Báo cáo khoa học: Gene expression silencing with ‘specific’ small interfering RNA goes beyond specificity – a study of key parameters to take into account in the onset of small interfering RNA off-target effects potx

Indeed, in some circum-stances, and especially in immune cells, siRNAs are Keywords cell type; gene expression; off-target effects; silencing; siRNA Correspondence T.. Results Verificati

RNA goes beyond specificity – a study of key parameters

to take into account in the onset of small interfering RNA off-target effects

Se´bastien Vankoningsloo1, Franc¸oise de Longueville2, Ste´phanie Evrard2, Pierre Rahier2, Andre´e Houbion1, Antoine Fattaccioli1, Me´lanie Gastellier1, Jose´ Remacle2, Martine Raes1, Patricia Renard1 and Thierry Arnould1

1 Laboratoire de Biochimie et Biologie Cellulaire, University of Namur (F.U.N.D.P), Belgium

2 Eppendorf Array Technologies, Namur, Belgium

RNA interference (RNAi) is a recently discovered

gene-silencing pathway [1] triggered by

dsRNA-derived molecules such as small interfering RNAs

(siRNAs) or microRNAs, leading to the degradation

of a particular mRNA (slicing) or to repression of

translation [2,3] Thereafter, the use of chemically

synthesized siRNAs as a new loss-of-function strategy

exploded during the last decade, mainly because the

RNAi pathway is believed to apply to all genes in

several species Therefore, siRNAs became useful

tools for the silencing of genes playing a role in many

biological processes

The extensive use of siRNAs, designed to match perfectly with a particular mRNA target, is based on the assumption of their high specificity Indeed, it was initially suggested that only one mismatch could abol-ish the siRNA-induced slicing activity [4] However, several studies using DNA microarray and⁄ or compu-tational approaches have shown that siRNAs can generate side effects, by inducing the degradation of nontarget mRNAs sharing sequence homology with the siRNA seed region, or by repressing the translation

of unintended proteins [5–10] Indeed, in some circum-stances, and especially in immune cells, siRNAs are

Keywords

cell type; gene expression; off-target

effects; silencing; siRNA

Correspondence

T Arnould, Laboratoire de Biochimie et

Biologie Cellulaire, University of Namur

(F.U.N.D.P), 61 rue de Bruxelles, 5000

Namur, Belgium

Fax: +32 81 724125

Tel: +32 81 724129

E-mail: thierry.arnould@fundp.ac.be

(Received 10 January 2008, revised 12

March 2008, accepted 19 March 2008)

doi:10.1111/j.1742-4658.2008.06415.x

RNA-mediated gene silencing (RNA interference) is a powerful way to knock down gene expression and has revolutionized the fields of cellular and molecular biology Indeed, the transfection of cultured cells with small interfering RNAs (siRNAs) is currently considered to be the best and easi-est approach to loss-of-function experiments However, several recent stud-ies underscore the off-target and potential cytotoxic effects of siRNAs, which can lead to the silencing of unintended mRNAs In this study, we used a low-density microarray to assess gene expression modifications in response to five different siRNAs in various cell types and transfection con-ditions We found major differences in off-target signature according to: (a) siRNA sequence; (b) cell type; (c) duration of transfection; and (d) post-transfection time before analysis These results contribute to a better understanding of important parameters that could impact on siRNA side effects in knockdown experiments

Abbreviations

DF, DharmaFECT1; IFN, interferon; IRF, interferon responsive factor; LAMP2, lysosome-associated membrane protein 2; NT, nontargeting; RISC, RNA-induced silencing complex; RNAi, RNA interference; siRNA, small interfering RNA; SREBF1, sterol-responsive element-binding protein 1; TLR3, Toll-like receptor 3.

also able to trigger an ‘interferon (IFN) response’

through the activation of cytosolic proteins such as

dsRNA-dependent protein kinase and⁄ or membrane

receptors such as Toll-like receptor 3 (TLR3), leading

to a general repression of translation [11] An

inflam-matory response was also observed in primary human

chondrocytes transfected with siRNAs [12] These

con-siderations cast some doubts on the validity of several

results previously published – particularly on the strict

specificity for targets supposed to be responsible for a

biological response – and highlight the importance of

studies intended to increase our understanding of the

extent of siRNA nonspecific effects and the conditions

under which they occur

In this work, we used a low-density DNA

micro-array that allows gene expression analysis of 273 genes,

in order to determine the off-target effects generated

by five siRNAs in different cell types and experimental

conditions We first studied the side effects of two

different siRNAs targeting the sterol-responsive

element-binding protein 1 (SREBF1) mRNA encoding

a transcription factor, two different siRNAs targeting

the lysosome-associated membrane protein 2 (LAMP2)

mRNA encoding a lysosomal glycoprotein, and a

non-targeting (NT) siRNA Gene expression profiles were

determined for each siRNA in transiently transfected

human osteosarcoma 143B cells, lung

adenocarci-noma A549 cells, and lung IMR-90 fibroblasts

Furthermore, in 143B cells, we studied the effects of

different transfection and post-transfection periods on

the modifications in gene expression triggered by

siRNA

Results

Verification of siRNA efficiency

We used two siRNAs designed to specifically knock

down expression of the transcription factor SREBF1

(SREBF1⁄ siRNA1 and SREBF1 ⁄ siRNA2), and two

siRNAs targeting the transcript coding for the

lyso-somal glycoprotein LAMP2, which is not known to be

directly involved in transcription events (LAMP2⁄

siR-NA1 and LAMP2⁄ siRNA2) The particular targets

were chosen on the basis of their interest for other

research programmes in our laboratory The efficiency

of these siRNAs at concentrations ranging from 5 nm

to 100 nm was first demonstrated by real-time PCR in

143B, A549 and IMR-90 cells (Fig 1) The choice of

these cell types was based on the selection of

trans-formed or nontranstrans-formed cells expressing or not

expressing the siRNA-responsive TLR3 receptor

Indeed, 143B and A549 are tumor-derived cell lines,

the latter being reported to express TLR3 [13], whereas IMR-90 is a nonimmortalized cell type We observed that the transfection reagent DharmaFECT1 (DF) has

no or little effect on the abundance of SREBF1 (Fig 1A,B) or LAMP2 (Fig 1C) transcript in these cell types The SREBF1-specific siRNAs (100 nm) were both very efficient at decreasing SREBF1 transcript abundance, with reductions of 81%, 66% and 69% for SREBF1⁄ siRNA1 and reductions of 79%, 78% and 71% for SREBF1⁄ siRNA2 in 143B, A549 and

IMR-90 cells, respectively (Fig 1A) Under these conditions, the effect of the SREBF1-targeting siRNA was pro-longed, at least, up to 72 h post-transfection, as dem-onstrated in 143B (Fig 1B) Both LAMP2-specific siRNAs (100 nm) were also efficient, as the abundance

of the corresponding transcript was decreased by 89%, 78% and 86% for LAMP2⁄ siRNA1 and by 64%, 58% and 66% for LAMP2⁄ siRNA2 in 143B, A549 and IMR-90 cells, respectively (Fig 1C) In contrast, an siRNA with an NT sequence did not dramatically alter the abundance of SREBF1 or LAMP2 mRNAs in these conditions The main observed effect was even a slight increase in the abundance of SREBF1 transcript

in each cell type

The efficiency of SREBF1⁄ siRNA1 was also investi-gated at the protein level by western blotting analysis

of SREBF1 abundance in 143B cells (Fig 2) We found a concentration-dependent and time-sustained decrease in SREBF1 protein abundance in 143B cells The signals present at 48 h and 72 h after cell transfec-tion with SREBF1⁄ siRNA1 (100 nm), apparently not correlated with mRNA levels (Fig 1B), probably result from differences in exposure times during western blot-ting We also observed a slight increase in SREBF1 protein level triggered in the presence of the NT siRNA, in agreement with the slight increase in SREBF1 mRNA observed under the same conditions (Fig 1A,B)

Off-target signatures elicited by five siRNAs in three different cell types

We next studied the effects of DF and of the five siRNAs at 100 nm on gene expression The side effects

of these siRNAs were systematically investigated in 143B (Fig 3), A549 (Fig 4) and IMR-90 cells (Fig 5) transiently transfected for 24 h before total RNA extraction and microarray analysis Please note that the scales are different for each heat map Each experi-ment was performed on biological triplicates, and the complete lists of relative transcript level values and corresponding standard deviations are provided in sup-plementary Tables S1–S12 Several transcripts were

0.5

1.0

1.5

2.0

2.5

Relative SREBF1 mRNA abundance 0 24 48 72 0 24 48 72 0 24 48 72

(100 nM)

NT siRNA (100 nM)

Time post-transfection (h)

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

Relative SREBF1 mRNA abundance DF 100 20 5

SREBF1 siRNA1 (n M )

100 20 5 SREBF1 siRNA2 (n M )

100 20 5

NT siRNA (n M )

DF 100 20 5 SREBF1 siRNA1 (n M )

100 20 5 SREBF1 siRNA2 (n M )

100 20 5

NT siRNA (n M )

DF 100 20 5 SREBF1 siRNA1 (n M )

100 20 5 SREBF1 siRNA2 (n M )

100 20 5

NT siRNA (n M )

DF 100 20 5 LAMP2 siRNA1 (n M )

100 20 5 LAMP2 siRNA2 (n M )

100 20 5

NT siRNA (n M )

DF 100 20 5 LAMP2 siRNA1 (n M )

100 20 5 LAMP2 siRNA2 (n M )

100 20 5

NT siRNA (n M )

DF 100 20 5 LAMP2 siRNA1 (n M )

100 20 5 LAMP2 siRNA2 (n M )

100 20 5

NT siRNA (n M )

A

B

C

Fig 1 Effect of the SREBF1-targeting and the LAMP2-targeting siRNAs on SREBF1 and LAMP2 mRNA levels analyzed by real-time PCR in 143B, A549 and IMR-90 cells (A) 143B, A549 and IMR-90 cells were incubated for 24 h with DF or transfected for 24 h with SREBF1 ⁄ siR-NA1, SREBF1 ⁄ siRNA2 or the NT siRNA at the indicated concentrations before RNA extraction, reverse transcription, and amplification in the presence of SYBR Green and specific primers (B) 143B cells were incubated for 24 h with DF or transfected for 24 h with SREBF1 ⁄ siRNA1

or the NT siRNA at 100 nM RNA was extracted 0, 24, 48 and 72 h post-transfection and processed for real-time PCR analysis (C) 143B, A549 and IMR-90 cells were incubated for 24 h with DF or transfected for 24 h with LAMP2 ⁄ siRNA1, LAMP2 ⁄ siRNA2 or the NT siRNA at the indicated concentrations before RNA extraction and processing for real-time PCR analysis TBP was used as a housekeeping gene for data normalization Results are expressed as relative SREBF1 or LAMP2 transcript abundance in treated cells as compared to untreated con-trol cells (n = 1).

not detected, most probably because of their absence

or low abundance: depending on the experiment, the

total number of mRNAs detected ranged between 185

and 260 out of 273 The results discussed here below

are only related to genes for which mRNA relative

abundance in siRNA-transfected cells was found to be

significantly different when compared with the mRNA

abundance determined in DF-treated cells

First, we observed that treatment with DF alone

affected the expression of a few genes, especially in

IMR-90 cells, such as IGFBP3 (insulin-like growth

fac-tor-binding protein 3) (3.3-fold decrease), ICAM1

(intercellular adhesion molecule 1) (1.9-fold decrease)

and PCNA (proliferating cell nuclear antigen) (1.8-fold

decrease) (supplementary Table S9) Second, we

estab-lished gene expression profiles for the five siRNAs in

the three cell types The number of genes differentially

expressed in response to siRNAs with statistical

signifi-cance ranged between one and 12, according to the

condition The main conclusion drawn from these

experiments is that each siRNA is associated with a

unique molecular signature on gene expression For

example, transcripts that are downregulated by

LAMP2⁄ siRNA2 in A549 cells, such as JUN (jun

oncogene), PLAU (plasminogen activator, urokinase),

PLAUR (plasminogen activator, urokinase receptor),

RRM1 (ribonucleotide reductase M1 polypeptide),

TERF1 (telomeric repeat binding factor 1) and TGFBR2 (transforming growth factor, beta recep-tor II) (Fig 4), were not systematically downregulated

by either LAMP2⁄ siRNA1, SREBF1⁄ siRNA1, SREBF1⁄ siRNA2 or the NT siRNA Importantly, the fact that two different siRNAs targeting the same tran-script do not provide the same gene expression profiles (see Venn diagrams in Figs 3–5) rules out potential secondary effects due to target knockdown, and indi-cates that the unintended mRNA downregulations observed are most probably siRNA off-target effects

To some extent, the signatures of siRNAs also seem

to be dependent on the cell type in which siRNAs are introduced Indeed, whereas several mRNAs were con-sistently downregulated by a given siRNA in every cell type, we found that the abundance of some transcripts was clearly differently affected by siRNA according to the cell type, as illustrated by the 2.3-fold downregula-tion of SOD2 (superoxide dismutase 2) found exclu-sively in IMR-90 cells transfected with SREBF1⁄ siRNA2 A global analysis of all data cross-ing siRNAs and cell types revealed that about 60% of the siRNA off-target effects observed in this study appear to be cell type-specific

Finally, in order to validate these data with another method, we performed real-time PCR analyses for some selected transcripts (CTGF, JUN, PLAU, SPARC, TGFBR2) on samples used for microarray experiments (RNAs extracted directly after a 24 h transfection of 143B or A549 cells with SREBF1⁄ siR-NA2 or LAMP2⁄ siRNA2) (supplementary Table S13)

We observed that mRNA abundances were modified similarly with both methods, attesting to the reliability

of the results

Kinetics of off-target effects induced by siRNA

In order to determine the time-course of siRNA side effects in 143B cells transfected for 24 h with SREBF1⁄ siRNA1 or the NT siRNA (100 nm), gene expression data obtained at 0, 24 and 48 h post-trans-fection were compared in experiments performed on biological triplicates (Fig 6) Again, we observed that the transfection reagent alone induced only small vari-ations in the abundance of gene transcripts, no matter what the post-transfection time was (Fig 6, col-umns 1–3) In contrast, the relative abundance of sev-eral mRNAs (between two and 15) was significantly modified in response to the introduction of SREBF1⁄ siRNA1 (Fig 6, columns 4–6) or the NT siRNA (Fig 6, columns 7–9) into 143B cells In these conditions, the highest number of modifications was observed 24 h post-transfection (Fig 6, columns 5 and

CTL DF 100 50 20 5 100 50 20 5

SREBF1 siRNA1 (n M )

NT siRNA (n M )

SREBF1

α-tubulin

SREBF1

α-tubulin

SREBF1

α-tubulin

SREBF1

α-tubulin

0 h

post-transfection

24 h

post-transfection

48 h

post-transfection

72 h

post-transfection

Fig 2 Effect of the SREBF1-targeting siRNA on SREBF1 protein

level analyzed by western blotting in 143B cells 143B cells were

incubated for 24 h with DF or transfected for 24 h with

SREBF1 ⁄ siRNA1 or the NT siRNA at the indicated concentrations.

Clear cell lysates were prepared 0, 24, 48 or 72 h post-transfection.

SREBF1 abundance was determined by western blotting on 25 lg

of protein, and immunodetection of a-tubulin was used as a loading

control.

8) (see also supplementary Tables S1 and S14)

Repre-sentative results are presented in Fig 7, which

summa-rizes and illustrates each kind of kinetic profile that we

obtained As shown in Fig 7A a moderate but

sus-tained upregulation of CDKN1B (cyclin-dependent

kinase inhibitor 1B, also known as p27Kip1) was

observed after the transfection of 143B cells with

SREBF1⁄ siRNA1 In Fig 7B, we illustrate the

upregu-lation of PLAU (plasminogen activator urokinase) in

cells responding to either SREBF1-specific or the NT

siRNA A similar profile was also obtained for

SPARC (secreted protein acidic cysteine-rich, also

known as osteonectin) The abundance of several

tran-scripts was also decreased in cells transfected with

SREBF1⁄ siRNA1, such as CCND2 (cyclin D2)

(Fig 7C), UNG (uracil-DNA glycosylase), ALDOA

(aldolase A), CENPF (centromere protein F), CKB

(brain creatine kinase) or CTGF (connective tissue

growth factor) The NT siRNA also downregulated

the expression of several genes, such as EGFR

(epider-mal growth factor receptor) (Fig 7D), MAP2K1 (also

known as MEK1, mitogen-activated protein kinase

kinase 1) and RAF1 (murine leukemia viral oncogene homolog 1) Finally, downregulation of IGFBP3 was observed in cells transfected with either SREBF1⁄ siRNA1 or the NT siRNA (Fig 7E)

Effect of duration of transfection period on siRNA off-target signature

To assess the putative effect of the transfection period

on the siRNA nonspecific effects, gene expression pro-files in 143B cells transfected for 24 or 48 h with SREBF1⁄ siRNA1 or the NT siRNA at 100 nm were next determined in three independent experiments RNA extractions were performed between 0 and 48 h post-transfection (see also supplementary Tables S14 and S15) As shown in Fig 8, the number of genes differentially expressed was higher after a 48 h than after a 24 h transfection period The heat map (Fig 9) compares, in all tested conditions, the relative abundances of mRNAs differentially expressed in at least one condition In the presence of SREBF1⁄ siRNA1, we usually observed higher upregulation or

SPARC

LAMP2-siRNA1 LAMP2-siRNA2

PLAU CCND3 CANX CAV1 MAP2K1 IGFBP3 RAF1 TNFRSF10B UNG YWHAZ CCND1 PLAUR EGFR DUSP1 CTGF JUN

SPARC

CCND3 CCND1

CANX

CAV1

MAP2K1 IGFBP3

RAF1 TNFRSF10B

UNG

YWHAZ

PLAU

PLAUR EGFR DUSP1

CTGF JUN

Fig 3 Effect of the SREBF1-targeting and the LAMP2-targeting siRNAs on gene expression profiles analysed by microarray in 143B cells Cells were incubated for 24 h with DF or transfected for 24 h with SREBF1 ⁄ siRNA1, SREBF1 ⁄ siRNA2 (A), LAMP2 ⁄ siRNA1, LAMP2 ⁄ siRNA2 (B) or the NT siRNA at 100 nM before RNA extraction, reverse transcription, and processing for microarray analysis Expression plots present the genes displaying significant differences in relative transcript level between siRNA-transfected cells and DF-treated cells (n = 3) Color key: green, downregulation; red, upregulation A scale for heat maps as minimum and maximum fold differences is presented The Venn diagrams present the numbers of mRNAs differentially expressed with statistical significance in the presence of the indicated siRNAs in 143B cells The numbers of transcripts differentially expressed in the presence of both siRNAs specific for the same target are indicated in diagram intersections.

downregulation magnitudes after a 48 h transfection

(Fig 9, columns 7 and 8) than after a 24 h transfection

(Fig 9, columns 5 and 6) Similar conclusions can be

drawn from data obtained for cells transfected with

the NT siRNA (Fig 9, columns 11 and 12 versus 9

and 10) Therefore, it seems that the longer the

trans-fection period, the stronger the off-target effects of

siRNA on gene expression

mRNA homology with siRNA seed region

Perfect mRNA⁄ siRNA pairing is not necessary for

siRNA off-target effects Indeed, homology between

mRNA and siRNA seed region (encompassing

nucleo-tides 2–8 or 2–7 of the antisense strand) was shown to

be sufficient to induce off-target silencing [6,7,10]

Hence, we searched for regions of sequence homology

between the guide strands of the five siRNAs used in

this study and their respective unspecific targets

(Fig 10) The transcripts used for this analysis were

found to be significantly downregulated in 143B, A549

and IMR-90 cells transfected for 24 h with the siRNAs

(Figs 3–5 and supplementary Tables S1–S12) For sev-eral mRNAs, we found small stretches of sequence identity with the 3¢-end of siRNA sense sequences (5¢-end of antisense sequences) However, only 65% of them (21 of 33) can lead to perfect mRNA pairing with siRNA seed regions, as defined above Therefore,

an important proportion (about 35%) of the siRNA side effects observed here cannot be directly explained

by seed homology This analysis was also repeated with the siRNA passenger strands, but no perfect seed match was found in these conditions (data not shown)

Discussion

It is now well established that off-target silencing is a fundamental feature of siRNAs [5,6,9,14] The present investigation was conducted in order to increase our knowledge about siRNA off-target effects under vari-ous experimental conditions Molecular signatures of siRNAs were determined with a commercial low-density microarray designed for siRNA side effect studies This microarray comprises 273 capture probes

0 0 12 SREBF1 siRNA1

SREBF1 siRNA2

8 0 10 LAMP2 siRNA1

LAMP2 siRNA2

RAF1

GPX1 TERF1 MAP2K1 IGFBP3 CTNNB1 TGFBR2 YWHAZ CCND1 RRM1 PLAUR IL8 JUND CDKN1A PLAU BIN1 MYC JUN CSF1 GADD45A

EGFR

RAF1

GPX1

MAP2K1

TGFBR2 TERF1 RRM1 CTNNB1 IGFBP3

YWHAZ

CCND1 PLAUR

IL8 JUND

CDKN1A

PLAU BIN1

MYC JUN CSF1

GADD45A

EGFR

Fig 4 Effect of the SREBF1-targeting and the LAMP2-targeting siRNAs on gene expression profiles analyzed by microarray in A549 cells Cells were incubated for 24 h with DF or transfected for 24 h with SREBF1⁄ siRNA1, SREBF1 ⁄ siRNA2 (A), LAMP2 ⁄ siRNA1, LAMP2 ⁄ siRNA2 (B) or the NT siRNA at 100 nM before RNA extraction, reverse transcription, and processing for microarray analysis Expression plots present the genes displaying significant differences in relative transcript level between siRNA-transfected cells and DF-treated cells (n = 3) Color key: green, downregulation, red, upregulation A scale for heat maps as minimum and maximum fold differences is presented The Venn diagrams present the numbers of mRNAs differentially expressed with statistical significance in the presence of the indicated siRNAs in A549 cells The numbers of transcripts differentially expressed in the presence of both siRNAs specific for the same target are indicated in diagram intersections.

allowing the expression analysis, at the transcriptomic

level, of genes mainly involved in cell responses to

IFN challenge, apoptosis, DNA repair, cell cycle, and

metabolism

The few effects of DF on gene expression were

found to be dependent on cell type Indeed, whereas

variations observed in both 143B and A549 cells

incu-bated with DF alone were generally negligible, they

were more numerous in IMR-90 cells, as illustrated by

the slight but reproducible downregulation of ADPRT

(ADP-ribosyltransferase), CCNB1 (cyclin B1), DDIT3

(DNA-damage-inducible transcript 3), ICAM1,

IG-FBP3, PCNA, PRKDC (protein kinase,

DNA-acti-vated, catalytic polypeptide), SERPINE1⁄ PAI-1

(serpin peptidase inhibitor 1⁄ plasminogen activator

inhibitor-1), TFDP1 (transcription factor Dp-1),

TNFRSF10B (tumor necrosis factor receptor

super-family, member 10b) and TYMS (thymidylate

synthe-tase) This transfection reagent might therefore alter

some cellular processes in a cell type-dependent

man-ner For instance, an increase in the cell cycle timing

could be expected following the downregulation of

PCNA, coding for a protein involved in the control of

DNA replication and CDK2-cyclin A activity [15]

In most cases (about 70%), and as expected, DF-induced effects on gene expression were also observed in the presence of any tested siRNA, as illus-trated by the comparable downregulation of ADPRT

in IMR-90 cells in the presence of DF alone (0.68 ± 0.25) or in combination with LAMP2⁄ siRNA2 (0.64 ± 0.17) or the NT siRNA (0.64 ± 0.12) (supplementary Table S12; see also supplementary Tables S9–S11) However, additional or antagonistic effects of DF and siRNAs were also observed For example, the SERPINE1 mRNA level was reduced by

DF alone (0.65 ± 0.08) but was increased with statisti-cal significance by LAMP2⁄ siRNA1 (2.24 ± 0.82) in IMR-90 cells (supplementary Table S11)

The four targeting siRNAs used in this study pro-vide efficient knockdown of their respective targets at

100 nm This concentration might seem rather high, but was chosen in order to generate side effects allow-ing a comparative study of the importance of siRNA sequence, cell type, transfection period and post-trans-fection time before analysis The differences in siRNA on-target efficiencies observed between 143B, A549 and IMR-90 cells (Fig 1), as previously found for other cell lines [16], could probably be explained by

SREBF1 siRNA1

SREBF1 siRNA2

9 0 1 LAMP2 siRNA1

LAMP2 siRNA2

DF DF LAMP2-siRNA1 LAMP2-siRNA2 NT_siRNA

MYBL2 BAD UNG MADH3 WARS SERPINE1 ICAM1 IGFBP3 HIST1H3I TGFBR2 HSPCA MAPK1 SOD2 EGFR

JUND

MYBL2

BAD UNG

MADH3 WARS SERPINE1

ICAM1

IGFBP3 HIST1H3I

TGFBR2

HSPCA MAPK1

SOD2 EGFR

JUND

Fig 5 Effect of the SREBF1-argeting and the LAMP2-targeting siRNAs on gene expression profiles analyzed by microarray in IMR-90 cells Cells were incubated for 24 h DF or transfected for 24 h with SREBF1 ⁄ siRNA1, SREBF1 ⁄ siRNA2 (A), LAMP2 ⁄ siRNA1, LAMP2 ⁄ siRNA2 (B)

or the NT siRNA at 100 nM before RNA extraction, reverse transcription, and processing for microarray analysis Expression plots present the genes displaying significant differences in relative transcript level between siRNA-transfected cells and DF-treated cells (n = 3) Color key: green, downregulation; red, upregulation A scale for heat maps as minimum and maximum fold differences is presented The Venn diagrams present the numbers of mRNAs differentially expressed with statistical significance in the presence of the indicated siRNAs in IMR-90 cells The numbers of transcripts differentially expressed in the presence of both siRNAs specific for the same target are indicated

in diagram intersections.

the expression level of the RNAi pathway components

in each cell type and⁄ or by different transfection

effi-ciencies These hypotheses highlight the importance of

the cellular environment in the determination of both

efficiency and specificity of siRNA molecules, not only

for in vitro studies, but also when siRNA-based

thera-peutic approaches are considered Moreover, it was

suggested that the cellular background could modify

the degree of siRNA off-target effects elicited through

1 2 3 4 5 6 7 8 9

PLAU

SPARC CTGF CDKN1B HSPB1 MLH1 IGFBP2 BCL2L1 HSPCB BIN1 K-ALPHA-1 ADPRT ALDOA CKB CENPF HPRT1 UNG CCND2 CANX CASP3 RAF1 UBE2V1 MAP2K1 EGFR PLAUR IGFBP3

A

B

Fig 6 Kinetics of the gene expression profiles induced by

SREBF1 ⁄ siRNA1 in 143B cells (A) Design of the experiment The

24 h transfection period is indicated on a gray background (B)

143B cells were incubated for 24 h with DF or transfected for 24 h

with SREBF1⁄ siRNA1 or the NT siRNA at 100 nM RNA was

extracted 0, 24 or 48 h post-transfection, reverse transcribed, and

processed for microarray analysis Expression plots present the

genes displaying significant differences in relative transcript level

between siRNA-transfected cells and DF-treated cells (n = 3) Color

key: green, downregulation; red, upregulation A scale for heat

maps as minimum and maximum fold differences is presented.

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 A

B

C

D

E

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5

0.0 0.2 0.4 0.6 0.8 1.0 1.2

0.0 0.2 0.4 0.6 0.8 1.0 1.2

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4

Fig 7 Representative kinetic profiles of gene expression in 143B cells incubated with DF (circles) or transfected with SREBF1 ⁄ siR-NA1 (squares) or the NT siRNA (triangles) Gene expression was analyzed by microarray 0, 24 and 48 h post-transfection, and pro-files are illustrated for CDKN1B (A), PLAU (B), CCND2 (C), EGFR (D) and IGFBP3 (E).

an IFN response pathway, as the IFN response was

found to be stronger in TLR3-expressing cells [11] and

in nontumor cells [17] However, genes classically

asso-ciated with the siRNA-induced IFN response, such as

IFITM2(interferon-induced transmembrane protein 2),

IFNAR1 (interferon receptor 1) or IRF1

(interferon-responsive factor 1), were not upregulated in the

presence of siRNAs, even in the TLR3-expressing

A549 cells or in the nonimmortalized IMR-90 cells

We showed that two different siRNAs designed to

knock down SREBF1 can also modify the expression

of unintended genes in 143B, A549 and IMR-90 cells

Interestingly, the sets of misregulated genes are not the

same for each siRNA This lack of overlapping effects

rules out an indirect effect resulting from the silencing

of the transcription factor SREBF1, which would

modify gene expression in an siRNA-independent

manner Therefore, these variations in mRNA

abun-dance can be considered as real siRNA off-target effects Similar conclusions can be drawn from experi-ments performed with two other siRNAs targeting the LAMP2 transcript Furthermore, we observed that an

NT siRNA, used as a negative control in our experi-ments, unexpectedly altered the expression of several genes affected or not affected by the siRNAs targeting SREBF1 or LAMP2 Thus, the unique nonspecific molecular signature generated by each siRNA supports previous studies showing that off-target effects are dependent on siRNA sequence [6,7] The role of sequence pairing in siRNA side effects is also supported by data showing that these effects can be dramatically reduced in the presence of another con-trol, the RNA-induced silencing complex (RISC)-free siRNA (data not shown) Unlike the NT siRNA, this negative control is not loaded onto RISC, is unable to interact with mRNA, and thus cannot direct slicing It

is also important to note that the unexpected effects of the NT siRNA on gene expression underline the diffi-culty of choosing the most relevant control in RNAi experiments in order to obtain reliable results, as emphasized recently [18]

The seed region is particularly important in siRNA side effects, because mRNA⁄ siRNA pairing in this short region may be sufficient to induce mRNA deg-radation [6,19] Thus, we investigated whether siRNA seed regions share homology with the sequences of mRNAs downregulated directly after cell transfection with SREBF1⁄ siRNA1, SREBF1 ⁄ siRNA2, LAMP2 ⁄ siRNA1, LAMP2⁄ siRNA2 or the NT siRNA We determined that about 35% of these downregulated mRNAs do not show perfect sequence matching with the seed region of the corresponding siRNA, suggest-ing that these off-target effects are not directed by seed pairing These results might seem inconsistent with the current description of siRNA off-targeting mechanisms, in which seed regions play a predomi-nant role [6,10] It is possible that these 35% of seed-independent variations represent a secondary effect resulting from the downregulation of the 65% seed-matching off-targets However, as these variations were observed at the earliest tested time point (0 h post-transfection), we could not establish whether these two categories of genes have different kinetics, and thus could not determine the mechanisms gener-ating all siRNA side effects, a point that will require further investigation

Sequence-dependent side effects of siRNAs on gene expression are expected to be identical in different cell types Gene expression profiles obtained for 143B, A549 and IMR-90 cells allow a cell type-to-cell type comparison of siRNA side effects, but only for

Trans-A

B

fection

Extraction

24 h post-T

Extraction

48 h post-T

20 genes

10 genes

4 genes

3 genes SREBF1/siRNA1

NT siRNA

Trans-fection

Extraction

0 h post-T

Extraction

24 h post-T

26 genes

27 genes

12 genes

15 genes

NT siRNA

SREBF1/siRNA1

Fig 8 Effect of two different transfection periods on gene

expres-sion profiles in 143B cells transfected with SREBF1 ⁄ siRNA1 or the

NT siRNA at 100 nM (A) Twenty-four hours of transfection and

RNA extraction 24 or 48 h post-transfection (B) Forty-eight hours

of transfection and RNA extraction 0 or 24 h post-transfection.

Design of the experiments and number of genes differentially

expressed in siRNA-transfected cells when compared with

DF- treated cells.

0.35 5.60

IGFBP5

DF_T24_E24 DF_T24_E48 DF_T48_E0 DF_T48_E24 SREBF1_T24_E24 SREBF1_T24_E48 SREBF1_T48_E0 SREBF_T48_E24 NT_T24_E24 NT_T24_E48 NT_T48_E0 NT_T48_E24

SPARC MMP2 BCL6 S100A4 FOS TFRC ENPP1 PLAU IGFBP4 CTGF CDH11 GSN HSPB1 ITGA5 MMP14 CDKN1B FGF2 PCNA DHFR UNG KIF23 HSPCB MLH1 IGFBP2 CAV1 EF21 BCL2L1 CDC42 PRAME MADH1 BAX CANX MAPK9 UBE2C RAD51 TFDP1 ADPRT BIN1 K-ALPHA-1 CKB ALDOA CDK2 CENPF TFDP2 TERT TGFBR2 FGFR1 CASP3 CTSL ABL1 UBE2V1 RAF1 BSG TIMP1 COL6A2 MAP2K1 EGFR CDH13 PLAUR JUN ITGA6 HPRT1 PLAT WARS TNFRSF10B CCND2 TK1 DUSP1

Fig 9 Effect of two different transfection

periods on gene expression profiles in 143B

cells transfected with SREBF1 ⁄ siRNA1 or

the NT siRNA at 100 nM 143B cells were

incubated for 24 h (T24) or 48 h (T48) with

DF or transfected for 24 h (T24) or 48 h

(T48) with SREBF1 ⁄ siRNA1 or the NT siRNA

at 100 nM RNA was extracted 0 h (E0),

24 h (E24) or 48 h (E48) post-transfection,

reverse transcribed, and processed for

microarray analysis Expression plots

pres-ent the genes displaying significant

differ-ences in relative transcript level between

siRNA-transfected cells and DF-treated cells

(n = 3) Color key: green, downregulation;

red, upregulation A scale for heat maps as

minimum and maximum fold differences is

presented.