Published online 2 September 2011 Nucleic Acids Research, 2011, Vol. 39, No.

21 e142
doi:10.1093/nar/gkr695

Generating the optimal mRNA for therapy:

HPLC purification eliminates immune activation
and improves translation of nucleoside-modified,
protein-encoding mRNA
Katalin Karikó1, Hiromi Muramatsu1, János Ludwig2 and Drew Weissman3,*
1
Department of Neurosurgery, University of Pennsylvania, Philadelphia, PA, USA, 2Institute of Clinical Chemistry

Downloaded from https://academic.oup.com/nar/article/39/21/e142/1104771 by guest on 29 December 2021

and Pharmacology, University of Bonn, Bonn, Germany and 3Department of Medicine, University of
Pennsylvania, Philadelphia, PA, USA

Received May 26, 2011; Revised August 9, 2011; Accepted August 10, 2011

ABSTRACT express therapeutic proteins (4). The recognition that the

immunogenicity of RNA could be reduced by the incorp-
In vitro-transcribed mRNA has great therapeutic po- oration of modified nucleosides with a concomitant
tential to transiently express the encoded protein increase in translation (5), potentially allows efficient ex-
without the adverse effects of viral and DNA-based pression of intra and extracellular proteins in vivo and ex
constructs. Mammalian cells, however, contain RNA vivo without activation of innate immune pathways.
sensors of the innate immune system that must Unfortunately, modified nucleoside-containing RNA
be considered in the generation of therapeutic transcribed by phage RNA polymerase transcription still
RNA. Incorporation of modified nucleosides both retains a low level of activation of such pathways (3,5–7).
reduces innate immune activation and increases The remaining activation of RNA sensors by nucleoside
translation of mRNA, but residual induction of modified RNA could be because the modifications do not
type I interferons (IFNs) and proinflammatory cyto- completely suppress the RNAs ability to activate sensors
or due to contaminants with structures that activate in the
kines remains. We identify that contaminants,
presence of nucleoside modification. It is well established
including double-stranded RNA, in nucleoside- that RNA transcribed in vitro by phage polymerase
modified in vitro-transcribed RNA are responsible contains multiple contaminants, including short RNAs
for innate immune activation and their removal by produced by abortive initiation events (8) and double-
high performance liquid chromatography (HPLC) stranded (ds)RNAs generated by self-complementary 30 ex-
results in mRNA that does not induce IFNs and in- tension (9), RNA-primed transcription from RNA tem-
flammatory cytokines and is translated at 10- to plates (10) and RNA-dependent RNA polymerase
1000-fold greater levels in primary cells. Although activity (11).
unmodified mRNAs were translated significantly Large quantities of RNA can be easily prepared by
better following purification, they still induced in vitro transcription from DNA templates using phage
high levels of cytokine secretion. HPLC purified RNA polymerase or solid-phase chemical synthesis. For
uses that require further purification, such as NMR (12),
nucleoside-modified mRNA is a powerful vector for
crystallography (13) and therapeutic applications (14), a
applications ranging from ex vivo stem cell gener- number of techniques have been developed. Preparative
ation to in vivo gene therapy. denaturing polyacrylamide gel electrophoresis is
commonly used to purify in vitro-transcribed RNA,
however, this method is suitable only for short RNAs
INTRODUCTION
[reviewed in (15)]. Long RNAs can be separated on
Our understanding of the importance of RNA in biologic- denaturing agarose gels, but they are not translatable
al processes and the therapeutic potential has substantially due to covalent modifications introduced by the denatur-
increased with the discovery of non-coding regulatory ants glyoxal and formaldehyde (16). Chromatography
RNAs. The use of mRNA has also expanded, including based on size exclusion can efficiently remove unincorpor-
the delivery of mRNA to generate induced pluripotent ated nucleoside triphosphates, small abortive transcripts
stem (iPS) cells (1–3) and in vivo administration to and plasmid template from the desired RNA product

*To whom correspondence should be addressed. Tel: +1 215 573 8491; Fax: +215 349 5111; Email: dreww@mail.med.upenn.edu

ß The Author(s) 2011. Published by Oxford University Press.

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/
by-nc/3.0), which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
e142 Nucleic Acids Research, 2011, Vol. 39, No. 21 PAGE 2 OF 10

under native conditions (17,18), but is limited in its ability polystyrene-divinylbenzene copolymer microspheres
to remove contaminants with similar sizes and contamin- (2.1 mm) (21  100 mm column). Buffer A contained
ants complementary to the RNA selected to purify. No 0.1 M triethylammonium acetate (TEAA), pH = 7.0 and
technique has been reported for purification and prepara- Buffer B contained 0.1 M TEAA, pH = 7.0 and 25%
tive isolation of long in vitro-transcribed mRNA that acetonitrile (Transgenomics). Columns were equilibrated
removes contaminating complementary strands and pre- with 38% Buffer B, loaded with RNA and run with a
serves its translatability. single or 2 linear gradients to 55 or 65% Buffer B over
The development of mRNA to use as a tool to replace 20–30 min at 5 ml/min. RNA analyses were performed
intra- and extracellular proteins in vivo and to with the same column matrix and buffer system using a
transdifferentiate, reprogram and differentiate cells ex 7.8 mm  50 mm column at 1.0 ml/min.
vivo requires the RNA to have high translatability and
no RNA sensor activation. In this report, we identify RNA isolation from column fractions

Downloaded from https://academic.oup.com/nar/article/39/21/e142/1104771 by guest on 29 December 2021

that contaminants from in vitro-transcribed RNA are a RNA content from desired fractions was concentrated and
source of innate immune activation and their removal in- desalted using Amicon Ultra-15 centrifugal filter units
creases RNA translation and eliminates type I interferon (30K membrane) (Millipore) by successive centrifugation
and inflammatory cytokine secretion. at 4000g for 10 min (4 C) in a Sorvall ST16R centrifuge
(Thermo Scientific) and dilution with nuclease free water.
MATERIALS AND METHODS The RNA was recovered by overnight precipitation at
20 C in NaOAc (0.3 M, pH 5.5), isopropanol
Cells (1 volume) (Fisher) and glycogen (3 ml) (Roche).
Human embryonic kidney 293T cells (American Type
Culture Collection) were cultured in Dulbecco’s modified Dot blot
Eagle’s medium (DMEM) supplemented with 2 mM L-glu- RNA (200 ng) was blotted onto super charged Nytran,
tamine (Life Technologies) and 10% fetal calf serum dried, blocked with 5% non-fat dried milk in TBS-T
(FCS) (HyClone) (complete medium). Human and buffer (50 mM Tris–HCl, 150 mM NaCl, 0.05%
murine dendritic cells (DCs) were generated as described Tween-20, pH 7.4), and incubated with dsRNA-specific
(5). Human keratinocytes were obtained from the Skin mAb J2 or K1 (English & Scientific Consulting) for
Disease Research Core (Penn) and grown in MCDB 60 min. Membranes were washed six times with TBS-T
with bovine pituitary extract (140 mg/ml) (Sigma) and and reacted with HRP-conjugated donkey anti-mouse Ig
70 mM Ca++ on collagen (0.01 mg/ml) (Invitrogen) (Jackson Immunology), washed six times and detected
coated plates. with ECL Plus Western blot detection reagent
(Amersham). Images were captured on a Fujifilm
mRNA synthesis LAS1000 digital imaging system. dsRNA (25 ng) used as
mRNAs were transcribed as previously described (5), a positive control was derived from sense and antisense
using linearized plasmids encoding firefly luciferase strands of T7TS UTR sequence (328 bp). Blots were
(pT7TSLuc and pTEVLuc), codon-optimized murine reprobed with 32P-labeled DNA complementary to the
erythropoietin (pTEVmEPO), enhanced green fluorescent 30 -UTR of the RNA to document the presence of RNA.
protein (pTEVeGFP), Metridia luciferase (pT7TSMetluc)
or Renilla luciferase (pT7TSRen and pTEVRen) and T7 Complexing of RNA
RNA polymerase (Megascript, Ambion). All mRNAs were Lipofectin (Invitrogen) complexing was performed as
transcribed to contain 30 or 51-nt long poly(A) tails. described previously (5) using 0.8 ml of Lipofectin and
Additional poly(A) tail was added with yeast poly(A) poly- 0.1 mg of RNA per well of a 96-well plate. Complexing
merase (USB) and noted as An. Triphosphate-derivatives of of RNA to TransIT mRNA (Mirus Bio) was performed
pseudouridine ( ) and 5-methylcytidine (m5C) (TriLink) according to the manufacturer combining RNA (0.1 mg)
were used to generate modified nucleoside containing with TransIT mRNA (0.3 ml) and boost (0.2 ml) reagents.
RNA. All RNAs were capped using the m7G capping
kit with or without 20 -O-methyltransferase (ScriptCap, Cell transfections
CellScript) to obtain cap1 or cap0. We did not observed
For Lipofectin complexed RNA, medium was removed
differences in the immunogenicity of cap0- and cap1-
and 50 ml of complexed RNA was added to 5 x 104 293T
containing nucleoside-modified RNAs. All RNAs were
or DCs per well. Cells were incubated for 1 h and the
analyzed by denaturing or native agarose gel electrophor-
Lipofectin-RNA mixture was replaced with 200 ml
esis. Pseudouridine-modified mRNAs encoding KLF4,
complete medium. For TransIT complexed RNA, 17 ml
LIN28, cMYC, NANOG, OCT4 and SOX2 were a kind
of complex was added to cells, 293T, DCs, or 2  105 kera-
gift of CellScript, Inc.
tinocytes cultured in 183 ml complete medium. Cells were
lysed in firefly or Renilla specific lysis reagents (Promega)
HPLC purification of RNA
at 24 h post RNA addition. Aliquots were assayed for
RNA was purified by High performance liquid enzyme activities using firefly and Renilla luciferase
chromatography (HPLC) (Akta Purifier, GE Healthcare) assay systems (Promega) and a LUMAT LB 950
using a column matrix of alkylated non-porous luminometer (Berthold/EG&G; Wallac). Expression of
PAGE 3 OF 10 Nucleic Acids Research, 2011, Vol. 39, No. 21 e142

eGFP in DCs was documented using an inverted

epifluorescent Nikon microscope mounted with a Nikon
D40 digital camera. Murine EPO protein was measured
with a specific ELISA assay (R&D Systems).

RNA immunogenicity analyses

DCs (murine or human) (5  104 cells/well) in 96-well
plates were treated with medium, R-848 (Invivogen), or
Lipofectin- or TransIT-complexed RNA or poly(I:C)
(Sigma). Supernatant was harvested after 24 h and the
levels of IFN-a, IFN-b (PBL InterferonSource), or
TNF-a (Biosource International) were measured by

Downloaded from https://academic.oup.com/nar/article/39/21/e142/1104771 by guest on 29 December 2021

ELISA.

Gene array analysis

Human DCs from three donors were generated in 5%
FCS. Cells (1  106 DCs/well of a 6-well plate) were
treated with TransIT-complexed TEVRenA51 RNA with
or without modification and with or without purification.
Six hours later, RNA was isolated using RNeasy (Qiagen).
RNA was amplified with the TargetAmp Nano-g
Biotin-aRNA labeling kit (Epicentre) and analyzed on
an Illumina Human HT12v4 chip in an Illumina
BeadStation 500GX. Raw data was processed by the
Bead Studio v.3.0 software. Levels in untreated DC were
used as the baseline for the calculation of fold increase.

Northern blot
Samples were processed and analyzed as previously
described (6). Probes were derived from plasmids and
were specific for the coding regions of human IFN-a13,
IFN-b (Open Biosystems), TNF-a, or GAPDH (ATCC).

RESULTS
A dot blot assay with J2 and K1 monoclonal antibodies
(mAbs) that recognize dsRNA (19) was used to determine
whether in vitro-transcribed RNA contains dsRNA. These
mAbs recognize continuous double stranded structure
of at least 40 bp in length (20), which is not found in Figure 1. In vitro-transcribed RNA is immunogenic and contains
any of the coding sequences or UTRs in the mRNAs dsRNA contaminants. (A) 200 ng of in vitro transcripts encoding
mEPO and containing the indicated modified nucleosides were
analyzed in this study. Testing mammalian and re- blotted and analyzed with K1 and J2 dsRNA-specific mAbs. The
porter protein-encoding in vitro transcripts containing dsRNA positive control contained a 328 bp long dsRNA (25 ng).
either no nucleoside modifications, pseudouridine- ( ), (B) DCs were treated with Lipofectin-complexed Renilla luciferase
or 5-methylcytidine- (m5C) and - (m5C/ ) nucleoside (T7TSRenA30), firefly and Metridia luciferases (T7TSLucA30,
modifications, we found that all samples contained T7TSMetlucA30), and mEPO (TEVmEPOA51) mRNAs. TNF-a levels
were measured in the supernatants at 24 h. (C) DCs were treated with
dsRNA contamination (Figure 1A and data not shown). TransIT-complexed in vitro transcripts encoding Renilla and firefly
Recognition of dsRNA by J2 mAb was not affected by the luciferases (T7TSRenA30, T7TSLucA30), eGFP (TEVeGFPA51) and
presence of modified nucleosides, while K1 had reduced mEPO (TEVmEPOA51). IFN-a levels were measured in the super-
binding to dsRNA containing or m5C/ nucleoside natants at 24 h. Error bars are standard error of the mean. Data
shown is from one experiment that is representative of 3–6.
modifications.
Others and we have previously demonstrated that in-
corporation of modified nucleotides into RNA reduced
its ability to activate RNA sensors including Toll-like of RNAs with coding sequences for mammalian and
receptor (TLR)3, TLR7 and TLR8 (21), retinoic acid reporter proteins flanked by different 50 - and 30 -UTRs
inducible gene I (RIG-I) (22) and RNA-dependent were analyzed. The RNAs were cell-delivered following
protein kinase (PKR) (6,23). Monocyte-derived DCs complexing with Lipofectin, a cationic lipid, or TransIT,
that express these and all other known RNA sensors a membrane active polymer and lipid mixture. RNA com-
(24) were used to measure residual immune activation plexed with Lipofectin induced high levels of TNF-a and
present in modified nucleoside-containing RNA. A series moderate levels of IFN-a, while RNA complexed with
e142 Nucleic Acids Research, 2011, Vol. 39, No. 21 PAGE 4 OF 10

TransIT induced low levels of TNF-a and high levels of HPLC-purified - or m5C/ -modified RNAs that were
IFN-a (data not shown) with some donor-dependent vari- complexed with Lipofectin or TransIT, respectively
ation. Typically, Lipofectin-complexed RNA with or (Figure 3C and D and data not shown). Similarly, no
m5C/ modifications induced less TNF-a. (Figure 1B), cytokine induction could be detected when HPLC-
while TransIT-complexed RNA with or without nucleo- purified modified nucleoside-containing RNAs were trans-
side modification induced variable, sequence-dependent fected into murine DCs. HPLC purification similarly
effects on IFN-a secretion (Figure 1C). These data ablated IFN-a secretion from DCs transfected with the
suggest that the presence of dsRNA and potentially clinically relevant -nucleoside modified mRNAs com-
other contaminants in in vitro-transcribed RNA could be plexed to TransIT used in Figure 3B. However,
responsible for innate immune activation. HPLC-purified RNA without nucleoside modification
Multiple HPLC bead matrix compositions and buffer remained potent inducers of TNF-a and IFN-a
systems were screened and alkylated non-porous (Figure 3C and D).

Downloaded from https://academic.oup.com/nar/article/39/21/e142/1104771 by guest on 29 December 2021

polystyrene-divinylbenzene copolymer matrix and The impact of HPLC purification of unmodified, - and
triethylammonium acetate buffer with an acetonitrile m5C/ -modified RNA on gene expression in human DCs
gradient was identified as a system capable of removing was analyzed using gene arrays. Total cellular RNA
dsRNA and other contaminants from in vitro-transcribed isolated from DCs from three different donors 6 h after
RNA. The HPLC chromatogram of -modified mRNA cells were transfected with TransIT-complexed RNA, were
encoding enhanced green fluorescent protein (eGFP) analyzed on an Illumina Human HT12v4 chip. RNA
demonstrated a major peak (Figure 2), which was col- modified with or m5C/ nucleosides induced less ex-
lected and identified as the expected RNA product using pression of type I interferons, interleukins, tumor necrosis
agarose gel electrophoresis. Additional UV-absorbing factor (TNF) family members, chemokines and markers
products with shorter and longer retention times relative associated with DC activation, while HPLC purification
of - and m5C/ -modified RNA further reduced induc-
to the main RNA product could also be observed.
tion of these genes to the levels observed in cells treated
Reanalysis of the purified RNA by HPLC demonstrated
only with TransIT (Figure 4A). The same sets of total
a single peak with the same retention time. RNAs with or
RNA from DCs that were tested on the gene arrays
without nucleoside modification encoding different se-
were also analyzed for levels of IFN-a, IFN-b and
quences yielded similar patterns with varying relative TNF-a mRNA by northern blot. Lower levels of IFN-a,
heights for the preceding and succeeding peaks. IFN-b and TNF-a mRNAs were detectable in DCs
HPLC purification of both unmodified and nucleoside- treated with nucleoside modified as compared to unmodi-
modified RNA reduced staining by dsRNA-specific mAb fied RNA. More importantly, none of these cytokine
to baseline levels (Figure 3A). Analysis of -modified mRNAs were detectable when DCs were transfected
RNA encoding clinically relevant proteins demonstrated with HPLC-purified RNAs containing or m5C/ modi-
that the amounts of dsRNA contamination in the in vitro fication. However, HPLC purified RNA without nucleo-
transcripts were dependent on the sequence, but HPLC side modification remained a potent inducer of IFN-a,
could successfully remove the contaminants from all of IFN-b and TNF-a mRNAs (Figure 4B).
them (Figure 3B). Next, the HPLC-purified RNAs were To determine whether HPLC purification affected
tested on human DCs. No TNF-a or type-I interferons translatability of in vitro transcripts, a series of mRNAs
(IFN-a and b) were induced following transfection of were tested following cell delivery. HPLC-purified Renilla
and mouse erythropoietin (mEPO) mRNAs were
translated at 2- to 20-fold higher levels compared to un-
purified RNA when delivered to 293 T cells by TransIT
(Figure 5A). In primary human DCs, the translational
enhancement was more robust, resulting in up to a
1000-fold increase when the same sets of unpurified and
HPLC-purified mRNAs were transfected with Lipofectin
(Figure 5B) or TransIT (Figure 5C). Similar increases in
translation were observed for other mRNAs after HPLC
purification, including mRNAs encoding firefly luciferase,
human EPO, macaque EPO and Metridia luciferase, and
other cell types, including mouse embryonic fibroblasts
and human primary keratinocytes. Translation levels
were much higher with - and m5C/ -modified eGFP
mRNA when the mRNA was HPLC-purified prior to
Figure 2. HPLC purification of RNA identifies contaminants eluting
before and after the expected product. Chromatogram of -modified transfection of human DCs (Figure 5D and data not
TEVeGFPAn mRNA. RNA was applied to the HPLC column and shown).
eluted using a linear gradient of Buffer B (0.1 M TEAA, pH 7.0, To characterize the contaminants being removed by
25% acetonitrile) in Buffer A (0.1 M TEAA, pH 7.0). The gradient HPLC purification, three fractions corresponding to RNAs
spanned 38–55% Buffer B over 22 min (red line). Absorbance at
260 nm was analyzed (black line), which demonstrated the expected eluting from the column prior to the major transcription
sized RNA as well as smaller and larger RNA species. Data shown product (fraction I), the full-length transcription product
are from one experiment that is representative of over 200. (fraction II), and RNAs eluting after the
PAGE 5 OF 10 Nucleic Acids Research, 2011, Vol. 39, No. 21 e142

Downloaded from https://academic.oup.com/nar/article/39/21/e142/1104771 by guest on 29 December 2021

Figure 3. HPLC purification of in vitro-transcribed nucleoside modified mRNA removes dsRNA contaminants and eliminates immunogenicity.
(A) 200 ng of RNA encoding the indicated protein and containing the indicated modified nucleosides with or without HPLC purification were
blotted and analyzed with the J2 dsRNA-specific mAb. (B) 200 ng of RNA encoding the indicated protein and containing -modifications with or
without HPLC purification were blotted and analyzed with the J2 dsRNA-specific mAb. Blots were reprobed with a 32P-labeled probe for the
30 -UTR of the RNAs to control for amount of RNA analyzed. (C) DCs were treated with TEVRenA51 RNA containing the indicated nucleoside
modifications with or without HPLC purification and complexed to Lipofectin. TNF-a levels were measured in the supernatants at 24 h. Differences
in the effect of nucleoside modification on immunogenicity of Renilla-encoding mRNA compared to Figure 1B is likely due to donor variation and
differences in UTRs of the RNAs. (D) DCs were treated with TEVLucA51 RNA containing the indicated nucleoside modifications with or without
HPLC purification and complexed to TransIT. IFN-a levels were measured in the supernatants at 24 h. Error bars are standard error of the mean.
Data shown is from one experiment that is representative of 3 or more.

expected transcription product (fraction III) were col- (Figure 7). HPLC purification of the RNA greatly
lected (Figure 6A). Nucleoside-modified RNA occasional- reduced toxicity in treated keratinocytes where unmodi-
ly demonstrated a second smaller peak overlying the large fied mRNA caused minimal cell rounding and death.
peak and isolation and purification of both peaks Daily treatment of keratinocytes for 10 days with
demonstrated similar RNA lengths on denaturing and HPLC-purified m5C/ -modified mRNA complexed to
non-denaturing agarose gel electrophoresis and RNAs TransIT showed no signs of cell toxicity and the rate of
with similar levels of translation and immunogenicity. proliferation was similar to that obtained with TransIT
RNA purified from the three fractions was analyzed for alone treated cells (Figure 7).
immunogenicity. RNA in fractions I and III induced
IFN-a secretion from transfected DCs, while the purified
full-length RNA (fraction II) was not immunogenic when DISCUSSION
it contained - or m5C/ -nucleoside modifications Modified nucleoside-containing mRNA has previously
(Figure 6B). Fraction I RNA had low levels of staining been used for the induction of iPS cells from fibroblasts
with the J2 mAb, while fraction III RNA had high levels with very high efficiency (3). The authors determined that
of staining similar to the unpurified RNA (Figure 6C). m5C/ -nucleoside modified mRNA yielded the least
Primary human keratinocytes and murine fibroblasts amount of RNA sensor activation and the highest level
treated once with unpurified, unmodified RNA delivered of translation, but needed to add the B18R protein, a
by TransIT complexing demonstrated detachment from vaccinia virus decoy receptor for type I interferon (25),
the collagen-coated plastic base as evidence of cell death. for optimal iPS cell generation. We previously reported
A second delivery of unmodified RNA 24 h later resulted that modified nucleoside-containing mRNA was efficient-
in the termination of the culture. Substantially less tox- ly delivered to primary dividing and non-dividing cells and
icity was observed when the RNA contained - or m5C/ produced high levels of encoded protein in an easily
-nucleoside modifications, but repeated daily delivery controlled manner (5). In addition, the RNA had
of m5C/ -nucleoside modified mRNA to keratino- reduced innate immune sensor activation. The residual
cytes reduced the final cell number by 75% on day 11 amount of activation of modified nucleoside-containing
e142 Nucleic Acids Research, 2011, Vol. 39, No. 21 PAGE 6 OF 10

Downloaded from https://academic.oup.com/nar/article/39/21/e142/1104771 by guest on 29 December 2021

Figure 4. HPLC purification of in vitro-transcribed nucleoside-modified mRNA eliminates activation of genes associated with RNA sensor activa-
tion. (A) Heatmap representing changes in expression of genes activated by RNA sensors were derived from microarray analyses of DCs treated for 6
hr with TransIT alone or transit-complexed TEVRenA51 RNA with the indicated modifications with or without HPLC purification. RNA from
medium-treated cells was used as the baseline for comparison. (B) Northern blot of RNA from DCs treated with medium or TransIT alone or
TransIT-complexed TEVRenA51 RNA with the indicated modifications with or without HPLC purification and probed for IFN-a, IFN-b, TNF-a
and GAPDH mRNAs.

mRNA depended on the sequence. In this report, we or TNF-a and no significant induction of genes associated
identify that m5C/ -nucleoside-modified RNA often has with RNA sensor activation. The purification procedure
the least ability to induce RNA sensor activation, but even can be easily scaled to produce large amounts of RNA
with these modifications, certain RNA sequences induce necessary for therapeutic applications and can be
high levels of cytokine production (Figure 1B and C). completed quickly and efficiently.
HPLC purification removes dsRNA and other contamin- The data suggests that different types of immunogenic
ants from in vitro-transcribed RNAs containing or contaminants are present in in vitro-transcribed mRNA. A
m5C/ nucleosides, yielding RNA with the highest series of RNAs that eluted before the major transcription
levels of translation, up to 1000-fold more than product, suggesting a smaller size, induced high levels of
non-HPLC purified RNA with no release of type I IFNs IFN-a, but had minimal staining with dsRNA-specific
PAGE 7 OF 10 Nucleic Acids Research, 2011, Vol. 39, No. 21 e142

Downloaded from https://academic.oup.com/nar/article/39/21/e142/1104771 by guest on 29 December 2021

Figure 5. HPLC purification of in vitro transcribed mRNA enhances translation. 293T (A) and human DCs (B and C) were transfected with
TransIT- (A and C) or Lipofectin- (B) complexed TEVRenA51 or TEVmEPOA51 mRNA with the indicated modifications with or without
HPLC purification and analyzed for Renilla luciferase activity or levels of supernatant-associated mEPO protein at 24 h. (D) Human DCs were
transfected with -modified TEVeGFPAn mRNA with or without HPLC purification (0.1 mg/well) complexed with Lipofectin or TransIT and
analyzed 24 h later. Error bars are standard error of the mean. Data shown is from one experiment that is representative of three or more.

mAbs. A second series of RNAs that eluted after the main to activate PKR (23), but whether m5C/ -nucleoside
transcription product were immunogenic when the RNA modifications in dsRNA block its ability to induce type
was unmodified or contained -modifications and had I IFNs through RNA sensors is unknown. The nature of
levels of dsRNA staining similar to the unpurified RNA the contaminants that elute prior to the main product is
(Figure 6). Long RNA (2 kb) with an added shorter length unknown. They have minimal staining with dsRNA
(276 nt) of complimentary RNA eluted at high concentra- specific mAb, but these mAbs require extended lengths
tions of acetonitrile, i.e. had a longer retention time, sug- of dsRNA for binding (20). We cannot rule out that
gesting that ds structures delay elution from the column shorter segments of dsRNA are present in the RNA that
matrix (K.K., H.M. and D.W., preliminary data). Three elutes prior to the main transcription product.
of the known mechanisms that produce contaminants HPLC purification of mRNAs enhanced their transla-
during in vitro transcription, self-complementary 30 exten- tion up to a 1000-fold in primary cells. The level of en-
sion (9), RNA-primed transcription from RNA templates hancement was greatest when the mRNA was unmodified,
(10) and RNA-dependent RNA polymerase activity (11) decreased when was incorporated and decreased fur-
can result in dsRNA of various lengths. If these contam- ther when m5C and were present (Figure 5A–C).
inants contain the main transcription product, they would These differences in translational enhancement of the
likely elute after the main transcription product. mRNAs are likely due to the RNA sensors PKR and
Interestingly, fraction III RNA from the purification of oligoadenylate synthetase (OAS) that directly decrease
m5C/ -modified RNA stained with dsRNA mAb, but translation with activation. We previously demonstrated
induced little IFN-a (Figure 6). dsRNA containing nu- that incorporation of m5C and into long mRNA
cleoside modifications, including , has reduced ability reduced its ability to activate PKR (6). Similarly, we
e142 Nucleic Acids Research, 2011, Vol. 39, No. 21 PAGE 8 OF 10

Downloaded from https://academic.oup.com/nar/article/39/21/e142/1104771 by guest on 29 December 2021

Figure 7. Daily transfection with HPLC-purified m5C/ -modified
mRNA does not reduce cell proliferation. Primary keratinocytes were
transfected daily with TransIT alone or m5C/ -modified
RNA-encoding Renilla luciferase with or without HPLC purification
complexed with TransIT. Every 2–3 days, cultures were split and equal
numbers of cells for each condition were plated. Total cell numbers for
each condition were divided by the total cell number in untreated cells
to calculate the percent of control proliferation.

dsRNA contaminants could involve RNA interference.

dsRNA greater than 27 bp in length is a substrate for
Dicer (27), whose action can lead to a specific suppression
of translation through the RNAi pathway (28). dsRNA
contaminants homologous to the desired mRNA would
lead to a specific suppression of translation in addition
to the non-specific suppression through RNA sensors of
the innate immune system.
The observation that the complexing agent used with
identical RNAs alters the type of cytokines released
from DCs could be caused by an effect of the complexing
agent on the interaction between the RNA and endosomal
Figure 6. Analysis of RNA contaminants removed by HPLC purifica-
tion. (A) One hundred microgram of -modified T7TSLucA30 RNA RNA sensors, the location of RNA after cytoplasmic
was applied to the HPLC column and 3 fractions were collected, all entrance, or the amount of time the RNA remains in the
RNAs eluting before the main transcription product (I), the expected endosome after endocytosis. RNA sensing TLRs require
RNA (II), and all RNAs eluting after the main transcription product acidification of endosomes to signal (29). A complexing
(III). The gradient began at 38% Buffer B and increased to 43% Buffer
B over 2.5 min and then spanned 43–65% Buffer B over 22 min.
agent that allows endocytosed RNA to exit the
Unmodified and m5C/ -modified T7TSLucA30 RNA had similar frac- endosome and enter the cytoplasm before acidification
tions obtained. (B) The RNAs from each fraction were complexed to and TLR signaling would result in reduced TNF-a secre-
TransIT and added to DCs and IFN-a in the supernatant was tion, as was observed with TransIT-complexed mRNA. In
measured 24 h later. Error bars are standard error of the mean.
(C) 200 ng of RNA from the 3 fractions and the starting unpurified
a comparison of cationic liposomes to linear
RNA were blotted and analyzed with the J2 dsRNA-specific mAb. polyethyleneimine (PEI) (a cationic polymer) delivery of
CpG containing DNA, it was found that PEI-DNA
induced less TNF-a, which was associated with a faster
exit from endosomes, while cationic lipid delivery resulted
recently demonstrated that RNA with -modifications in in DNA remaining in vesicular structures for extended
the absence of dsRNA contaminants induced less activa- periods of time and high levels of TNF-a (30). We simi-
tion of OAS compared to unmodified mRNA (26). It is larly observed that cationic liposome complexed RNA
also possible that RNA samples with nucleoside modifica- induced higher levels of TNF-a compared to the cationic
tions contain a smaller amount of short dsRNA contam- polymer with lipid TransIT. Another possibility for the
inants as the dsRNA-specific mAbs do not recognize discrepant cytokine response could be due to the sizes of
dsRNA shorter than 40 bp. An additional mechanism the complexed mRNA particles. A recent report by
for the inhibition of translation by mRNA containing Retting et al. (31) demonstrated that nanometric particles
PAGE 9 OF 10 Nucleic Acids Research, 2011, Vol. 39, No. 21 e142

induced IFN-a, while larger, micrometric particles proteins after delivery of chemically modified mRNA in mice.
induced TNF-a even when the particles were made with Nat. Biotechnol., 29, 154–157.
5. Kariko,K., Muramatsu,H., Welsh,F.A., Ludwig,J., Kato,H.,
the same RNA and complexing agent. Akira,S. and Weissman,D. (2008) Incorporation of pseudouridine
Current gene therapy methods in clinical trials use viral into mRNA yields superior nonimmunogenic vector with
vectors that can activate innate immune sensors (32) and increased translational capacity and biological stability.
integrate into host cell DNA (33) with potentially serious Mol. Ther., 16, 1833–1840.
complications. Although the use of mRNA for such 6. Anderson,B.R., Muramatsu,H., Nallagatla,S.R., Bevilacqua,P.C.,
Sansing,L.H., Weissman,D. and Kariko,K. (2010) Incorporation
therapies seems a safer alternative, activation of RNA of pseudouridine into mRNA enhances translation by diminishing
sensors that alter cellular activities and impair the PKR activation. Nucleic Acids Res., 38, 5884–5892.
intended function of the delivered mRNA (3) limit its use- 7. Kariko,K. and Weissman,D. (2007) Naturally occurring
fulness. With the sequencing of the human genome and nucleoside modifications suppress the immunostimulatory
the increasing ease and ability to sequence the genome of activity of RNA: implication for therapeutic RNA development.
Curr. Opin. Drug Discov. Devel., 10, 523–532.

Downloaded from https://academic.oup.com/nar/article/39/21/e142/1104771 by guest on 29 December 2021

individuals, new approaches are needed for the treatment 8. Milligan,J.F., Groebe,D.R., Witherell,G.W. and Uhlenbeck,O.C.
of inherited genetic disorders and therapies that require (1987) Oligoribonucleotide synthesis using T7 RNA
in vivo expression of extracellular or intracellular polymerase and synthetic DNA templates. Nucleic Acids Res., 15,
proteins. In addition, advances in regenerative medicine 8783–8798.
therapies where transient expression of transcription 9. Triana-Alonso,F.J., Dabrowski,M., Wadzack,J. and
Nierhaus,K.H. (1995) Self-coded 30 -extension of run-off
factors ex vivo transdifferentiate, reprogram and differen- transcripts produces aberrant products during in vitro
tiate cells are leading to new approaches for organ and transcription with T7 RNA polymerase. J. Biol. Chem., 270,
tissue restoration with the hope of treating many 6298–6307.
diseases including diabetes, spinal cord injury, neurologic 10. Nacheva,G.A. and Berzal-Herranz,A. (2003) Preventing
syndromes and organ repair. As we report here, purifica- nondesired RNA-primed RNA extension catalyzed by T7 RNA
polymerase. Eur. J. Biochem., 270, 1458–1465.
tion of modified nucleoside-containing mRNA avoids 11. Arnaud-Barbe,N., Cheynet-Sauvion,V., Oriol,G., Mandrand,B.
RNA sensor activation and increases the translational and Mallet,F. (1998) Transcription of RNA templates by T7
levels of the encoded protein. These mRNAs can be de- RNA polymerase. Nucleic Acids Res., 26, 3550–3554.
livered repeatedly without toxicity or the need to add in- 12. Thakur,C.S., Sama,J.N., Jackson,M.E., Chen,B. and Dayie,T.K.
hibitors of innate immune activation and can be used for (2010) Selective 13C labeling of nucleotides for large RNA NMR
spectroscopy using an E. coli strain disabled in the TCA cycle.
both ex vivo and in vivo therapies. J. Biomol. NMR, 48, 179–192.
13. Pallan,P.S. and Egli,M. (2007) Selenium modification of nucleic
acids: preparation of oligonucleotides with incorporated 20 -
ACKNOWLEDGEMENTS SeMe-uridine for crystallographic phasing of nucleic acid
structures. Nat. Protoc., 2, 647–651.
We thank Dr Louise Showe and the Wistar Genomics 14. Fotin-Mleczek,M., Duchardt,K.M., Lorenz,C., Pfeiffer,R., Ojkic-
Core Facility for performing the gene array analyses, Zrna,S., Probst,J. and Kallen,K.J. (2011) Messenger RNA-based
Houping Ni for technical assistance and the Skin vaccines with dual activity induce balanced TLR-7 dependent
Disease Research Core at Penn for primary keratinocytes. adaptive immune responses and provide antitumor activity.
J. Immunother., 34, 1–15.
15. Summer,H., Gramer,R. and Droge,P. (2009) Denaturing urea
polyacrylamide gel electrophoresis (Urea PAGE). J. Vis. Exp., 29,
FUNDING 1485.
Funding for open access charge: National Institutes of 16. Masuda,N., Ohnishi,T., Kawamoto,S., Monden,M. and Okubo,K.
(1999) Analysis of chemical modification of RNA from
Health (grant number HL87688 to K.K., AI050484, formalin-fixed samples and optimization of molecular biology
DE019059 to D.W.). applications for such samples. Nucleic Acids Res., 27, 4436–4443.
17. McKenna,S.A., Kim,I., Puglisi,E.V., Lindhout,D.A., Aitken,C.E.,
Conflict of interest statement. Drs Weissman and Kariko Marshall,R.A. and Puglisi,J.D. (2007) Purification and
are named on a patent for modified RNA submitted by characterization of transcribed RNAs using gel filtration
the University of Pennsylvania. chromatography. Nat. Protoc., 2, 3270–3277.
18. Kim,I., McKenna,S.A., Puglisi,E.V. and Puglisi,J.D. (2007)
Rapid purification of RNAs using fast performance liquid
chromatography (FPLC). RNA, 13, 289–294.
REFERENCES 19. Schonborn,J., Oberstrass,J., Breyel,E., Tittgen,J., Schumacher,J.
1. Angel,M. and Yanik,M.F. (2010) Innate immune suppression and Lukacs,N. (1991) Monoclonal antibodies to double-stranded
enables frequent transfection with RNA encoding reprogramming RNA as probes of RNA structure in crude nucleic acid extracts.
proteins. PLoS One, 5, e11756. Nucleic Acids Res., 19, 2993–3000.
2. Yakubov,E., Rechavi,G., Rozenblatt,S. and Givol,D. (2010) 20. Bonin,M., Oberstrass,J., Lukacs,N., Ewert,K., Oesterschulze,E.,
Reprogramming of human fibroblasts to pluripotent stem cells Kassing,R. and Nellen,W. (2000) Determination of preferential
using mRNA of four transcription factors. Biochem. Biophys. Res. binding sites for anti-dsRNA antibodies on double-stranded RNA
Commun., 394, 189–193. by scanning force microscopy. RNA, 6, 563–570.
3. Warren,L., Manos,P.D., Ahfeldt,T., Loh,Y.H., Li,H., Lau,F., 21. Kariko,K., Buckstein,M., Ni,H. and Weissman,D. (2005)
Ebina,W., Mandal,P.K., Smith,Z.D., Meissner,A. et al. (2010) Suppression of RNA recognition by Toll-like receptors: the
Highly efficient reprogramming to pluripotency and directed impact of nucleoside modification and the evolutionary origin of
differentiation of human cells with synthetic modified mRNA. RNA. Immunity, 23, 165–175.
Cell Stem Cell., 7, 618–630. 22. Hornung,V., Ellegast,J., Kim,S., Brzozka,K., Jung,A., Kato,H.,
4. Kormann,M.S., Hasenpusch,G., Aneja,M.K., Nica,G., Poeck,H., Akira,S., Conzelmann,K.K., Schlee,M. et al. (2006)
Flemmer,A.W., Herber-Jonat,S., Huppmann,M., Mays,L.E., 50 -Triphosphate RNA is the ligand for RIG-I. Science, 314,
Illenyi,M., Schams,A. et al. (2011) Expression of therapeutic 994–997.
e142 Nucleic Acids Research, 2011, Vol. 39, No. 21 PAGE 10 OF 10

23. Nallagatla,S.R. and Bevilacqua,P.C. (2008) Nucleoside 29. Lund,J.M., Alexopoulou,L., Sato,A., Karow,M., Adams,N.C.,
modifications modulate activation of the protein kinase PKR in Gale,N.W., Iwasaki,A. and Flavell,R.A. (2004) Recognition of
an RNA structure-specific manner. RNA, 14, 1201–1213. single-stranded RNA viruses by Toll-like receptor 7. Proc. Natl
24. Kramer,M., Schulte,B.M., Eleveld-Trancikova,D., van Hout- Acad. Sci. USA, 101, 5598–5603.
Kuijer,M., Toonen,L.W., Tel,J., de Vries,I.J., van Kuppeveld,F.J., 30. Saito,Y., Higuchi,Y., Kawakami,S., Yamashita,F. and
Jansen,B.J. and Adema,G.J. (2010) Cross-talk between human Hashida,M. (2009) Immunostimulatory characteristics induced by
dendritic cell subsets influences expression of RNA sensors and linear polyethyleneimine-plasmid DNA complexes in cultured
inhibits picornavirus infection. J. Innate Immun., 2, 360–370. macrophages. Hum. Gene Ther., 20, 137–145.
25. Symons,J.A., Alcami,A. and Smith,G.L. (1995) Vaccinia virus 31. Rettig,L., Haen,S.P., Bittermann,A.G., von Boehmer,L.,
encodes a soluble type I interferon receptor of novel structure Curioni,A., Kramer,S.D., Knuth,A. and Pascolo,S. (2010) Particle
and broad species specificity. Cell, 81, 551–560. size and activation threshold: a new dimension of danger
26. Anderson,B.R., Muramatsu,H., Jha,B.K., Silverman,R.H., signaling. Blood, 115, 4533–4541.
Weissman,D. and Kariko,K. (2011) Nucleoside modifications in 32. Raper,S.E., Chirmule,N., Lee,F.S., Wivel,N.A., Bagg,A.,
RNA limit activation of 20 -50 oligoadenylate synthetase and Gao,G.P., Wilson,J.M. and Batshaw,M.L. (2003) Fatal systemic
increase resistance to cleavage by RNase L. Nucleic Acids Res. inflammatory response syndrome in a ornithine transcarbamylase

Downloaded from https://academic.oup.com/nar/article/39/21/e142/1104771 by guest on 29 December 2021

(in press) PMID: 21813458. deficient patient following adenoviral gene transfer. Mol. Genet.
27. Kim,D.H., Behlke,M.A., Rose,S.D., Chang,M.S., Choi,S. and Metab., 80, 148–158.
Rossi,J.J. (2005) Synthetic dsRNA Dicer substrates enhance 33. Hacein-Bey-Abina,S., Hauer,J., Lim,A., Picard,C., Wang,G.P.,
RNAi potency and efficacy. Nat. Biotechnol., 23, 222–226. Berry,C.C., Martinache,C., Rieux-Laucat,F., Latour,S.,
28. Fire,A., Xu,S., Montgomery,M.K., Kostas,S.A., Driver,S.E. and Belohradsky,B.H. et al. (2003) Efficacy of gene therapy for
Mello,C.C. (1998) Potent and specific genetic interference by X-linked severe combined immunodeficiency. N. Engl. J. Med.,
double-stranded RNA in Caenorhabditis elegans. Nature, 391, 363, 355–364.
806–811.