Journal of Controlled Release 143 (2010) 350–358

Contents lists available at ScienceDirect

Journal of Controlled Release

j ourna l homepage: www.e lsev ie r.com/ locate / jconre l

GENEDELIVERY

Targeted gene delivery into peripheral sensorial neurons mediated by self-assembledvectors composed of poly(ethylene imine) and tetanus toxin fragment c

Hugo Oliveira a,b, Ramon Fernandez a, Liliana R. Pires a, M. Cristina L. Martins a, Sérgio Simões c,Mário A. Barbosa a,b, Ana P. Pêgo a,⁎a INEB–Instituto de Engenharia Biomédica, Divisão de Biomateriais, Universidade do Porto, Rua do Campo Alegre, 823, 4150-180 Porto, Portugalb Universidade do Porto–Faculdade de Engenharia–Departamento de Engenharia Metalúrgica e Materiais, Rua Roberto Frias, s/n, 4200-465 Porto, Portugalc Centro de Neurociências e Biologia Celular, Universidade de Coimbra, 3004-517 Coimbra, Portugal and Departamento de Tecnologia Farmacêutica, Faculdade de Farmácia,Universidade de Coimbra, 3000-295 Coimbra, Portugal

⁎ Corresponding author. Tel.: +351 226074982; fax:E-mail address: apego@ineb.up.pt (A.P. Pêgo).

0168-3659/$ – see front matter © 2010 Elsevier B.V. Adoi:10.1016/j.jconrel.2010.01.018

a b s t r a c t

a r t i c l e i n f o

Article history:Received 10 November 2009Accepted 12 January 2010Available online 20 January 2010

Keywords:Peripheral nervous systemNon-viralNanoparticleTargetedRegenerative medicine

A simple, safe and efficient system that can specifically transfect peripheral sensorial neurons can bring newanswers to address peripheral neuropathies. A multi-component non-viral gene delivery vector targeted toperipheral nervous system cells was developed, using poly(ethylene imine) (PEI) as starting material. Abinary DNA/polymer complex based on thiolated PEI (PEISH) was optimized, considering complex size andzeta potential and the ability to transfect a sensorial neuron cell line (ND7/23). The 50 kDa non-toxicfragment from tetanus toxin (HC), which has been previously shown to interact specifically with peripheralneurons and to undergo retrograde transport, was grafted to the complex core via a bifunctional PEG (HC-PEG) reactive for the thiol moieties present in the complex surface. Several formulations of HC-PEG ternarycomplexes were tested for targeting, by assessing the extent of cellular internalization and levels oftransfection, in both the ND7/23 and NIH 3 T3 (fibroblast) cell lines. Targeted gene transfer to the neuronalcell line was observed for the complex formulations containing 5 and 7.5 μg of HC-PEG. Finally, our resultsdemonstrate that the developed ternary vectors are able to transfect primary cultures of dorsal root gangliondissociated neurons in a targeted manner and elicit the expression of a relevant neurotrophic factor.

+351 22 6094567.

ll rights reserved.

© 2010 Elsevier B.V. All rights reserved.

1. Introduction

Peripheral nervous system (PNS) problems are common andencompass a large spectrum of traumatic injuries, diseases, tumoursor iatrogenic lesions [1]. Conventional treatments for peripheralneuropathies have primarily been palliative rather than curative, butmost importantly, they have often been ineffective. In the recentyears, enormous progress has been made in our understanding of thebiology of neurotrophic factors and how they may be applied in thetreatment of neurologic diseases. These advances have paved the wayfor new therapeutic approaches that may arrest or reverse the diseaseprocess underlyingmany types of peripheral neuropathies [2]. Studieswith recombinant peptides have demonstrated that a number ofneurotrophic factors, including nerve growth factor, neurotrophin-3,brain derived neurotrophic factor (BDNF), insulin-like growth factorsand vascular epithelial growth factor can prevent the degeneration ofperipheral sensory axons [2,3]. However, these potent peptidescannot be administrated to patients for long periods of time, due tounwanted systemic effects or short systemic half life [4]. The lack of

effective response presented bymodern therapeutics and the crescentneed for new treatments have brought new interest in gene therapyapproaches. The local expression of neurotrophic factors achieved bygene transfer may be used to attain the desired outcome avoidingunwanted adverse effects that could result from the systemicadministration of drugs [5].

The present work aims at the development of a non-viral genedelivery system able to specifically transfect a neuronal cellpopulation with a therapeutic final purpose. In recent years extensivework has been developed in order to accomplish successful genetransfer to mammalian cells. Both viral and non-viral approaches havebeen attempted, but an effective and safe system has not yet beenfound as the drawbacks presented by each of the strategies remain tobe solved. Poly(ethylene imine) (PEI) was introduced by Behr in 1995[6], and has become one of the gold standards of non-viral genedelivery. PEI has been shown to effectively condense plasmids intocolloidal particles able to transfect cells, both in vitro and in vivo [7].These condensed particles are of spherical shape and can be preparedwith a narrow particle size distribution, which presumably allowshigh cellular uptake of the plasmids, resulting in a high transfectionefficiency of the system [7]. However, PEI based complexes interactwith cells in a non-specific manner mainly due to electrostaticinteractions resulting from the complex high positive charge [8]. In

Fig. 1. Branched poly(ethylene imine) thiolation reaction with 2-iminothiolane.

351H. Oliveira et al. / Journal of Controlled Release 143 (2010) 350–358

GENEDELIVERY

order to promote cell specificity, efforts have been made to combineor even exchange the non-specific electrostatic polyplex-cell surfaceinteraction with a specific receptor-mediated cellular uptake, by theincorporation of cell binding ligands into the transfection complexes.A variety of ligands has been successfully coupled to PEI, includinggalactose for hepatocyte targeting [9], mannose for enhanced uptakeby dendritic cells [10], epidermal growth factor for enhanced uptakeby epithelial cells [11], integrin-binding peptides [12] and anti-CD3antibodies for gene delivery to CD3-expressing cells [13]. Efforts havealso been developed in order to attain neuron-specific complexes. Thegrafting of PEI with neurotensin and tet-1 peptides was pursued, andalthough an increased complex uptake was observed in PC-12 anddorsal root ganglia (DRG) primary cultures [14,15], no transfectionwas attained in fully differentiated PC12 cells or DRG cultures.

Neurotoxins have been for long explored to target the nervoussystem. Such an example is the use of the tetanus toxin (TeNT). Whenapplied intramuscularly TeNT binds to the pre-synaptic neuronterminals, and after internalization is trafficked to the cell body byretrograde axonal transport [16], thus allowing the possibility ofdelivery by a minimally invasive method. The non-toxic carboxylicterminal fragment from TeNT (HC) has been chemically or geneticallyconjugated to several enzymes such as superoxide dismutase [17],horseradish peroxidase [18], and β-galactosidase [19]. The in vivosystemic administration of these conjugates resulted in the efficientdelivery to the brain stem, motor neurons of the spinal cord and to theDRG. Fairweather et al. showed that HC-grafted poly(lysine)mediatedtransfection ten times more efficiently than polyplexes without HC, inneuronal cell lines [20].

Focusing on a therapeutic application, we developed a PEI basedsystem to deliver genes to the PNS. Based on the design of a modularstructure, and therefore a more versatile one, we aimed at a targetednon-viral gene delivery system able to undergo retrograde transport.In a first step the core of the complex and its correspondenttransfection efficiency was optimized using PEI as starting material.Polymer thiolation was performed in order to obtain a thiol-functionalized complex core. Upon optimization of the basal formu-lation, the grafting of the targeting moieties–HC fragment–wasperformed via a bifunctional PEG spacer. Using this strategy theamount of protein on the complex surface was easily tuned in order toobtain neuron specificity and transfection.

2. Materials and methods

2.1. Polymer modification

Branched PEI (25 kDa, Sigma) was purified by dialysis, using amembrane with a molecular weight cut-off of 2.5 kDa (Spectrum labs,USA), for 3 days against a 5 mM HCl solution at 4 °C (renewed daily).After freeze-drying, the purified PEI was dissolved in a 5% (w/v)glucose solution (pH 7.4) at 1 mg ml−1. Alternatively, purified PEIwas modified with 2-iminothiolane (Sigma) as described elsewhere[21]. Briefly, a 2-iminothiolane solution in dimethylformamide wasadded to a PEI solution in phosphate saline buffer (0.05 M NaH2PO4/Na2PO4, 0.1 M NaCl, 0.01 M EDTA, pH 7.4), at a ratio of 1 mol of 2-iminothiolane per mole of PEI primary amine, and let to react for 3 hwhile stirring at room temperature (RT). The thiolated PEI (PEISH;Fig. 1) was purified as previously described and freeze-dried. PEISHwas dissolved in a 5% (w/v) glucose solution (pH 7.4) at 1 mg ml−1

concentration and stored at −80 °C until further use.

2.2. Plasmid DNA

The plasmid DNAs used encoded for the β-galactosidase (β-Gal;pCMV-Sport-βGal, 7.8 kb), green fluorescent protein (GFP; pCMV-GFP, 7.4 kb), luciferase (Luc; pCMV-LUC, 6.4 kb) or BDNF (pCMV-BDNF, 7.7 kb). Plasmids were produced in a DH5α competent E. coli

strain transformed with the respective plasmid. Subsequently, DNApurification was performed using an endotoxin-free Maxiprep kitfollowing the manufacturer's instruction (GenElute, Sigma). Plasmidconcentration and purity were assessed by UV spectroscopy. Plasmidsolutions with Abs (260 nm/280 nm) ratio comprised between 1.7and 2.0 were used in all studies.

2.3. Complex core formation (binary complex)

DNA–polymer complexes were prepared as described elsewhere[22] by mixing, while vortexing, equal volumes of plasmid DNA (in5% (w/v) glucose aqueous solution, pH 7.4) and PEI or PEISH solutions.PEI based complexes were allowed to form for 15 min at RT beforefurther use. PEISH based complexes were let to oxidize for 24 h at RT.PEI based complexes with different molar ratios of primary aminegroups (N) to moles of DNA phosphate groups (P)–N/P molar ratio–were prepared. In the case of PEISH based complexes the weight ratioof polymer/DNA was the same as used for the PEI based complexes.

2.4. Ternary complex formation

2.4.1. HC production, purification and modificationThe non-toxic fragment of the tetanus toxin (HC) was produced

recombinantly using the BL21 E. coli strain. The plasmid encoding forthe HC fragment, together with a coding sequence for six histidines inthe N terminal, was a kind offer from Prof. Neil Fairweather (King'sCollege, UK). The HC production in the BL21 E. coli strain andpurification was performed as described by Sinha et al. [23].Subsequently, the obtained HC fragment was covalently linked to aPEG spacer bearing a maleimide (MAL) end group. Briefly, abifunctional 5 kDa PEG (JenkemUSA, China) bearing an N-hydro-xysuccinimide (NHS) and an MAL end group was used as indicated bythe manufacturer, at a 2.5 PEG/HC protein molar ratio. The amount ofreactive MAL groups in the HC protein was determined using amodified Ellman's assay [24], and found to be 1.5±0.4 mol PEG/HC.Hereafter, the PEG-modified HC will be designated as HC-PEG.

2.4.2. Ternary complex formulationThe core complexes were formed using PEISH at an N/Pmolar ratio

of 3 and let to stabilize for 15 min. Subsequently, at a finalconcentration ranging from 1.25 to 7.5 µg per 2 µg of plasmid DNA,HC-PEG was added to the complex mixture and let to react for 24 h atRT. The efficiency of HC-PEG binding to the complex was assessed bymeans of radiolabelling as follows. HC-PEG was labelled with 125Iusing the iodo-gen method [25]. Purification of the 125I-HC-PEG wasperformed using Sephadex G-25 M columns (PD-10, AmershamPharmacia Biotech). The yield of the iodination reaction was 97%, asestimated by trichloroacetic acid precipitation (TCA, 10% v/v, Merck)of the protein. The HC-PEGworking solution was prepared by diluting125I-HC-PEG with non-labelled HC-PEG (1 mg ml−1) to a finalconcentration of (15×106 cpm mg−1) and left at 4 °C. Ternarycomplexes corresponding to 20 μg of DNA were prepared using theHC-PEG working solution, as previously described. In parallel, thesame amount of HC-PEG working solution was diluted to the samefinal volume with 5% (w/v) glucose solution (pH 7.4) and used ascontrols. After 24 h, the complexes and free-125I-HC-PEG dispersions

352 H. Oliveira et al. / Journal of Controlled Release 143 (2010) 350–358

GENEDELIVERY

were separated using a 100 kDa cut-off filter (Vivaspin 500, Sartorius)by centrifugation at 1300 rpm for 30 min, and washed 2 times with0.2 ml of a 5% (w/v) glucose solution (pH 7.4). The eluted solution wastransferred to radioactive immunoassay tubes and the correspondingradioactivities were counted in a gamma-counter. The HC-PEGpercentage of binding to the complex core was determined bycomparing the counts of free 125I-HC-PEG from the ternary complexesand the corresponding control.

2.5. Complex size and zeta potential determination

Complexes were prepared as previously described. Ten μg ofplasmid DNA (pCMV-GFP) was used for each formulation. The zetapotential and size of the complexes were assessed using a ZetasizerNano Zs (Malvern, UK) following the manufacturer's instructions. TheSmoluchowski model was applied for zeta potential determinationand cumulant analysis was used for mean particle size determination.All measurements were performed in triplicate, at 25 °C.

2.6. Cell culture

2.6.1. Cell linesND7/23 (mouse neuroblastoma (N18 tg 2)×rat dorsal root

ganglion neurone hybrid) or NIH 3 T3 (mouse embryonic fibroblast)cell lines (both obtained from ECACC) were routinely cultured inDulbecco's Modified Eagle Medium (DMEM) with Glutamax, supple-mented with 10% (w/v) foetal bovine serum (FBS) (heat inactivatedat 57 °C for 30 min) and 1% PS (10,000U ml−1 penicillin and10,000 μg ml−1 streptomycin), all supplied by Gibco, and maintainedat 37 °C in a 5% CO2 humidified incubator. The ND7/23 cell line waschosen as a sensorial PNS cell model and the NIH 3 T3 cell line as afibroblast model. Cells were routinely tested for mycoplasma bystandard PCR.

2.6.2. DRG primary culture cellsEmbryos (E17), obtained from euthanized pregnant Wistar rats,

were placed in cold Hanks' balanced salt solution (HBSS, Sigma). Thespinal cord with the DRGs attached was isolated from the ventralregion of the embryos. The DRGs were gently detached and placed inHBSS. Subsequently, the DRGs were incubated in 0.1% (w/v) trypsin(Sigma) in HBSS without Ca2+ and Mg2+ (Sigma) for 15 min at 37 °Cand collected by centrifugation (2 min at 1700 rpm). The DRGs werethen dissociated in complete medium (DMEM/F12 high glucose, fromGibco), 10% (w/v) FBS and 1% (w/v) PS) using fired Pasteur pipettes.The obtained cell suspension was plated on a tissue culturepolystyrene (TCPS) flask for 90 min, in order to purify the DRGdissociated culture (DRGc) from TCPS adherent cells. From the finalcell suspension, 2×104 cells cm−2 were seeded in a 24-well plate onglass coverslips treated with poly D-Lysine (PDL, 0.1 mg ml−1, Sigma)and laminin (10 μg ml−1, Sigma), in DMEM/F12 supplemented with50 ng ml−1 of nerve growth factor (NGF 7s; Invitrogen) and 5-fluoro-2′-deoxyuridine (60 µM, Sigma). Medium was supplemented withNGF (50 ng ml−1) every 2 days and renewed every week. Thisprotocol allowed a neuron purity >80% in DRGc, at 24 h post-plating,as determined by immunocytochemistry. Briefly, DRGc were fixedwith 4% (w/v) paraformaldehyde in phosphate-buffered saline (PBS)for 15 min at RT, permeabilized for 20 min with PBS containing0.1% (v/v) Tween 20 and blocked for 30 min with PBS containing3% BSA (w/v). Cells were incubated with rabbit polyclonal anti-200 kDa Neurofilament (N200; 1:1000; Abcam, UK) in PBS with3% (w/v) BSA overnight. After washing with PBS, cells were incubatedwith goat anti-rabbit Alexafluor 488 labelled antibody (1:500,Invitrogen) in a 1% (w/v) BSA solution on PBS for 1 h. Finally, thecell nuclei were stained with DAPI (0.1 µg ml−1, Sigma) and mountedusing Vectashield (Vector, UK). The percentage of DRG cells wasdetermined by counting 5 different fields per replicate (n=3).

2.7. HC binding studies

For this study the HC protein was labelled with a rhodaminefluorochrome. Briefly, 1 mg of rhodamine-NHS (ROX-5(6)-Carboxy-X-rhodamine N-succinimidyl ester, Sigma) dissolved in 500 µl ofanydride dimethyl formamide (Sigma) was added drop wise to 6 mgof HC (1 mg ml−1) in buffer A (300 mM NaCl, 50 mM NaH2PO4, pH8.0) and left to react for 1 h. The obtained solution was washed andconcentrated using a 30 kDa cut-off filter (Millipore, Ireland). Afterfiltration (0.22 μm, TPP), the rhodamine-labelled HC (HC-rox)concentration was assessed using the BCA assay (Pierce, USA), set to1 mg ml−1 with buffer A and stored at −20 °C until further use.

To investigate the cell binding specificity of the HC fragment, cellswere incubated with HC-rox. The ND7/23 and NIH 3 T3 cell lines wereseeded at a cell density of 2.0×104 and 2.5×104 cells cm−2,respectively, on glass coverslips coated with PDL placed in a 24-wellplate. Briefly, 24 h post-seeding, cells were incubated at 4 °C for15 min and then washed twice with cold HBSS before a 1 h incubationat 4 °C with HC-rox at a concentration of 5 µg ml−1 in HBSScontaining 0.1% (w/v) BSA. Additionally, in the case of the ND7/23cell line, a control was performed in which a 15-min incubation with10 µg ml−1 of unlabelled HC was performed prior to the incubationwith HC-rox. After washing twice with HBSS, cells were fixed with4% (w/v) paraformaldehyde in PBS for 15 min at RT, stainedwith DAPI(1 µg ml−1, Sigma) for 5 min and mounted on slides with Vectashield(Vector). Cells were analyzed using an inverted fluorescent micro-scope (Axiovert 200M, Zeiss, Germany).

2.8. Cellular internalization studies

ND7/23 and NIH 3 T3 cells were sub-cultured in 24-well TCPS platesfor 24 h as described in Section 2.7. Cells were exposed, for 1 h at 37 °C,to ternary complexes (2 µg of plasmid DNA/well) formulated withdifferent HC-PEG amounts, respectively, 0, 1.25, 2.5, 5 and 7.5 µg per2 µg of DNA. Plasmid DNA was labelled with YOYO-1 (Invitrogen, USA)as indicated by themanufacturer (1 mol YOYO-1 per 100 bp of plasmidDNA). For the competition assays, prior addition of the ternarycomplexes, cells were incubated for 20 min at 4 °C with a 100-foldexcess of free HC or a 5-fold excess of the correspondent unlabeledternary complex. For complex internalization assessment, cells wereincubated with a trypan blue solution (0.2 mgml−1 in PBS, Sigma) for5 min, trypsinized and processed for flow cytometry. Ten thousandgated events were taken for each replicate (n=3) using a FACSCaliburcytometer (BD Biosciences, USA) and analyzed by histogram forpositivity for YOYO-1 using the FlowJo software (version 8.3.7, USA).

2.9. Transfection studies

2.9.1. Cell linesThe ND7/23 and NIH 3 T3 cells were sub-cultured for 24 h as

described for the HC binding studies. Two hours prior transfection,medium was removed and replaced by 0.5 ml of fresh medium.Different N/P ratios of the core complexes were tested (2, 3, 6 and 12)using PEI and PEISH. For the ternary complexes different formulations(0, 1.25, 2.5, 5 and 7.5 µg of HC-PEG per 2 µg of plasmid DNA) weretested. Cells were exposed, for 6 h at 37 °C, to complexes containing2 µg of plasmid DNA before the medium was renewed. Cells wereincubated for an additional 48, 72 or 96-hour period and processed fortransfection activity or efficiency assessment, using β-galactosidase orGFP as the reporter genes.

The transfection activity, corresponding to the β-Gal activity(ortho-nitrophenyl-β-galactoside (ONPG) hydrolyses), was deter-mined by an enzymatic assay (Promega). All experiments wereperformed in triplicate and expressed in terms of specific activity(nmol of ONPG hydrolyzed/min/mg total protein). BCA assay (Pierce)was used to quantify the total protein content. For transfection

353H. Oliveira et al. / Journal of Controlled Release 143 (2010) 350–358

GENEDELIVERY

efficiency assessment, cells were analyzed by flow cytometry for GFPexpression. Twenty thousand gated events were taken for eachreplicate (n=3) as previously described.

2.9.2. Primary culturesDRGc were seeded at a density of 2.0×104 cells cm−2 and 3 to

4 days post-seeding they were incubated with complexes for 6 h.Complexes based on PEISH prepared at an N/P=3, with or without7.5 μg of HC-PEG moieties, were tested. 72 h post-transfection bothluciferase, GFP or BDNF gene expression was determined, and 96 hpost-transfection the expression of the first gene. The luciferase assaywas performed according to the manufacturer instructions (Pro-mega). In these conditions, 5 pg of Luciferase (Sigma) had a value of703 Relative light units (RLU). RLU where collected for 10 s andnormalized to the cell extract total protein. Immunostaining wasperformed to assess the percentage of GFP positive cells. Discrimina-tion between neurons and non-neuron cells was achieved by N200staining. Briefly, DRGc were fixed, permeabilized for 20 min with PBScontaining 0.1% (v/v) Tween 20 and blocked for 30 min with PBScontaining 3% BSA (w/v). Biotin conjugated mouse monoclonal anti-GFP (1:200; Novus Biologicals, USA) and N200 rabbit polyclonal(1:1000; Abcam, UK) were incubated in PBS-3% BSA overnight. Afterwashing with PBS, cells were incubated with PBS containing 1% (v/v)H2O2 for 10 min, then, with peroxidase coupled ABC complex(Vectastain Elite, Vector, UK) for 30 min and finally with DAB/Nisolution (Vector, UK). To reveal the N200 staining, the DRGc wereincubated with goat anti-rabbit Alexafluor 488 labelled antibody in a1% (w/v) BSA solution in PBS (1:500, Invitrogen) for 1 h. Finally, thecells were stainedwith DAPI (0.1 µg ml−1, Sigma) andmounted usingVectashield. The percentage of transfected cells, DAB/Ni positive, wasdetermined by counting 5 different fields per replicate (n=3). Imageswere collected using an inverted fluorescence microscope (Axiovert200 M, Zeiss, Germany). BDNF protein secretion was determined incell media (24-hour culture period), using a BDNF enzyme-linkedimmunosorbant assay (ELISA) kit (Chemikine, Chemicon) followingthe manufacturer instructions and expressed as pg ml−1.

2.10. Cytotoxicity assay

To assess cell viability after transfection, a resazurin based assaywas used [26]. Briefly, 24 h post-transfection a solution of resazurin(0.1 mgml−1 in PBS, Sigma)was added to eachwell to a final 10% (v/v)concentration. After 4 h of incubation at 37 °C, 200 µL of the mediumwas transferred to a black-walled 96-well plate (Greiner) andfluorescence was measured (λexc=530 nm, λem=590 nm, SpectraMax GeminiXS — Molecular Devices). Results were expressed aspercentage of metabolic activity of treated cells relative to untreatedcells.

2.11. Statistical analysis

Using the Graphpad Prism 5.0 software, statistically significantdifferences between several groups were analyzed by non-parametricKruskal–Wallis test, followed by Dunns post-test. Non-parametricMann–Whitney test was used to compare two groups. A p value lowerthan 0.05 was considered statistically significant.

3. Results

3.1. Complex core characterization and transfection evaluation

As determined by the Ellman's assay [24], 7.8±0.2% of primaryamines of PEI were substituted by thiol groups in the PEISH polymer.

Cationic polymers like PEI form self-assembled complexes withanionic plasmid DNA by electrostatic interactions. When DNA isloosely complexed or in a free form, it freely migrates through an

agarose gel. When PEI and PEISH based complexes were applied in anagarose gel and submitted to an electric field, no free DNA wasdetected for all N/P molar ratios tested (Fig. S1, Supplementary data(SD)). The complex size and zeta potential were determined as afunction of the N/P molar ratio used in the preparation of the binarycomplexes. No significant differences were found in terms ofpolydispersity index (Pdi) between PEI and PEISH based complexes.The Pdi values ranged between 0.222 and 0.409 for all N/P molarratios tested. PEI based complexes ranged from 55±3 to 130±5 nmin diameter (Fig. S2 A, SD) and−26±1 to 46±1 mV in terms of zetapotential (Fig. S2 B, SD), whichwas directly relatedwith the N/Pmolarratio. PEI thiolation enabled the formation of significantly smallerparticles at N/P molar ratios of 1 and 2 (Fig. S2 A, SD), as well as withlower zeta potential values for N/P molar ratios ≥2 (Fig. S2 B, SD).

Transfection activity mediated by PEI and PEISH based complexesprepared with N/P molar ratios of 2, 3, 6 and 12 was assessed in theND7/23 cell line, 48 and 72 h post-transfection (Fig. S3, SD). β-Galactivity was significantly higher for PEISH based vectors for N/P=2, atboth time points tested. For N/P=3, the transfection activity wasfound to be similar for both polymer systems. At higher N/P values,the transfection activity was significantly reduced when mediatedby PEISH based complexes. Relative viability assessed 24 h post-transfection showed significant cell toxicity of the complexesprepared at N/P molar ratios higher than 3 (Fig. S4, SD).

Taking together the results of the core complex characterization,the cytocompatibility and transfection profile, we selected the N/Pmolar ratio of 3 as the formulation for the development of the ternarycomplex. By FACS analysis, the transfection efficiency of the thiolatedbinary complexes prepared at an N/P=3 was found to be 13.3±0.5%and 12.8±0.8% on the ND7/23 cell population, at 48 and 72 h post-transfection, respectively (data not shown).

3.2. HC fragment cell specificity

The determination of the binding specificity of the purified HCprotein to the ND7/23 cell line is an important step in order to validatethis cell line as an in vitro model for neurospecificity evaluation of thedeveloped complexes. The HC fragment of the tetanus toxin wasfluorescently labelled with rhodamine (HC-rox) and found tospecifically bind to the ND7/23 cell line (Fig. 2A), as compared tothe control NIH 3 T3 cell line (Fig. 2C). When cells were pre-incubatedwith a 100-fold excess of unlabelled HC, a significant reduction of thefluorescent signal was observed (Fig. 2B).

3.3. Ternary complex formation

The strategy followed for the formation of a ternary polyplexconsiders two steps. In a first step the polymer was thiolated in order toobtain a thiol-decorated polyplex. The addition of the HC fragment wasperformed via a 5 kDa bifunctional poly(ethylene glycol) (PEG) linkerreactive for the thiol groups present in the binary complexes (Fig. 3).

The coupling efficiency of HC-PEG to the complexes, as determinedby radiolabelling, ranged between 89±3% and 95±3% (Table 1).

The characterization of the ternary complexes by light scatteringshowed that the particle size peaked for the 1.25 µg HC-PEGformulation (Fig. 4A). For higher HC-PEG content, a reduction of thecomplex size was observed, which stabilized in the same size range ofthe binary complex (Fig. 4A). Additionally, no significant differenceswere found in terms of Pdi for the complex formulations tested(Fig. 4A). The addition of HC-PEG to the complexes, in theconcentration range tested,was shown tohave no significant influenceon the complex zeta potential (Fig. 4B), although a reduction tendencywas observed for both ternary formulations prepared with the lowestHC-PEG content (i.e. 1.25 and 2.5 µg HC-PEG).

Fig. 2.Microscope images corresponding to the binding affinity of the rhodamine-labelled tetanus toxin fragment (HC-rox). Lane A corresponds to the ND7/23 cell line, lane B to theND7/23 cell line previously incubated with an excess of unlabelled tetanus toxin fragment and lane C to the NIH 3T3 cell line. Scale bar=50 µm.

354 H. Oliveira et al. / Journal of Controlled Release 143 (2010) 350–358

GENEDELIVERY

3.4. Ternary complex internalization

In the ND7/23 cell line, the extent of ternary complex cellinternalization was found to slightly decrease for the two formula-tions prepared with the lowest HC-PEG concentrations tested (1.25and 2.5 µg of HC-PEG), when compared with the binary polyplexes(Fig. 5A). However, for the two highest concentrations of HC-PEGtested (5 and 7.5 µg of HC-PEG per formulation), the same extent ofcell internalization values as for the binary complexes were observed(Fig. 5A). In contrast, the increase of the HC-PEG content in thecomplexes resulted in an opposite effect for the NIH 3 T3 cell line.Although the two formulations prepared with the lowest HC-PEGconcentration (1.25 and 2.5 µg of HC-PEG) had similar internalizationpattern as the binary complexes, the formulations with the highestHC-PEG concentration were significantly less internalized by thisfibroblast cell line (Fig. 5B).

The receptor-mediated internalization of the formulation pre-pared with 7.5 μg HC-PEGwas further tested by means of competitiontests. Following pre-incubationwith a 100-fold excess of HC fragment,the ternary complex internalization was significantly reduced in theND7/23 cells (Fig. 6B and C). When pre-incubated with a 5-fold excessof the correspondent unlabelled ternary complexes, a similar effectwas observed, even though to a lower extent (Fig. 6B and D).

Fig. 3. Proposed model for the formation of a ternary complex. Two steps are considered: thethe tetanus toxin fragment (HC) is coupled to the polyplex surface via a maleimide (MAL)

Additional internalization studies with the binary formulation ofPEISH (N/P=3) were performed in the ND7/23 cell line pre-incubated with a 100-fold excess of HC fragment. In this case theinternalization levels were found not to significantly vary from theones shown in Fig. 5A for this formulation.

3.5. Transfection efficiency of cell lines mediated by ternary complexes

In the ND7/23 cell line, at 48 h post-transfection, approximately10% of the cell population was positive for GFP after transfection withthe binary complexes (Fig. 7A). The introduction of HC moieties at thecomplex surface initially leads to a reduction of transfection(formulations containing 1.25 and 2.5 μg of HC-PEG) at 48 h post-transfection. Nevertheless, at higher HC protein content (formulationscontaining 5 and 7.5 μg of HC-PEG), the transfection efficiency wassignificantly increased when compared with the binary formulation(Fig. 7A). In the case of the 5 μg HC-PEG formulation a slight increasein transfection efficiency along time was observed, in contrast withthe clear reduction in the case of the 7.5 μg HC-PEG formulation(Fig. 7A). For the NIH 3 T3 cells, a decrease in transfection efficiencywas seen with the increase of HC protein content on the complexes,which was significantly reduced for the two highest HC-PEG contentcomplex formulations (Fig. 7B). A significant decrease in cell viability

introduction of thiol moieties in the binary polyplex and a subsequential step in whichgrafted (thiol reactive) 5 kDa PEG linker.

Table 1Theoretical thiol/HC-PEG fold content for each formulation tested and percentage ofHC-PEG coupled to the complexes, as determined by radiolabelling.

Footnotes: SH— thiol groups; HC— tetanus toxin fragment c; PEG— polyethyleneglycol; PEISH— thiolated poly(ethylene imine).

355H. Oliveira et al. / Journal of Controlled Release 143 (2010) 350–358

GENEDELIVERY

was observed in the ND7/23 cell line for the highest HC-PEGformulation at 24 h post-transfection, when comparedwith untreatedcells (Fig. S5 A, SD).

Fig. 5. Extent of cell internalization of ternary complexes preparedwith PEISH (N/P=3)and with increasing amounts of HC-PEG in (A) ND7/23 and (B) NIH 3T3 cell lines,respectively (n=3, Aver±SD; ⁎ denotes statistically difference between groups and thebinary formulation p<0.05).

3.6. Ternary complex transfection in primary cultures

The 7.5 µg HC-PEG ternary complex formulation was tested indissociated DRG cultures and the transfection mediated by thesecomplexes evaluated in terms of luciferase expression, 72 and 96 hpost-transfection. As seen in Fig. 8A, 72 h post-transfection thetransfection activity was significantly reduced for the HC-PEG graftedcomplexes, in comparison with the binary complexes. Nevertheless,96 h post-transfection no significant differences were observedbetween groups (Fig. 8A). GFP immunostaining revealed that at theearly evaluation point no significant differences were observed interms of the % of total transfected cells (Fig. 8B). Nevertheless, bydiscriminating the neuron from non-neuron cell populations, onecould observe significant differences between the 0 and 7.5 µg of HC-PEG formulations, with p=0.0044 and p=0.0349, respectively(Fig. 8C).

Fig. 4. Characterization of ternary complexes based on PEISH (N/P=3) as a function ofµg of HC-PEG per formulation (A) size and Pdi and (B) zeta potential (n=3, Aver±SD).

The non-neuron cell transfection was reduced while the neuroncell transfection was increased when the HC-PEG moiety wasintroduced to the complexes (Fig. 8C–I). No significant toxicity wasfound in terms of cell metabolic activity for both PEISH and PEISH7.5 μg HC-PEG (Fig. S7, SD).

Additionally, the ability of the PEISH based ternary complexes tomediate the transfection of the DRGc with a plasmid encoding for theBDNF was also tested. As seen in Fig. 9 PEISH based complexes wereable to elicit significantly higher BDNF expression (>100 pg ml−1) inrelation to naked DNA.

In addition, BDNF levels of cells treated with vectors transportingthe GFP genewere also tested in order to assess the effect of the vectoralone in endogenous BDNF expression. The levels of BDNF wereshown not to significantly vary from untreated cells (data not shown).

4. Discussion

A simple, safe and efficient system that can specifically transfectperipheral sensorial neurons can bring new answers to addressperipheral neuropathies. Towards the achievement of such a system,our approach is based on the use of the cationic polymer PEI, whichwas shown to be an efficient transfection agent, both in vitro [6,27]and in vivo [28–30]. However, the use of PEI based complexes cancompromise the intended cellular targeting, due to their excessivepositive charge [8].

In order to develop neuron specific complexes, the thiolation of PEIwas performed to allow the subsequent functionalization of PEI:DNAcomplexes with a neuron targeting protein. Firstly, the N/Pmolar ratioof thiolated PEI based complexes was optimized aiming at establish-ing the required complex properties to attain efficient transfection ofneuronal cells and, in a second stage a non-toxic fragment from thetetanus toxin (HC)was grafted to the surface of the nanocomplexes, inorder to prone these specifically for neurons.

Fig. 6. Internalization in the ND7/23 cell line of ternary complexes based on PEISH N/P=3; (A) ND7/23 isotype; (B) 7.5 μg HC-PEG formulation; (C) competition with a 100-foldexcess of free HC and (D) with a 5-fold excess of the correspondent unlabelled ternary complexes. (n=3, Aver±SD; ⁎ denotes statistically difference between groups, p<0.05).

356 H. Oliveira et al. / Journal of Controlled Release 143 (2010) 350–358

GENEDELIVERY

A binary complex based on PEISH prepared at an N/P ratio of 3 wasfound to be optimal in terms of size and zeta potential characteristicsand ability to transfect the ND7/23 cell line. Moreover, complexesprepared at this N/P molar ratio were found to be non-toxic under thetested conditions. PEI thiolation has been previously explored in orderto develop triggered complexes that could be stable in the oxidizingextracellular environment but unstable in the reducing intracellularenvironment [31]. Carlisle et al. [21] showed that thiolated PEI based

Fig. 7. GFP positive cells at 48, 72 and 96 h post-transfection of ternary complexes basedon PEISH (N/P=3) as a function of HC-PEG in (A) ND7/23 and (B) NIH 3T3 cell lines.(n=3, Aver±SD; ⁎ denotes statistically difference between groups and control,p<0.05).

complexes, when used at a total amine to phosphate molar ratio≥10,inhibited complex transfection capacity, indicating an over stabiliza-tion of the complex. Our results are in agreement with this study,showing that for N/P molar ratios higher than 3 (corresponding to atotal amines to phosphate molar ratio of 12) a significant reduction oftransfection activity was observed in the ND7/23 cell line.

Having established the conditions for the preparation of the core ofthe complex, the grafting of the proteinmoieties was pursued. The useof a PEG spacer is expected to increase complex hydrophilicity andtherefore to enhance complex stability in an aqueous environment, aswell as increase the protein moieties exposure. The grafting of thebifunctional PEG to HC was optimized in order to obtain approxi-mately one active PEG per protein. The amount of HC-PEG coupled tothe complexes was found to be over 89% for all the preparedformulations. The ternary complex size, zeta potential and Pdi wereshown not to vary significantly for the formulations prepared withHC-PEG content higher than 2.5 µg, when compared to the binarycomplex.

The binding specificity of the developed ternary complexes wastested both in the ND7/23 sensorial neuron cell model and NIH 3 T3control cell line. A significant decrease in the extent of complexinternalization with increasing amounts of HC-PEG was found for theNIH 3 T3 cells, clearly indicating the targeting potential of the ternarycomplexes. Additionally, one could observe that the high levels ofternary complex internalization were reduced both by the competi-tion with free HC or the correspondent unlabeled ternary complexes(tested for the 7.5 µg HC-PEG formulation). Upon increase of HC-PEGmoieties, the transfection efficiency was augmented in the ND7/23cell line (up to 7 fold), whereas in the NIH 3 T3 cells a significantdecrease was observed, corroborating the results obtained in theinternalization and competition studies. Our results indicate thatalthough the overall levels of internalization are at constant levels inthe ND7/23 cell line, the incorporation of HC-PEG moieties clearlytriggers a specific internalization of the complexes.

Signs of toxicity were observed in the ND7/23 cells in theconditions that mediated the higher transfection levels (7.5 µg HC-PEG formulation). Considering that the levels of gene expressionwerealso the highest for this formulation (see Fig. S6, SD), one can associatethis toxic effect with the high levels of GFP protein, which has beendescribed as cytotoxic when over-expressed [32–34].

Fig. 8. A) Transfection activity of complexes based on PEISH at N/P=3 with or without HC-PEG moieties, at 72 and 96 h post-transfection in dissociated dorsal root ganglia cultures(DRGc; Aver±SD; ⁎p<0.05); B) transfection efficiency of the complexes at 72 h post-transfection in DRGc (median±interquartile range), C) classification of DRGc transfected cellsas neuron and non-neuron cells; D, E, F) immunocytochemistry of DRGc with N200 and DAPI for untreated cells, and cultures treated with binary complexes (PEISH) and ternarycomplexes with 7.5 µg HC-PEG, respectively (bar=50um). G, H, I) immunocytochemistry of DRGc with anti-GFP (DAB/Ni) for untreated, PEISH and PEISH with 7.5 µg HC-PEG,respectively. Scale bar=50 μm.

357H. Oliveira et al. / Journal of Controlled Release 143 (2010) 350–358

GENEDELIVERY

Both transfection activity and efficiency mediated by the binaryand ternary complexes (7.5 µg HC-PEG formulation) were assessed ina DRG primary culture, in order to evaluate the transfection specificityof the developed ternary vectors. Transfection activity, as evaluated by

Fig. 9. BDNF secretion (24-hour period) in dissociated dorsal root ganglia cultures(DRGc) mediated by complexes based on PEISH at N/P=3 with or without HC-PEGmoieties, at 72 h post-transfection, the naked plasmid encoding for the BDNF was usedas control (DNA; Aver±SD; ⁎p<0.05).

luciferase assay, was shown to be significantly lower or similar, at 72and 96 h post-transfection, respectively, when cells were treated withtargeted complexes with relation to the binary complexes. Further-more, in terms of transfection efficiency, the % of total transfected cellswas found not to significantly vary between the binary and theternary complexes. However, when cell population discriminationwas undertaken–neuron versus non-neuron cells–a significant in-crease in the % of transfected neurons together with an inverselyproportional decrease in the % of transfected non-neuron cells wasobserved for the ternary complexes. Taken together, these findingsindicate that upon functionalization of the complexes, transfectionwas promoted in primary neurons, thus reinforcing the targetingpotential of the developed vectors. Moreover, we showed that thedeveloped targeted vectors could transfect DRG primary cultures witha plasmid enconding for BDNF, leading to the efficient expression ofthis relevant neurotrophic factor.

Pun et al. [14] showed that targeted cellular binding to differen-tiated PC-12 and dissociated DRG cells can be obtained by grafting PEIwith the tet-1 peptide, a peptide obtained by phage display exhibitingsimilar binding activities to tetanus toxin [35]. Nevertheless, notransfection was observed in fully differentiated PC-12 or dissociatedDRG cultures [14]. We hypothesise that by retaining the retrogradetransport ability [36], the HC fragment will enable an improvedintracellular trafficking of the ternary complexes, thus leading toefficient transfection of neuron cells. Further studies are currentlybeing performed to shade light over this process.

358 H. Oliveira et al. / Journal of Controlled Release 143 (2010) 350–358

GENEDELIVERY

By using an integrative approach we have developed a multi-component cell-targeted gene delivery vector that can be formed bythe simple mixture of the different components. To our knowledgethis is the first report of a non-viral vector not only able to transfectDRG neuron cells, but also in a targeted fashion. The application of thedeveloped ternary complexes in vivo in a regeneration model,following a minimally invasive application is currently beinginvestigated. In addition, both the stability of the ternary complexesin an in vivo environment and their ability to undergo retrogradetransport will be studied.

5. Conclusions

Neuron targeted gene delivery systems encounter numerousbarriers implying newer approaches in their development. In thisstudy, we proposed a novel multi-component nanoparticle systemthat successfully targeted and transfected neuronal cell lines as well asDRG neuron primary cells. Being able to target the PNS, this systembrings new perspectives for the in vivo application of gene therapyapproaches towards PNS mediated by non-viral vectors.

Acknowledgements

This project was carried out under the Portuguese Foundation forScience and Technology (FCT) contract POCI/SAU-BMA/58170/2004.The authors would like to thank Manuela Brás (INEB) for the FTIRmeasurements, Patrícia Cardoso (INEB) and Susana Carrilho (IBMC) forthemycoplasma screeningof the cell cultures andRaquelGonçalves andSimonMonard for the precious helpwithflow cytometry. HugoOliveiraand Liliana R. Pires acknowledge FCT for their PhD scholarships (SFRH/BD/22090/2005 and SFRH/BD/46015/2008, respectively).

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at doi:10.1016/j.jconrel.2010.01.018.

References

[1] F.J. Rodriguez, A. Valero-Cabre, X. Navarro, Regeneration and functional recoveryfollowing peripheral nerve injury, Drug Discov. Today 1 (2) (2004) 177–185.

[2] Z. Sahenk, Neurotrophins and peripheral neuropathies, Brain Pathol. 16 (4)(2006) 311–319.

[3] S.C. Apfel, J.A. Kessler, Neurotrophic factors in the therapy of peripheralneuropathy, Baillieres Clin. Neurol. 4 (3) (1995) 593–606.

[4] S.C. Apfel, Is the therapeutic application of neurotrophic factors dead? Ann.Neurol. 51 (1) (2002) 8–11.

[5] J.C. Glorioso, M. Mata, D.J. Fink, Therapeutic gene transfer to the nervous systemusing viral vectors, J. Neurovirol. 9 (2) (2003) 165–172.

[6] O. Boussif, F. Lezoualc'h, M.A. Zanta, M.D. Mergny, D. Scherman, B. Demeneix,et al., A versatile vector for gene and oligonucleotide transfer into cells in cultureand in vivo: polyethylenimine, Proc. Natl. Acad. Sci. U.S.A. 92 (16) (1995)7297–7301.

[7] J.L. Coll, P. Chollet, E. Brambilla, D. Desplanques, J.P. Behr, M. Favrot, In vivodelivery to tumors of DNA complexed with linear polyethylenimine, Hum. GeneTher. 10 (10) (1999) 1659–1666.

[8] D.V. Schaffer, D.A. Lauffenburger, Optimization of cell surface binding enhancesefficiencyand specificity ofmolecular conjugate genedelivery, J. Biol. Chem. 273 (43)(1998) 28004–28009.

[9] M.A. Zanta, O. Boussif, A. Adib, J.P. Behr, In vitro gene delivery to hepatocytes withgalactosylated polyethylenimine, Bioconjug. Chem. 8 (6) (1997) 839–844.

[10] S.S. Diebold, H. Lehrmann, M. Kursa, E.Wagner, M. Cotten, M. Zenke, Efficient genedelivery into human dendritic cells by adenovirus polyethylenimine andmannosepolyethylenimine transfection, Hum. Gene Ther. 10 (5) (1999) 775–786.

[11] T. Blessing, M. Kursa, R. Holzhauser, R. Kircheis, E. Wagner, Different strategies forformation of PEGylated EGF-conjugated PEI/DNA complexes for targeted genedelivery, Bioconjug. Chem. 12 (4) (2001) 529–537.

[12] P. Erbacher, J.S. Remy, J.P. Behr, Gene transfer with synthetic virus-like particles viathe integrin-mediated endocytosis pathway, Gene Ther. 6 (1) (1999) 138–145.

[13] R. Kircheis, A. Kichler, G. Wallner, M. Kursa, M. Ogris, T. Felzmann, et al., Couplingof cell-binding ligands to polyethylenimine for targeted gene delivery, Gene Ther.4 (5) (1997) 409–418.

[14] I.K. Park, J. Lasiene, S.H. Chou, P.J. Horner, S.H. Pun, Neuron-specific delivery ofnucleic acids mediated by Tet1-modified poly(ethylenimine), J. Gene Med. 9 (8)(2007) 691–702.

[15] E.J. Kwon, J.M. Bergen, I.K. Park, S.H. Pun, Peptide-modified vectors for nucleic aciddelivery to neurons, J. Control. Release 132 (3) (2008) 230–235.

[16] D.L. Price, J. Griffin, A. Young, K. Peck, A. Stocks, Tetanus toxin: direct evidence forretrograde intraaxonal transport, Science 188 (4191) (1975) 945–947.

[17] D.M. Figueiredo, R.A. Hallewell, L.L. Chen, N.F. Fairweather, G. Dougan, J.M. Savitt,et al., Delivery of recombinant tetanus-superoxide dismutase proteins to centralnervous system neurons by retrograde axonal transport, Exp. Neurol. 145 (2 Pt 1)(1997) 546–554.

[18] P.S. Fishman, J.M. Savitt, D.A. Farrand, Enhanced CNS uptake of systemicallyadministered proteins through conjugation with tetanus C-fragment, J. Neurol.Sci. 98 (2–3) (1990) 311–325.

[19] L. Coen, R. Osta, M. Maury, P. Brulet, Construction of hybrid proteins that migrateretrogradely and transynaptically into the central nervous system, Proc. Natl.Acad. Sci. U.S.A. 94 (17) (1997) 9400–9405.

[20] A. Knight, J. Carvajal, H. Schneider, C. Coutelle, S. Chamberlain, N. Fairweather,Non-viral neuronal gene delivery mediated by the HC fragment of tetanus toxin,Eur. J. Biochem. 259 (3) (1999) 762–769.

[21] R.C. Carlisle, T. Etrych, S.S. Briggs, J.A. Preece, K. Ulbrich, L.W. Seymour, Polymer-coated polyethylenimine/DNA complexes designed for triggered activation byintracellular reduction, J. Gene Med. 6 (3) (2004) 337–344.

[22] D. Goula, J.S. Remy, P. Erbacher, M. Wasowicz, G. Levi, B. Abdallah, et al., Size,diffusibility and transfection performance of linear PEI/DNA complexes in themouse central nervous system, Gene Ther. 5 (5) (1998) 712–717.

[23] K. Sinha, M. Box, G. Lalli, G. Schiavo, H. Schneider, M. Groves, et al., Analysis ofmutants of tetanus toxin Hc fragment: ganglioside binding, cell binding andretrograde axonal transport properties, Mol. Microbiol. 37 (5) (2000) 1041–1051.

[24] G.L. Ellman, Tissue sulfhydryl groups, Arch. Biochem. Biophys. 82 (1) (1959)70–77.

[25] Iodine-125, a Guide to Radioionation Techniques, Amersham Life Science, 1993.[26] J. O'brien, I. Wilson, T. Orton, F. Pognan, Investigation of the Alamar Blue

(resazurin) fluorescent dye for the assessment of mammalian cell cytotoxicity,Eur. J. Biochem. 267 (17) (2000) 5421–5426.

[27] H. Pollard, J.S. Remy, G. Loussouarn, S. Demolombe, J.P. Behr, D. Escande,Polyethylenimine but not cationic lipids promotes transgene delivery to thenucleus in mammalian cells, J. Biol. Chem. 273 (13) (1998) 7507–7511.

[28] M. Neu, D. Fischer, T. Kissel, Recent advances in rational gene transfer vectordesign based on poly(ethylene imine) and its derivatives, J. Gene Med. 7 (8)(2005) 992–1009.

[29] S. Moffatt, C. Papasakelariou, S. Wiehle, R. Cristiano, Successful in vivo tumortargeting of prostate-specific membrane antigen with a highly efficient J591/PEI/DNA molecular conjugate, Gene Ther. 13 (9) (2006) 761–772.

[30] R. Kircheis, E. Ostermann, M.F. Wolschek, C. Lichtenberger, C. Magin-Lachmann, L.Wightman, et al., Tumor-targeted gene delivery of tumor necrosis factor-alphainduces tumor necrosis and tumor regression without systemic toxicity, CancerGene Ther. 9 (8) (2002) 673–680.

[31] G. Bellomo, M. Vairetti, L. Stivala, F. Mirabelli, P. Richelmi, S. Orrenius,Demonstration of nuclear compartmentalization of glutathione in hepatocytes,Proc. Natl. Acad. Sci. U.S.A. 89 (10) (1992) 4412–4416.

[32] H.S. Liu, M.S. Jan, C.K. Chou, P.H. Chen, N.J. Ke, Is green fluorescent protein toxic tothe living cells? Biochem. Biophys. Res. Commun. 260 (3) (1999) 712–717.

[33] M. Baens, H. Noels, V. Broeckx, S. Hagens, S. Fevery, A.D. Billiau, et al., The dark sideof EGFP: defective polyubiquitination, PLoS One 1 (2006) e54.

[34] R.R. Taghizadeh, J.L. Sherley, CFP and YFP, but not GFP, provide stable fluorescentmarking of rat hepatic adult stem cells, J. Biomed. Biotechnol. 2008 (2008)453590.

[35] J.K. Liu, Q.S. Tenga, M. Garrity-Moses, T. Federici, D. Tanase, M.J. Imperiale, et al., Anovel peptide defined through phage display for therapeutic protein and vectorneuronal targeting, Neurobiol. Dis. 19 (3) (2005) 407–418.

[36] P.S. Fishman, D.R. Carrigan, Retrograde transneuronal transfer of the C-fragmentof tetanus toxin, Brain Res. 406 (1–2) (1987) 275–279.