents inring I formed specific interactions with the backbone phosphates ofA1824 and A1825 that stabilized the flipped-out conformation(Fig. 2E). Simulation with gentamicin B, which is inactive in PTCreadthrough but differs from B1 only in lacking the 6′ methylsubstituent (Fig. 1D–F), revealed that the absence of the methyl groupwas sufficient to enable A1824 to move in, whereas A1825 remainedout (Fig. 2 C, F, andG), consistent with an eRF1-bound conformation

(9) and termination at a PTC. Interestingly, the absence of the6′ methyl group not only prevented steric blocking of A1824 butalso allowed ring I to move partially below the backbone phos-phates of the rRNA and make specific interactions that stabi-lized the “in” position (Fig. 2F). Stabilization of the flipped-outconformation of both A1824 and A1825 by gentamicin B1 at aPTC likely enhances binding of a near-cognate tRNA and per-mits insertion of an amino acid while also impeding binding ofeRF1 to prevent termination.

Gentamicin B1 Induces PTC Readthrough at All Three Termination Codonsand at Different Positions of the TP53 Gene. The TGA nonsense codonthat is efficiently suppressed by B1 in HDQ-P1 cells is known to bemore susceptible to basal- and gentamicin-induced readthrough thanthe other two nonsense codons TAG and TAA (14). We askedwhether B1 can induce readthrough in TP53-null NCI-H1299 cellstransiently expressing TP53 cDNAs bearing TGA, TAG, or TAA atR213. B1 induced robust readthrough at all three nonsense codonswithout affecting the expression of WT TP53 (Fig. 3A). By com-parison, a 10- or 20-fold higher concentration of gentamicin eliciteda low level of readthrough at TGA only (Fig. 3A). The position of anonsense codon within a gene-coding sequence can also influencePTC readthrough. Five human cancer cell lines bearing homozygousnonsense mutations at different positions in the TP53 gene (7) weretested. Exposure to 100 μg/mL B1 for 3, 6, or 13 d induced varyinglevels of PTC readthrough (full-length p53) and also increasedtruncated p53 in all cell lines (Fig. 3B). By contrast, B1 had no effectin NCI-H1299 TP53-null cells or in HCT116 cells with WT TP53.These findings show that gentamicin B1 is effective on a range ofTP53 nonsense mutations.

A1825

A1824 A1825

A1823

A1824A1825

A1823

A1824

A1825

A1823

A1825

CBA

FED

4

G418 (pdb:4U4O)

Gentamicin B1

Gentamicin B

G418 (pdb:4U4O) Gentamicin

B1Gentamicin

B

ekil-noitanimreTekil-noitagnolEekil-noitagnolE

0 5 10 15 200

200

400

600

800

1000

A1824-C1710distance (Å)

Tim

e (p

s)

G418B1

B

No cpd

G

Fig. 2. Conformations of A1824 and A1825 upon binding of G418, gentamicin B1, and gentamicin B. (A–C) Overview of the effect of binding of G418 (A, green, pdb:4U4O), gentamicin B1 (B, purple,MD simulation), and gentamicin B (C, blue, MD simulation) to rRNA helix 44. B and C depict themost important conformations during the1-ns MD simulation. The rRNA is shown as cartoon except for A1824 and A1825, which are shown as sticks. A1824 and A1825 are flipped out, in an “elongation-like”conformation in the presence of G418 or B1, whereas only A1825 is flipped out, in a “termination-like” conformation, in the presence of gentamicin B. (D–F) Close-up viewof the areas boxed in A–C (rotated ∼90° around the y axis), illustrating interactions (dashed lines) between ring I and rRNA. (G) Distance between atom N6 of A1824 andatom O2 of C1710 on the opposite side of the helix during the course of the simulation (see Fig. S2 for conformations in control simulations without and with G418).

Baradaran-Heravi et al. PNAS | March 28, 2017 | vol. 114 | no. 13 | 3481

GEN

ETICS

See Retraction Published November 26, 2018

Dow

nloa

ded

by g

uest

on

Apr

il 18

, 202

0

Gentamicin B1 Induces PTC Readthrough in Vivo. To investigatewhether B1 is also capable of inducing PTC readthrough in vivo, weexamined its activity in a mouse xenograft model. HDQ-P1 cellsused as our cell culture model do not grow in immunodeficient mice(15). Therefore, we stably transfected NCI-H1299 cells, whichreadily form tumors in mice, with TP53 cDNA bearing the samemutation as HDQ-P1 and verified that B1 induces PTC read-through in this cell line (Fig. S3). Xenografts were then establishedin 129S6/SvEvTac-Rag2tm1Fwa (Rag2M) or NOD-Rag1null IL2rgnull

(NRG) immunodeficient mice and B1 was administered as a single i.p.injection, at either 200 or 400 μg/g. B1 induced robust dose-dependentPTC readthrough in all mice (Fig. 3C, Left). By comparison, similardoses of gentamicin did not induce PTC readthrough (Fig. 3C, Left).Five consecutive daily injections of lower doses of B1 (25 and 50 μg/g)also induced readthrough, whereas gentamicin did not (Fig. 3C,Right).

Gentamicin B1 Induces PTC Readthrough in Rare Genetic Disease Models.In principle, chemicals that induce PTC readthrough could be usedto treat rare genetic disorders caused by nonsense mutations. Toevaluate the potential of B1 as a readthrough treatment, we tested it

in cells from four patients with different genetic diseases: primaryfibroblasts from a patient with neuronal ceroid lipofuscinosis [TPP1(tripeptidyl peptidase 1): p.R127X/R208X], a myoblast line from apatient with Duchenne muscular dystrophy [DMD (dystrophin):p.E2035X], a fibroblast line from a patient with Schimke immuno-osseous dysplasia (SIOD) [SMARCAL1 (SWI/SNF-related, matrix-associated, actin-dependent regulator of chromatin, subfamilya-like 1): p.R17X/R17X], and a keratinocyte line from a patientwith recessive dystrophic epidermolysis bullosa (RDEB) [COL7A1(collagen type VII alpha 1 chain): p.Q251X/Q251X]. TPP1 en-codes the lysosomal enzyme tripeptidyl peptidase 1 and the patient’sfibroblasts displayed no detectable enzyme activity (Fig. 4A). Ex-posure to 25 μg/mL B1 induced a time-dependent increase in ac-tivity that was first detectable at day 3 and by day 10 increased to14% of the activity found in fibroblasts from unaffected individualswith WT TPP1 (Fig. 4A). Consistent with the activity assay, thepatient’s fibroblasts expressed no TPP1 protein in Western analysis,whereas B1 induced a time-dependent increase in both proenzymeand mature forms of TPP1 (Fig. 4B). Similarly, untreated cells frompatients with Duchenne muscular dystrophy, SIOD, and RDEB

66

40

p53-WT Mock p53-R213X (TGA)

p53-R213X (TAG)

p53-R213X (TAA)

untre

ated

B1 50 μ

g/ml

B1 100

μg/m

l

Gent. 1

mg/m

l

untre

ated

B1 50 μ

g/ml

B1 100

μg/m

l

Gent. 1

mg/m

l

untre

ated

B1 50 μ

g/ml

B1 100

μg/m

l

Gent. 1

mg/m

l

untre

ated

B1 50 μ

g/ml

B1 100

μg/m

l

Gent. 1

mg/m

l

untre

ated

B1 50 μ

g/ml

B1 100

μg/m

l

Gent. 1

mg/m

l

FL-p53TR-p53

1 0.8 0.8 1.1 0 0 0 0 0.2 3.3 7.1 0.6 0 2.1 4.1 0.1 0 0.9 1.9 0

0 0 0 0 0 0 0 0 1 1.2 1.4 1 1 1.6 1.8 1.2 1 1.7 1.9 1

FL-p53

TR-p53

66

40

116

66

40

116

66

40

116

66

40

116

66

40

116

66

40

116

66

40

116

SW900(Q167X)

NCI-H1688(Q192X)

ESS-1(R213X)

SK-MES-1(E298X)

HCC1937(R306X)

H1299 (TP53 del.)

HCT116(TP53 WT)

untre

ated

B1 100

μg/m

l

untre

ated

B1 100

μg/m

l

untre

ated

B1 100

μg/m

l

untre

ated

B1 100

μg/m

l

untre

ated

B1 100

μg/m

l

untre

ated

B1 100

μg/m

l

B1 100

μg/m

l

B1 100

μg/m

l

untre

ated

B1 100

μg/m

l

B1 100

μg/m

l

B1 100

μg/m

l

66

40

116

Gentamicin B1 Gentamicin Saline

25 25 25 50 50 50 25 25 25 50 50 50 (μg/g)

TR-p53

FL-p53

Vinculin

FL-p53 0 1 5.5 2.1 2 5.2 8.9 4 1 1.7 1.9 1.7 0.6 2.5

66

40

116

Gentamicin B1 Gentamicin Saline

200 200 200 400 400 400 200 200 200 400 400 400 (μg/g)

TR-p53

FL-p53

Vinculin

* * *FL-p53 1 1.1 12 6.7 8.2 12 20 22 0.5 1.2 1.9 2 2.1 0.7

6 days 13 days3 days 6 days 3d 6d 13d 3d 6d 13d 3 days

A

B

C

Vinculin

Fig. 3. PTC readthrough in cell culture and in vivo. (A) PTC readthrough in TP53-null NCI-H1299 cells transiently transfected with TP53 R213X TGA, TAG, orTAA constructs and exposed for 48 h to B1 or gentamicin (Gent.). Full-length and truncated p53 were quantified relative to the amount of p53 in untreatedcells. p53-WT: R213; mock: transfection reagents only. (B) PTC readthrough in cancer cell lines with different TP53 nonsense mutations exposed to gentamicinB1. The mutations are shown in parentheses. Vinculin was used as a protein loading control. Red arrowhead: full-length p53; black arrowhead: truncated p53.(C) PTC readthrough in NRG mice bearing xenografts of NCI-H1299 cells stably expressing TP53 R213X . Each lane is from an individual mouse xenograft. (Left)Mice were injected once with saline or the indicated concentrations of B1 or USP gentamicin and the amounts of truncated p53 and full-length p53 weremeasured 48 h later. *These mice were killed shortly after administration because of severe toxicity. (Right) Mice were injected with the indicated doses on5 consecutive days and p53 levels were measured 72 h after the last injection. Vinculin was used as a protein loading control. The numbers below the panelsindicate the amounts of normalized FL-p53 against vinculin relative to the ones of saline-treated samples.

3482 | www.pnas.org/cgi/doi/10.1073/pnas.1620982114 Baradaran-Heravi et al.

See Retraction Published November 26, 2018

Dow

nloa

ded

by g

uest

on

Apr

il 18

, 202

0

showed little to no detectable full-length dystrophin, SMARCAL1, orcollagen VII, respectively. Treatment with B1 induced concentration-dependent increases in full-length protein in these cells (Fig. 4C–E).By comparison, much higher concentrations of gentamicin did notinduce production of full-length protein in any of these patient-derived cells (Fig. 4). Exposure to PTC readthrough-inducing con-centrations (≤100 μg/mL) of B1 for 3 d caused a small decrease inthe doubling time of HDQ-P1, NCI-H1299, and fibroblast but notmyoblast cells, comparable to exposure to 1 mg/mL gentamicin (Fig.S4). By contrast, at these concentrations, B1 and gentamicin weretoxic to keratinocytes in culture (Fig. S4). In aggregate, these resultsshow that B1 can induce significant levels of PTC readthrough in anumber of genes and cell types in culture without evident toxicity.

Major Components Present in Pharmaceutical Gentamicin Reduce PTCReadthrough Activity of Gentamicin B1. Although B1 is a knownminor component of the approved drug gentamicin, developing it as atreatment for patients with nonsense mutations would require thesame preclinical and clinical studies as any new drug candidate. On theother hand, repurposing gentamicin as a readthrough drug could be

much more expedient. Having uncovered the central role of B1 inreadthrough, we wondered whether gentamicin batches containing themaximum admissible amount of B1 under USP and European Phar-macopoeia rules might show sufficient readthrough for repurposing.To address this point, we tested whether the activity of B1 is influencedpositively or negatively by the presence of other gentamicin compo-nents. HDQ-P1 cells were coincubated with a fixed concentration ofB1 and increasing concentrations of inactive USP gentamicin. Addi-tion of gentamicin significantly reduced the PTC readthrough activityof B1 (Fig. S5A). This effect was not accompanied by increased toxicityto the cells (Fig. S5B). A similar experiment was carried out by addingincreasing concentrations of individual gentamicin components. Themajor gentamicins C1, C1a, C2, C2a, and C2b and the minor com-ponent sisomicin all reduced the readthrough activity of B1 withoutincreased toxicity (Fig. S5 C and D). These results indicate that unlikepure B1, gentamicin batches with high admissible amounts ofB1 would probably be unsuccessful as therapy agents.

DiscussionReadthrough compounds enable formation of full-length protein viainsertion of an amino acid at the PTC. The application of PTCreadthrough as a therapeutic approach is premised on the insertedamino acid not negatively impacting protein function. The nature ofthe amino acid introduced at PTCs during readthrough stimulated bythe aminoglycoside G418 was recently studied in human HEK293 cellsusing a proteomic approach (16). Readthrough introduced Arg, Trp,and Cys in a 3/1/1 ratio at a UGA PTC, whereas Gln and Tyr wereincorporated in a 8.5/1.5 ratio at UAG, and Gln and Tyr were insertedin equal proportions at UAA (16). The most frequent mutation pro-ducing a PTC is a C-to-T transition at the CGAArg codon, responsiblefor 21% of disease-causing nonsense mutations (1). Readthrough ofthis mutation is predicted to produce a mix of 60% WT protein and40% mutant protein with a Trp or Cys substitution. The second mostcommon mutation (19% of cases) (1), produces TAG from the Glncodon CAG. In this case, readthrough would generate 85% WTprotein and 15% Tyr-substituted protein. In either case, PTC read-through compounds such as B1 would likely produce a preponderanceof WT protein, with a smaller proportion of protein where insertion ofthe incorrect amino acid might be neutral or lead to reduced activity. Inour study of TPP1, in which we measured both protein levels andenzyme activity following readthrough of TGA, the amount of full-length protein (9% of WT after 6 d of treatment and 15% after10 d) correlated well with enzyme activity (6% and 14%, respectively;Fig. 4), indicating that PTC readthrough produces functional TPP1.Another potential concern is promotion of readthrough of nor-

mal termination codons. Unlike PTCs, normal termination codonsare in proximity to 3′-UTR sequences and the poly(A) tail. Poly(A)-binding protein at the poly(A) tail interacts with the release factoreRF3 to stimulate rapid translation termination (reviewed in ref. 17).By contrast, PTCs are farther from the poly(A) tail. This limits theinteraction between poly(A) binding protein and eRF3, leading toribosomal pausing at PTCs, which makes them susceptible to read-through (17). It was also recently found that cells can block theaccumulation of proteins containing C-terminal extensions caused bytranslation past a normal termination codon and into the 3′-UTR bya mechanism that has not yet been elucidated (18). Thus, a combi-nation of mechanisms prevent readthrough at normal terminationcodons. Our examination of five proteins (7 and this study) for whichstrong PTC readthrough was observed did not reveal any increasedsize beyond the full-length protein. This finding agrees with previousobservations suggesting readthrough compounds selectively targetPTCs over normal termination codons (19).Gentamicin is the approved aminoglycoside antibiotic that has

been most extensively tested in clinical trials for treatment of diseasescaused by nonsense mutations. However, gentamicin induces very lowlevels of PTC readthrough and has shown unexplained variability in itseffect (5). Pharmaceutical gentamicin is composed of major and minorcomponents. We found that gentamicin B1, present only in minoramounts in gentamicin, shows potent PTC readthrough activity andthat its activity is reduced in the presence of other gentamicins. Wealso found that the very close structural analog gentamicin B lacks PTC

A

B

C

DMD (E2035X)WT(5%)

untre

ated

B1 25 μ

g/ml

G41

8 10

0 μg

/mlB1 5

0 μg/m

l

B1 100

μg/m

l

Gent. 4

00 μg

/ml

Ata

lure

n 10

μM

Ata

lure

n 25

μMDystrophin

beta-actin

untre

ated

D

WT(10%)

(R17X/R17X)

untre

ated

B1 25 μ

g/ml

B1 50 μ

g/ml

B1 100

μg/m

l

Gent. 4

00 μg

/ml

untre

ated

SMARCAL1

beta-actin

E

TPP1 (R127X/R208X)WT (20%)

untre

ated

B1 25 μ

g/ml

Gent. 1

00 μg

/ml

untre

ated

B1 25 μ

g/ml

B1 25 μ

g/ml

Gent. 1

00 μg

/ml

Gent. 1

00 μg

/ml

3 6 10 3 6 10 days

Pro-TPP1Mature TPP1

Vinculin

Gentamicin B1 (25 g/ml)

Gentamicin (100 g/ml)

02468

10121416

0 2 4 6 8 10

TPP1

act

ivity

(% o

f WT)

days

WT(20%)

(Q251X/Q251X)

untre

ated

B1 5 μg

/ml

B1 10 μ

g/ml

B1 15 μ

g/ml

Gent. 6

0 μg/m

l

untre

ated

Collagen VIIbeta-actin

Fig. 4. Induction of PTC readthrough by gentamicin B1 in cells derived frompatients with rare genetic diseases. (A and B) TPP1 enzyme activity and proteinlevels measured after exposure of GM16485 fibroblasts with TPP1 nonsensemutations to B1 or gentamicin for up to 10 d. (A) TPP1 activity was determinedand expressed relative to the average activity of untreated primary fibroblastsfrom two unaffected individuals (WT). (B) The same cell extracts were analyzedfor production of TPP1 by automated capillary electrophoresis Western analysis.Extracts from WT fibroblasts were also analyzed, using 20% of the amount ofprotein used for GM16485. Vinculin was used as a loading control. (C) Full-lengthdystrophin protein levels measured after HSK001 myoblasts with a DMD nonsensemutation were differentiated into myotubes and exposed to B1 or gentamicin for3 d. Extracts from WT myotubes were also analyzed, using 5% of the amount ofprotein used for Duchenne muscular dystrophy (DMD) cells. β-Actin was used as aloading control. (D) Full-length SMARCAL1 protein levels measured after exposureof SD123 fibroblasts with SMARCAL1 nonsense mutations to gentamicin B1 orgentamicin for 6 d. Extracts fromWT fibroblasts were also analyzed, using 10% ofthe amount of protein used for SIOD cells. β-Actin was used as a loading control.(E) Full-length collagen VII measured after exposure of EB14 keratinocytes withCOL7A1 nonsense mutations to B1 or gentamicin for 72 h. Extracts from WTkeratinocytes were also analyzed, using 20% of the amount of protein used forEB14 cells. β-Actin was used as a loading control.

Baradaran-Heravi et al. PNAS | March 28, 2017 | vol. 114 | no. 13 | 3483

GEN

ETICS

See Retraction Published November 26, 2018

Dow

nloa

ded

by g

uest

on

Apr

il 18

, 202

0

readthrough activity. Our molecular dynamics simulations show that,whereas B1 stabilizes the flipped-out conformation of both A1824 andA1825 rRNA residues and favors an elongation-like conformation,lack of the 6′ methyl substituent in ring I in gentamicin B enablesresidue A1824 to move in, which favors the termination-like confor-mation, highlighting the critical importance of the 6′ methyl group.Unlike pharmaceutical gentamicin, B1 can induce PTC read-

through at all three nonsense mutations (TGA > TAG > TAA),in multiple genes (TP53, TPP1, DMD, SMARCAL1, and COL7A1),in multiple cancer- and patient-derived cells, and also in a xenograftmouse model. These observations offer hope that purified B1 mightshow potential as a therapy agent alone or in combination withrecently discovered compounds that potentiate PTC readthrough byaminoglycosides (7).

Materials and MethodsAutomated p53 Immunofluorescence 96-Well Plate Assay. The assay was per-formed using a Cellomics ArrayScan VTI automated fluorescence imager aspreviously described (7).

Automated Capillary Electrophoresis Western Analysis. The assays for p53 andTPP1 detection were performed as previously described (7).

SDS/PAGE and Immunoblotting. Western blotting for SIOD fibroblasts wasperformed as previously described (7). RDEB keratinocytes were treated withvarious concentrations of gentamicin B1 or gentamicin for 72 h, lysed, and20 μg protein from each lysate was separated on a 5% polyacrylamide gel andelectroblotted onto a nitrocellulose membrane. The membrane was incubatedin a denaturation buffer (50 mM Tris·HCl, pH 8.3, 1 M guanidine HCl, 50 mMDTT, and 2 mMEDTA) for 1 h at room temperature. After a brief rinse with Tris-buffered saline (TBS) the membrane was incubated in a renaturation buffer[10 mM Tris·HCl, pH 7.4, 140 mM NaCl, 2 mM DTT, 2 mM EDTA, 1% BSA, 0.1%Nonidet P-40] overnight at 4 °C followed by 1 h blocking with 5% (wt/vol)nonfat milk in TBS. The membrane was incubated with primary and secondaryantibodies and visualized using enhanced chemiluminescence.

In Vitro Transcription and Translation Assays. In vitro transcription andtranslation assays were performed as previously described (7) with minormodifications. For further details please refer to SI Materials and Methods.

TPP1 Activity Assay. GM16485 primary fibroblasts were exposed to genta-micin sulfate or B1 for up to 10 d with medium and compounds replenished

every 3 d. At the end of the experiment the cells were lysed and TPP1 enzymeactivity was determined as previously described (7).

Molecular Dynamics Simulations. Molecular dynamics simulations were per-formed using GROMACS (20, 21) as described previously (22, 23), with case-specific modifications. For details refer to SI Materials and Methods.

Mouse Experiments. The mouse studies were approved by the University ofBritish Columbia Animal Care Committee and performed in accordance withthe Canadian Council on Animal Care guidelines (protocol A14-0290). MaleNRG (NOD-Rag1null IL2rgnull) or Rag2M (129S6/SvEvTac-Rag2tm1Fwa) mice (22–30 g)were inoculated s.c. in the lower back with 2 million NCI-H1299 cells stablyexpressing TP53 R213X in a volume of 50 μL. The site of tumor cell inoculationwas monitored daily. When tumors becamemeasurable, their size was estimatedby converting dimensions (in millimeters) measured by caliper to tumor weight(in milligrams) using the equation: length × (width2) ÷ 2. The mice were injectedintraperitoneally with gentamicin B1 dissolved in 0.9% NaCl or the vehicle con-trol (0.9% NaCl). The health status of the animals was monitored daily followingan established standard operating procedure. In particular, signs of ill healthwere based on body weight loss, change in appetite, and behavioral changessuch as altered gait, lethargy, and gross manifestations of stress. The monitoringstaff were blinded to the treatment groups. When signs of severe toxicity werepresent, the animals were killed (isoflurane overdose followed by CO2 asphyxi-ation) for humane reasons. The tumors were excised, cut in two, weighed, andsnap frozen in liquid nitrogen. Tumor pieces were thawed in cold lysis buffer,minced into small pieces using scissors, and homogenized using 20 strokes of aloose-fitting piston in a Dounce homogenizer on ice. The homogenized tissuewas kept on ice for 1 h, vortexing every 10 min, and centrifuged at 15,000 × g for15 min at 4 °C. The supernatants were transferred to fresh tubes and proteinconcentration was measured by Bradford assay (Bio-Rad).

Statistical Analysis. Data are presented as mean ± SD unless otherwise stated.Comparisons were made using the two-tailed Student’s t test and differ-ences were considered significant at a P value of <0.01.

Further experimental details are provided in SI Materials and Methods.

ACKNOWLEDGMENTS. We thank Vincent Mouly and Association Françaisecontre les Myopathies (Myobank) for myoblasts and Cornelius Boerkoel forcells, an antibody, and useful comments. This research was funded by theCanadian Cancer Society (Grants 702681 and 704700), Genome BC (GrantPOC027), the Canadian Institutes of Health Research (Grant PPP144245),Mitacs (Award IT03752), and the Little Giants Foundation.

1. Mort M, Ivanov D, Cooper DN, Chuzhanova NA (2008) A meta-analysis of nonsensemutations causing human genetic disease. Hum Mutat 29(8):1037–1047.

2. Electronic Medicines Compendium (2017) Gentamicin 10mg/ml solution for injectionor infusion. Available at https://www.medicines.org.uk/emc/medicine/32933. AccessedMarch 3, 2017.

3. Malik V, et al. (2010) Gentamicin-induced readthrough of stop codons in Duchennemuscular dystrophy. Ann Neurol 67(6):771–780.

4. Clancy JP, et al. (2001) Evidence that systemic gentamicin suppresses premature stopmutations in patients with cystic fibrosis. Am J Respir Crit Care Med 163(7):1683–1692.

5. Linde L, Kerem B (2008) Introducing sense into nonsense in treatments of humangenetic diseases. Trends Genet 24(11):552–563.

6. Stypulkowska K, Blazewicz A, Fijalek Z, Sarna K (2010) Determination of gentamicinsulphate composition and related substances in pharmaceutical preparations by LCwith charged aerosol detection. Chromatographia 72(11–12):1225–1229.

7. Baradaran-Heravi A, et al. (2016) Novel small molecules potentiate premature ter-mination codon readthrough by aminoglycosides. Nucleic Acids Res 44(14):6583–6598.

8. Dabrowski M, Bukowy-Bieryllo Z, Zietkiewicz E (2015) Translational readthroughpotential of natural termination codons in eucaryotes: The impact of RNA sequence.RNA Biol 12(9):950–958.

9. Brown A, Shao S, Murray J, Hegde RS, Ramakrishnan V (2015) Structural basis for stopcodon recognition in eukaryotes. Nature 524(7566):493–496.

10. Ogle JM, et al. (2001) Recognition of cognate transfer RNA by the 30S ribosomalsubunit. Science 292(5518):897–902.

11. Budkevich TV, et al. (2014) Regulation of the mammalian elongation cycle by subunitrolling: A eukaryotic-specific ribosome rearrangement. Cell 158(1):121–131.

12. Ogle JM, Murphy FV, Tarry MJ, Ramakrishnan V (2002) Selection of tRNA by the ri-bosome requires a transition from an open to a closed form. Cell 111(5):721–732.

13. Garreau de Loubresse N, et al. (2014) Structural basis for the inhibition of the eu-karyotic ribosome. Nature 513(7519):517–522.

14. Floquet C, Hatin I, Rousset J-P, Bidou L (2012) Statistical analysis of readthrough levelsfor nonsense mutations in mammalian cells reveals a major determinant of responseto gentamicin. PLoS Genet 8(3):e1002608.

15. Wang CS, et al. (2000) Establishment and characterization of a new cell line derivedfrom a human primary breast carcinoma. Cancer Genet Cytogenet 120(1):58–72.

16. Roy B, et al. (2016) Ataluren stimulates ribosomal selection of near-cognate tRNAs topromote nonsense suppression. Proc Natl Acad Sci USA 113(44):12508–12513.

17. Keeling KM, Xue X, Gunn G, Bedwell DM (2014) Therapeutics based on stop codonreadthrough. Annu Rev Genomics Hum Genet 15:371–394.

18. Arribere JA, et al. (2016) Translation readthrough mitigation. Nature 534(7609):719–723.19. Welch EM, et al. (2007) PTC124 targets genetic disorders caused by nonsense muta-

tions. Nature 447(7140):87–91.20. Hess B, Kutzner C, van der Spoel D, Lindahl E (2008) GROMACS 4: Algorithms for

highly efficient, load-balanced, and scalable molecular simulation. J Chem TheoryComput 4(3):435–447.

21. Pronk S, et al. (2013) GROMACS 4.5: A high-throughput and highly parallel opensource molecular simulation toolkit. Bioinformatics 29(7):845–854.

22. Tou WI, Chang S-S, Lee C-C, Chen CY-C (2013) Drug design for neuropathic painregulation from traditional Chinese medicine. Sci Rep 3:844.

23. Aghdam EM, Hejazi ME, Hejazi MS, Barzegar A (2014) Riboswitches as potentialtargets for aminoglycosides compared with rRNA molecules: In silico studies. J MicrobBiochem Technol S9:002.

24. Pourreyron C, et al. (2014) High levels of type VII collagen expression in recessive dystrophicepidermolysis bullosa cutaneous squamous cell carcinoma keratinocytes increases PI3K andMAPK signalling, cell migration and invasion. Br J Dermatol 170(6):1256–1265.

25. Pettersen EF, et al. (2004) UCSF Chimera: A visualization system for exploratory re-search and analysis. J Comput Chem 25(13):1605–1612.

26. Zoete V, Cuendet MA, Grosdidier A, Michielin O (2011) SwissParam: A fast force fieldgeneration tool for small organic molecules. J Comput Chem 32(11):2359–2368.

27. Bjelkmar P, Larsson P, Cuendet MA, Hess B, Lindahl E (2010) Implementation of theCHARMM force field in GROMACS: Analysis of protein stability effects from correctionmaps, virtual interaction sites, and water models. J Chem Theory Comput 6(2):459–466.

28. Jorgensen WL, Chandrasekhar J, Madura JD, Impey RW, Klein ML (1983) Comparisonof simple potential functions for simulating liquid water. J Chem Phys 79:926–935.

29. Darden TA, Pedersen LG (1993) Molecular modeling: An experimental tool. EnvironHealth Perspect 101(5):410–412.

30. Hess B, Bekker H, Berendsen HJC, Fraaije JEGM (1997) LINCS: A linear constraint solverfor molecular simulations. J Comput Chem 18:1463–1472.

3484 | www.pnas.org/cgi/doi/10.1073/pnas.1620982114 Baradaran-Heravi et al.

See Retraction Published , November 26 2018

Dow

nloa

ded

by g

uest

on

Apr

il 18

, 202

0

Retraction

GENETICSRetraction for “Gentamicin B1 is a minor gentamicin componentwith major nonsense mutation suppression activity,” by AlirezaBaradaran-Heravi, Jürgen Niesser, Aruna D. Balgi, Kunho Choi,Carla Zimmerman, Andrew P. South, Hilary J. Anderson, Na-talie C. Strynadka, Marcel B. Bally, and Michel Roberge, whichwas first published March 13, 2017; 10.1073/pnas.1620982114(Proc Natl Acad Sci USA 114:3479–3484).The authors wish to note the following: “We have now de-

termined that the gentamicin B1 compound we acquired com-mercially and used in our study was not gentamicin B1 but theclosely related aminoglycoside G418. At the time we carried outthe work, the only commercial source for gentamicin B1 wasMicroCombiChem. The company provided two batches ofcompound, purified from gentamicin sulfate complex c, withcertificates of analysis verifying the compound to be >97% puregentamicin B1, together with HPLC-MS and NMR analysisshowing the data conformed to gentamicin B1. Moreover, weasked a third party to carry out independent NMR analysis andthey determined that the MicroCombiChem compound had anitrogen in the vicinity of a methyl group, which is the case forgentamicin B1 but not for G418.“In July 2018, while carrying out structure–activity relationship

studies, we observed that a newly synthesized aminoglycosidederivative containing ring 1 of gentamicin B1 unexpectedlylacked nonsense suppression activity. This finding made us sus-pect the nature of the compound purchased from Micro-CombiChem. At this time, gentamicin B1 became available froma second commercial source—Toronto Research Chemicals.Chemistry collaborators David Powell, David Williams, andRaymond Andersen graciously agreed to carry out NMR analysisof G418 disulfate salt from Sigma, gentamicin B1 free base fromMicroCombiChem and gentamicin B1 diacetate salt from TorontoResearch Chemicals. They determined that gentamicin B1 providedby MicroCombiChem was G418 whereas gentamicin B1 provided byToronto Research Chemicals was composed principally (>95%) ofgentamicin B1.“We note that gentamicin B1 and G418 are closely related

compounds, having the same atomic composition and molecularmass but differ in the location of amine and hydroxyl functionalitiesin ring 1. Ascertaining the structure of gentamicin B1 and identifyingthe NMR signals that distinguish it from G418 was not a trivial taskbecause there are no published NMR data for gentamicin B1, and itwas observed that the resonances of methylene protons vicinal toamino nitrogens in G418 and gentamicin B1 shifted considerablydepending on pH and solvent. This could have led to the mis-identification of G418 as gentamicin B1. Details of the NMR andbiological analyses have been deposited in bioRxiv.“All biological data presented in the PNAS paper (Figs. 1, 3, and

4, and Figs. S1, S3, S4, and S5) are accurate except that the resultsattributed to gentamicin B1 should now be considered to pertain toG418. We wish to apologize to the scientific community for anyadverse consequences that may have resulted from following ourwork. Accordingly, we hereby retract the article.”

Published under the PNAS license.

Published online November 26, 2018.

www.pnas.org/cgi/doi/10.1073/pnas.1818172115

www.pnas.org PNAS | December 11, 2018 | vol. 115 | no. 50 | E11885

RETR

ACT

ION