Rapid crossing of the pulmonary endothelial barrier by polyethylenimine/DNA complexes


Intravenous administration could become a delivery route of choice for prophylactic and curative gene therapies on condition that genes cross the capillary barrier and reach target tissues without being degraded. We investigated the kinetics and process of transgene delivery through mouse lung capillaries following DNA complexation with linear polyethylenimine (L-PEI) and intravenous injection. Using digoxin-labeled DNA we followed the cellular localization of DNA at different times after injection and correlated these findings with cell markers and transgene expression. At 2 h after injection some DNA was still localized on the interior of the capillary lumen, but other complexes had already crossed the barrier and resulted in gene expression. At 24 h after injection most labeled DNA was localised in pulmonary cells, as was transgene expression. Only rarely was transgene expression found in endothelial cells, suggesting that the complexes cross the capillary barrier rapidly. Levels of caspase-1-like activity did not increase following transfection implying that L-PEI/DNA complexes are transported across cellular barriers by a non-damaging, physiological process, without causing inflammation. The high levels of expression of different transgenes in pneumocytes indicates that transport of L-PEI/DNA complexes through the endothelial barrier does not affect their transfection capacity. These findings open up new possibilities for gene delivery and its application to the lung.


The systemic route could provide a relatively non-invasive means of gene delivery to the lung for a number of prophylactic and therapeutic situations. For instance, expression in pneumocytes of a dominant FHIT (fragile histidine triad) gene has been suggested for prophylaxis of certain forms of lung cancer123 and intravenous delivery of therapeutic genes is a realistic alternative to aerosol or intratracheal injections for cystic fibrosis (CF) therapy. Indeed, in CF patients with established lung disease the lung epithelium can become inaccessible for vector–DNA complexes because of thick mucus accumulation, rendering the intravenous route more attractive, on condition that genes reach the appropriate cellular sites.4

A series of recent studies have shown the lung capillaries to be a privileged site for transgene expression following intravenous gene delivery with various nonviral vectors.567 We too have shown that a synthetic cationic polymer, 22 kDa linear polyethylenimine (L-PEI or ExGen 500; Euromedex, Souffleweyersheim, France), provides particularly high levels of transgene expression in the mouse lung.8 However, with L-PEI, the detection of β-galactosidase activity in lung sections seemed to indicate that the transgene was expressed in pneumocytes beyond the endothelium barrier. This observation, which required confirmation by the use of cell markers to identify unambiguously the cell types, raised a key question. If the genes were reaching cells beyond the endothelium, how is this cell barrier crossed? Two main hypotheses can be proposed: either rupture of the intracellular junctions or a cell permeation mechanism such as transcytosis.9

Here we set out, first, to confirm that passage of the endothelium barrier does occur. To this end we used double immunocytochemistry to identify the cell types and to co-localize them with digoxin-labeled DNA or transgene expression. Next, we analyzed the kinetics and the cellular basis of endothelium crossing.

In a first series of experiments we assayed luciferase expression in lungs and livers of animals injected with either L-PEI/DNA complexes or naked DNA at time-points between 2 h and 72 h. As shown in Figure 1a and b, luciferase expression was detected in both liver and lung within 2 h of injection at two to three levels of magnitude higher than in uninjected animals. In the lung, expression at 2 h was over 105 RLU/mg protein and then peaked close to 107 RLU/mg protein at 6 and 12 h post-injection (p.i.), two orders of magnitude higher than in liver at the same time-points. In contrast, in control animals that had received naked DNA, expression levels did not exceed background values in the liver, and only reached 103 RLU/mg protein in the lung between 2 and 12 h p.i. Thus, L-PEI vectorization strongly favored transfection of lung tissue in both the short and longer term.

Figure 1

Time-course of transgene expression in lung and liver after intravenous injection. Luciferase expression in extracts of lung (a) and liver (b) of mice injected with 125 μg pCMV-luciferase either alone (naked DNA, black triangles) or complexed with with 1.5 μl of 1 M 22 kDa linear PEI in 500 μl 5% glucose at an amine/phosphate ratio of 4 (black circles). Means ± s.e.m. are given, n = 6 in all cases. Endotoxin-free DNA was prepared using affinity columns according to the manufacturer's instructions (Genomed, Research Triangle Park, NC, USA). DNA was washed with 70% ethanol, resuspended in TE and stored at 4°C. Vasodilatation was induced by warming mice tails in water (45°C) before injection. The dissected organs were homogenized in 500 μl of luciferase assay buffer, centrifuged for 10 min at 4°C, and 20 μl of each extract mixed with 100 μl of Luciferase assay substrate (Promega, Madison, WI, USA). Activity was estimated for 10 s, and normalized against protein content. Proteins were assayed by the Bradford method.

The remarkably rapid and efficient transgene expression in the lung raised the question of the cellular localization of the DNA at the earliest time-point and whether DNA localization could be correlated with gene expression. The cellular localization of vectorized DNA was assessed by injecting digoxin-labeled DNA and then examining sections from injected animals with an antibody against digoxin and using specific markers for different cell types. Antibodies against factor VIII (Dako, Glostrup, Denmark) and surfactant protein A4 were used to identify endothelial and pulmonary cells, respectively. As shown in Figure 2a–d, at 2 h p.i. labeled complexes were found in the lumen of small capillaries (Figure 2a), within the endothelial cell layer (arrowheads in Figure 2b, c), in pulmonary cells (arrow in Figure 2b) and in cells located in close proximity of blood vessels (arrows in Figure 2c). This latter localization correlated with β-galctosidase expression already seen at 2 h p.i. in groups of pulmonary cells adjacent to blood vessels (Figure 2d).

Figure 2

DNA distribution at 2 h and 24 h after injection. The distribution of digoxin-labeled DNA was revealed 2 h (a–d) and 24 h (e–h) after intravenous injection by an alkaline phosphatase-coupled anti-digoxigenin antibody (dark violet colour in a–c, e–f). Sections were immunostained either with an anti-factor VIII antibody (DAKO, Glostrup, Denmark) (a, b, f) to identify endothelial cells or with an anti-surfactant protein A antibody4 (g, h) for respiratory epithelial cells (1:200 dilution in PBS, with 5% fetal calf serum) and revealed with a peroxidase ABC kit (DAKO); in all cases the immunostaining product appears dark brown. Panel d shows the expression of a LacZ reporter 2 h after injection. At 2 h p.i. large (a) aggregates containing DNA are occasionally seen adhering to the wall of major blood vessels (arrow). DNA can be detected in endothelial cells of blood vessels (arrowheads in b, c) and in the alveolar cells in close proximity to the vessels (arrows in b and c). Two hours after injection with a LacZ reporter construct, some alveolar cells already express β-galactosidase (d). At 24 h p.i. DNA was no longer found in the lumen of blood vessels but could be detected between blood vessels and bronchioli (e; arrows in f, g) and in groups of cells in the alveolar region (arrows in h). DNA distribution at 24 h is recapitulated by the expression of LacZ observed at the same time after injection (see Figure 3). Digoxin labeling was carried out according to the manufacturer's instructions (Panvera, Madison, WI, USA). Briefly, 50 μg of DNA was mixed with the labeling buffer, dye added and incubated at 37°C for 1 h. Ethanol/NaCl precipitation was carried to separate labeled from non-labeled DNA. Labeling was verified by dot blotting, and did not modify the L-PEI-based transfection activity of DNA (data not shown). Digoxin-labeled DNA was injected in a 1:9 (w/w) mix with unlabeled DNA and was revealed with an alkaline-phosphatase-coupled anti-digoxigenin antibody followed by NBT/BCIP solution according to the manufacturer's instructions (Boehringer-Roche, Mannheim, Germany). In sections double stained with a lung marker (either factor VIII or surfactant protein A) the reaction with the anti-digoxigenin antibody (Amersham, Les Ulis, France) was performed after completion of the peroxidase reaction. Anti-digoxigenin antibody fully cross-reacts with digoxin. as, alveolar space; bv, blood vessel; br, bronchioli. Bar: 40 μm in a, b, e–h; 60 μm in d; 200 μm in c. Representative sections are shown. The experiment was carried out three times with a minimum of four animals per group (total of 12 lungs examined).

At 24 h after intravenous injection digoxin-labeled DNA was found in the space between blood vessels and bronchioli and in endothelial and respiratory cells (Figure 2e). The identity of endothelial (Figure 2f) and pulmonary cells (Figure 2g, h) was confirmed by immunocytochemistry with appropriate markers. DNA localization was well correlated with sites of transgene expression. Indeed, when digoxin-labeled DNA was coinjected with a CMV-LacZ reporter construct, β-galactosidase activity was detected in the same sites where labeled DNA was present. In particular, reporter gene expression was present in large groups of cells located between major blood vessels and bronchioles (Figure 3a). High power examination of the same sections immunostained with the appropriate markers (Figure 3b, c) revealed that β-galactosidase expression was present both in endothelial cells and in pneumocytes.

Figure 3

LacZ expression in the lung 24 h after intravenous injection. Mice were injected with a plasmid (125 μg) containing the coding sequence of β-galactosidase (LacZ) under the CMV promoter (Vical, San Diego, CA, USA). Mice were killed 2 h and 24 h after injection, sequentially perfused with 5 mg/l heparin in saline, 2% paraformaldehyde, saline and a standard β-galactosidase revealing solution (0.8 mg/ml X-gal, Boehringher-Roche). After perfusion, the lungs were dissected and the β-galactosidase reaction continued for 48 h by immersion in the X-gal solution at 30°C. After staining, the lungs were embedded in paraffin and sectioned (5 μm). The sections were subsequently immuno-stained either with anti-factor VIII (a, b) or with anti-surfactant protein A (c) antibodies as in Figure 2; in all cases nuclei were identified by nuclear fast red counterstaining. β-Galactosidase activity is prominent in groups of cells located in close proximity to major blood vessels (arrowheads in a). At higher magnification it is clear that both respiratory endothelial cells (immunostained by the anti-factor VIII antibody; arrows in b) and type II pneumocytes (immunostained by the anti-surfactant protein A antibody; arrows in c) express the reporter gene. bv, blood vessel; br, bronchioli. Bar 100 μm in a, 25 μm in b, c. Methods as in Figure 2. Representative sections are shown. The experiment was carried out three times with a minimum of four animals per group.

These observations raised the question of the cellular mechanism underlying the rapid translocation of complexes through the endothelium. One possibility could be rupture of the endothelium cell junctions. Such an event would most likely be followed by necrosis and inflammation. We measured a biochemical indicator of inflammation: caspase 1-like activity.1011 No increased caspase 1-like activity was detected in lungs at any time-point p.i. irrespective of whether animals were injected with L-PEI and a luciferase expressing plasmid (Figure 4a) or with L-PEI and an empty plasmid (Figure 4b). Similarly, injection of L-PEI alone, uncomplexed to DNA, did not cause capsase 1 activation at any time-point, in fact levels in treated animals were lower than in controls (Figure 4c).

Figure 4

L-PEI/DNA injection does not induce inflammation. Caspase-1-like activity in lung extracts of mice injected with either 125 μg pCMV-luciferase (a), pCMV with no coding sequence (b) or L-PEI alone (c). In a and b, DNA was complexed with 22 kDa L-PEI in 500 μl 5% glucose at an amine/phosphate ratio of 4. Organs were dissected and frozen in liquid nitrogen. Caspase activities were measured using a commercially available kit (Promega). Succinctly, 100 μl tissue extracts containing 100 μg protein were mixed with the appropriate caspase 3- or caspase 1-like substrates (Ac-DEVD-MCA and Ac-YVAD-MCA, respectively) or inhibitors (Ac-DEVD-CHO and Ac-YVAD-CHO, respectively) at a final concentration of 0.05 mM. Upon cleavage by the caspase, MCA (Methyl Coumarin Amide) is liberated and quantified against an MCA standard on a microplate reader (Wallac Oy, Turku, Finland). Fluorescence is calculated as enzymatic activity (picomoles of MCA liberated/μg protein/min). Protein content was measured using a Bradford assay. Means ± s.e.m. are given, n = 6 in all cases.

A final point we addressed was the cause of the rapid loss of expression between 24 and 48 h. We used semiquantitative PCR to assay luciferase plasmid present in lung extracts at given time points p.i. As shown in Figure 5, pCMV-luc expression in lung at 48 h p.i. is less intense than at 2 and 6 h p.i. Quantification by densitometry and normalization against those for endogenous β-actin amplified from the same samples, showed the ratio of luc/actin to be less than 0.2 at 48 h as compared with ratios of 0.4 to 0.75 at earlier time-points.

Figure 5

Plasmid degradation is seen at 48 h after injection. Animals were killed at the time-points indicated, tissues removed and homogenized. Total DNA was extracted from a whole lung; after ethanol precipitation, PCR was performed on 1/20th of the total extract, with primers corresponding to the luciferase cDNA (5′ ACTGCATAAGGCTATGAAGAGA 3′ and 5′ CTTGTAATCCTGAAGGCTCCTC 3′). Twenty cycles of amplification were carried out at 94°C for 1 min, 60°C for 1 min and 72°C for 1 min. Products of amplification were loaded on gel electrophoresis and stained with ethidium bromide. The expected size fragments (768 base pairs) were quantified, using the NCSA GELREADER 2.0.5 program (University of Illinois, Urbana-Champaign, IL, USA). The values obtained for the luciferase insert were normalized on those obtained for β-actin.25 Each value is the average of three separate experiments. Means ± s.e.m. are given.

To determine whether plasmid loss was associated with apoptosis, we measured caspase 3 (CPP32)-like activity in tissues. Even though caspase 3-like activity in injected lungs was increased slightly (two-fold), and significantly, at 12 and 48 h p.i. but not at 2, 6 and 24 h, the values observed were always an order of magnitude less than those in the liver at the same time (Figure 6). These low levels of caspase 3-like activity in the L-PEI-transfected lung suggest that plasmid loss is unlikely to be due to apoptosis. Indeed, we co-transfected CMV-luc with a plasmid expressing the anti-apoptotic gene bclXL1213 and found no increase in duration of reporter gene expression in the lung (data not shown).

Figure 6

Injection of L-PEI/DNA complexes cause only slight activation of caspase-3-like activity in the lung. Mice were injected with 125 μg pCMV-luciferase complexed with 22 kDa L-PEI. Non-injected animals (NT) were used as controls. Subsequent steps were carried as in Figure 4. Means ± s.e.m. are given, n = 6 in all cases. **P < 0.01; ***P < 0.001.

In this study, we analyzed the kinetics and process of L-PEI-based gene delivery to the mouse lung. The findings bolster our postulate that L-PEI/DNA complexes can pass the capillary barrier in the lung, so as to reach and transfect other pulmonary cells types. Three series of observations corroborate the hypothesis that transcytosis of complexes through the lung endothelium is occurring: first, the localization and identity of the cells expressing the transgenes, second, the rapidity of transfer and gene expression, and third, the lack of inflammatory activity.

Our primary argument is that transgene expression is found in pulmonary cells identified unambiguously by specific cell markers. Indeed, we used double immunocytochemistry and histochemistry to localize both digoxin-labeled DNA and transgene expression with respect to cellular markers identifying either endothelial cells (anti-factor VIII) or surfactant-expressing pulmonary cells (anti-surfactant protein A). Various combinations of markers showed that plasmid DNA vectorized with L-PEI is found in the alveolar cells as early as 2 h p.i. and in these same cells as well as pulmonary cells at the base of the bronchioles at 24 h p.i. Localization of digoxin-labeled DNA in these cell types was correlated with transgene expression. Transgene expression was found in endothelial cells, but this was by no means the major site of expression.

The passage of the endothelium is extremely rapid, as one would expect for a transcytotic mechanism; labeled DNA and transgene expression being found in the alveolar cells as early as 2 h p.i. Both luciferase activity and β-gal transgenes showed significant levels of activity at the earliest time-point examined. This means that within 2 h, not only have the complexes moved through the endothelial cells but that they have re-entered the pulmonary cells, undergone trafficking to the nucleus and produced functional protein.

Our final argument in favor of a transcytotic mechanism is the lack of inflammation resulting from L-PEI/DNA gene transfer. If transport is not occurring by a cell based-transfer mechanism, then passage across the endothelium would involve rupture of the tight intercellular junctions. Indeed, one could suppose that injection of a large volume (up to 20% of blood volume) such as used here, could cause a surge in blood pressure and rupture of the endothelium cell junctions. This would be followed by cell damage, necrotic cell death and inflammation associated with a caspase 1 increase. Previous histological evidence8 has suggested the innocuous nature of this vectorization method, and tends to rule out the rupture theory. These biochemical results showing no increase in caspase 1-like activity in the lung reinforce the idea that passage occurs through a physiological transport process, such as transcytosis.

Transgene expression occurring in the pulmonary cells appears to be favored over expression in the endothelial cells themselves. This seems to be a hallmark characteristic of gene delivery to the lung with L-PEI, and is not seen, at least not at the high level reported here, with other cationic vectors. Indeed, although some authors have reported the presence of either plasmid4 or limited gene expression in pulmonary cells14 following delivery with cationic lipids, the majority of the signal is found in endothelial cells.

Two key factors can be proposed to explain the particularly high levels of transgene expression in pulmonary cells following transfection with L-PEI: the method used for formulation of complexes and the inherent buffering capacity of L-PEI. We formulated DNA/L-PEI complexes in a non-ionic solution (5% glucose) that results in the formation of stable particles of 30–60 nm diameter.15 These particles seem to be remarkably stable in physiological fluids.15 The endosomal buffering capacity of PEI,161718 could well explain the fact that PEI/DNA complexes can survive the passage through the endosome/transcytotic pathway. In PEI, one in every third atom is an amino nitrogen that can be protonated, making PEI the cationic polymer having the highest charge density potential known. The protonation level of PEI increases from 20 to 45% between pH 7 and 5,19 thus increasing its buffering capacity as endosomal pH decreases. Another enhancing factor is probably the high level of transcytosis that occurs in the lung.20 Moreover, although, this transcytotic delivery of PEI/complexes across the capillary barrier appears particularly favorable in the lung, the fact that PEI is an excellent vector for conjugating cell-specific ligands21222324 means the approach could be adapted to other target tissues, for both prophylactic and curative procedures.

A final point is that plasmid loss in the lung has no obvious correlation with increased apoptosis. Even in the liver where caspase 3-like activity was increased, levels were only one order of magnitude over controls. A possible explanation of the plasmid loss in the lungs is that a mitotic population of cells is being transfected, and that during mitosis plasmid is diluted and degraded.

In conclusion, the data presented here underline both the high transfection capacity of L-PEI /DNA complexes and the lack of toxicity associated with their use. Taken together, the results strengthen arguments favoring the use of this vector as a method of choice for efficient gene delivery in vivo.


  1. 1

    Croce CM, Sozzi G, Huebner K . Role of FHIT in human cancer J Clin Oncol 1999 17: 1618–1624

  2. 2

    Sozzi GS et al. Absence of Fhit protein in primary lung tumors and cell lines with FHIT gene abnormalities Cancer Res 1997 57: 5207–5212

  3. 3

    Sozzi G et al. Loss of FHIT functions in lung cancer and preinvasive bronchial lesions Cancer Res 1998 58: 5032–5037

  4. 4

    Griesenbach U et al. Comparison between intrathracheal and intravenous administration of liposome–DNA complexes for cystic fibrosis lung gene therapy Gene Therapy 1998 5: 181–188

  5. 5

    Li S, Huang L . In vivo gene transfer via intravenous administration of cationic lipid-protamine–DNA (LPD) complexes Gene Therapy 1997 4: 891–900

  6. 6

    McLean JW et al. Organ-specific endothelial cell uptake of cationic liposome^DNA complexes in mice Am J Physiol 1997 273: H387–H404

  7. 7

    Rodman DM et al. In vivo gene delivery to the pulmonary circulation in rats: transgene distribution and vascular inflammatory response Am J Respir Cell Mol Biol 1997 16: 640–649

  8. 8

    Goula D et al. Polyethylenimine-based intravenous delivery of transgenes to mouse lung Gene Therapy 1998 5: 1291–1295

  9. 9

    Renkin EM . Cellular and intercellular transport pathways in exchange vessels Am Rev Respir Dis 1992 146: S28–S31

  10. 10

    William R et al. The IL-1 beta-converting enzyme (caspase-1) inhibits apoptosis of inflammatory neutrophils through activation of IL-1 beta J Immunol 1998 161: 957–962

  11. 11

    Wong WW . ICE family proteases in inflammation and apoptosis Agents Actions Suppl 1998 49: 5–13

  12. 12

    Chao DT, Korsmeyer SJ . BCL-2 family: regulators of cell death Annu Rev Immunol 1998 16: 395–419

  13. 13

    Dietrich JB . Apoptosis and anti-apoptosis genes in the Bcl-2 family Arch Physiol Biochem 1997 105: 125–135

  14. 14

    Liu F, Qi H, Huang L, Liu D . Factors controlling the efficiency of cationic lipid-mediated transfection in vivo via intravenous administration Gene Therapy 1997 4: 517–523

  15. 15

    Goula D et al. Size, diffusibility and transfection performance of linear PEI/DNA complexes in the mouse central nervous system Gene Therapy 1998 5: 712–717

  16. 16

    Boussif O et al. A novel, versatile vector for gene and oligonucleotide transfer into cells in culture and in vivo: polyethylenimine Proc Natl Acad Sci USA 1995 92: 7297–7303

  17. 17

    Behr JP . Gene transfer with synthetic cationic amphiphiles: prospects for gene therapy Bioconj Chem 1994 5: 382–389

  18. 18

    Zou SM, Behr JP, Goula D, Demeneix BA . Gene delivery with polyethylenimine. In: Templeton NS, Lasic DD (eds). Gene Therapy: Therapeutic Mechanisms and Strategies Marcel Dekker: NY 1999 (in press)

  19. 19

    Suh J, Paik H-J, Hwang BK . Ionization of polyethylenimine and polyallylamine at various pHs Bioorg Chem 1994 22: 318–327

  20. 20

    Schnitzer JE, Oh P, Jacobson BS, Dvorak AM . Caveolae from luminal plasmalemma of rat lung endothelium: microdomains enriched in caveolin, Ca(2+)-ATPase, and inositol trisphosphate receptor Proc Natl Acad Sci USA 1995 92: 1759–1763

  21. 21

    Zanta MA, Boussif O, Adib A, Behr JP . In vitro gene delivery to hepatocytes with galactosylated polyethylenimine Bioconjug Chem 1997 8: 839–844

  22. 22

    Coll JL et al. In vitro targeting and specific transfection of human neuroblastoma cells by chCE7 antibody-mediated gene transfer Gene Therapy 1997 4: 156–161

  23. 23

    Kircheis R et al. Coupling of cell-binding ligands to polyethylenimine for targeted gene delivery Gene Therapy 1997 4: 409–418

  24. 24

    Erbacher P, Remy JS, Behr JP . Gene transfer with synthetic virus-like particles via the integrin-mediated endocytosis pathway Gene Therapy 1999 6: 138–145

  25. 25

    Guissouma H et al. Physiological regulation of hypothalamic TRH transcription in vivo is T3 receptor isoform specific FASEB J 1998 12: 1755–1764

Download references


We are grateful to the Association Française pour la Lutte contre la Mucoviscidose (AFLM) the Association Française contre les Myopathies (AFM) and ARC for support. GL was supported by ARSEP, AISM and Consiglio Nazionale delle Ricerche (Progetto Finalizzato ‘Biotecnologie’). The support from Telethon (Italy) for the project: ‘Use of transgenic mutant mice as a model to study the molecular control of bone development and peripheral myelination and to develop new gene therapy strategies in the embryo’ (Project D76) is gratefully acknowledged. Dr M Post (Toronto) kindly provided the anti-surfactant antibody. Daniel Goula is a recipient of a graduate fellowship from the French government (MRE).

Author information

Correspondence to B A Demeneix.

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Goula, D., Becker, N., Lemkine, G. et al. Rapid crossing of the pulmonary endothelial barrier by polyethylenimine/DNA complexes. Gene Ther 7, 499–504 (2000). https://doi.org/10.1038/sj.gt.3301113

Download citation


  • apoptosis
  • cationic polymers
  • lung
  • in vivo gene transfer
  • synthetic vectors

Further reading