We have assessed the transfection efficiency and efficacy of small (approximately 20 bp) nucleic acids (siRNA and antisense DNA) complexed to GL67 in conducting airway epithelium in vivo. In general, uptake of small oligonucleotides was inefficient and although mRNA levels were reduced, we could not reduce protein levels or ENaC function. This study suggests that although siRNAs and asODNs can be developed to inhibit gene expression in culture systems and certain organs in vivo, barriers to nucleic acid transfer in airway epithelial cells seen with large DNA molecules also affect the efficiency of in vivo uptake of small nucleic acid molecules.
It is unlikely that alteration in nucleic acid size alone will result in improved airway nucleic acid transfer. We suggest, that the asODN and siRNA-based strategies may not be successful in conducting airway epithelium, until nucleic acid transfer is optimised further.
Interestingly, the intracellular distribution of asODN and siRNA following in vitro
transfection into epithelial cells was very different, with asODN-lipoplexes accumulating rapidly in the nucleus while siRNA-lipoplexes were seen as a diffuse staining in the cytoplasm. For asODN this distribution was independent of nucleotide sequence, and has also recently been seen in other cell lines, such as A549 and HEK293 cells (Chris Kitson, GlaxoSmithKline, personal communication). To the best of our knowledge, we show, for the first time, that the respective localisations in vitro
are consistent with the presumed sites of action of each of the molecules in inhibiting gene expression [23
Various cationic lipids and polymers have been assessed for gene transfer to the airways, but information on a "best buy" non-viral transfer agents for small nucleic acid delivery does not currently exist. In our hands, GL67 has been most efficient for airway gene transfer and we have assessed its safety in phase 1 trials in normal volunteers and CF patients [20
]. Here, we assessed the efficacy for GL67 to delivery asODN and siRNA to the airways. In vivo
, siRNA-lipoplexes were found predominantly associated with alveolar macrophages, whereas asODNs were, as has been reported previously, associated with pneumocytes. The reasons for this difference in cellular distribution is currently unclear, but may relate to differences in RNA versus DNA-lipoplex properties. Small differences in overall complex structure may for example alter the surface charge slightly, which in turn may affect cell uptake. Zhang and co-workers recently described administration of "naked" biotin-labelled siRNA to the lung and reported diffuse staining in airways and parenchyma [24
]. We were unable to detect any signal after administration of "naked" FITC-labelled siRNA in the lung (data not shown). This may in part reflect different detection limits of these methods (FITC versus biotin-streptavidin) in the lung, possibly due to high auto-fluorescence of lung tissue. We speculate that the differences in cellular distribution of asODN in vitro
(mainly nuclear) and in vivo
(mainly cytoplasmic) may be due to differences in the proliferation status of the transfected cells. M1 cells replicate rapidly, whereas the majority of pneumocytes are terminally differentiated non-dividing cells. The nuclear membrane is likely to present a significant barrier to ODN uptake in vivo
The inefficient and variable degree of transfection of the airway epithelium was also reflected in the experiments testing inhibition of gene expression. In the airway epithelium of K18-lacZ transgenic mice the administration of lacZ asODN significantly reduced βgal mRNA 72 and 96, but not 48 hours, after transfection. In contrast, following lacZ siRNA transfection lacZ mRNA levels were significantly reduced 48 hours after transfection. The delayed response of asODNSs, may be due to the different sites and modes of action. In contrast to mRNA there was no reduction in βgal protein at any time point tested. There are a number of possible explanations. The predicted half-life of βgal protein is 20 hours. Harvesting lungs 48 hours after transfection should, therefore, have allowed a reduction in protein expression to be seen. However, the half-life of βgal protein in vivo may be longer than 20 hours, reducing the likelihood of detecting a significant change in protein levels during the time course of these experiments. In addition, we cannot exclude a non-linear response between protein amounts and enzymatic activity, which may require a substantial reduction in protein before any changes in enzymatic activity can be seen. Finally, even with the relatively large numbers studied, we were only powered to detect a protein reduction of 60% or more.
Recently several studies described the successful use of siRNA to inhibit pulmonary influenza virus and respiratory syncytial virus infections [25
] using intravenous (IV) or intranasal (IN) administration of plasmid DNA encoding the siRNA or IV plus IN administration of siRNA. These studies did not specifically assess the effects of RNA interference in difficult to transfect conducting airway epithelial cells, but looked at global anti-viral responses in the lung. To the best of our knowledge efficacy of RNAi in airway epithelium has not been studied yet.
We tried to improve transfection efficiency by prolonging contact time with the airway epithelial cells. For practical reasons this necessitated a change to the nose. Despite this, we could not detect any intracellular fluorescent signal after administration of FITC-labelled siRNA/GL67 complexes and only a low and highly variable signal of asODN/GL67 complexes. Importantly, the overall transfection efficiency of nasal and lung airway epithelial cells was identical, despite anatomical differences, lending credit to using the nasal epithelium as a surrogate for airway nucleic acid transfer.
Despite using antisense ODNs that reduced αENaC mRNA by up to 60% in vitro
, results consistent with other gene targets in our hands, these did not reduce ENaC mRNA or protein function in murine nasal epithelium in vivo
. The reported half-life of the ENaC protein varies and appears to be cell-type specific, but is of the order of 40–120 min in cultured cells [28
] and 3–4 hours in Xenopus oocytes [29
]. We know of no data for the murine airways. Thus, although the stability of the protein may play a role, we suggest that poor transfection is more likely the reason for the lack of effect.
The efficiency of siRNA and antisense inhibition is sequence dependent and though some attempts have been made to develop computer-assisted design of these types of molecules extensive biological testing is still required. The LacZ siRNAs described in this study were developed using first generation design guidelines that focused on the overall nucleotide composition and distribution (50:50 AU:GC ratio, with approximately equal distribution of pyrimidines and purines throughout the molecule). These siRNAs were analyzed prior to the development of second-generation guidelines that focus on more mechanistically relevant features of siRNAs, particularly a nucleotide composition that favours asymmetrically loading of siRNAs into the ribonucleoprotein complex that mediates RNAi. Using the first generation guidelines 3 out of 10 siRNAs mediated a significant degree of silencing. In our experience this proportion of effective siRNAs and asODNs reflected the success rate seen for the design of siRNAs against most transcripts using first generation guidelines.
A key development in improving small nucleic acid delivery in vivo
, particularly of asODNs, has been modification of the molecules or their delivery agent to improve their stability in biological fluids. Neither asODN nor siRNA stability has been studied in the context of the extracellular lung environment in humans, though Templin et al
nebulised asODN into the murine lung and reported a half-life of >20 hours [30
]. The airway surface liquid (ASL) lining the conducting airways is important for host defence and contains numerous anti-microbial peptides [31
]. Collection of unperturbed ASL in human lung is impossible and broncho-alveolar lavages (using comparatively large volumes) lead to dilution of ASL components. Although exhaled breath condensate (EBC), which can be collected non-invasively and consists of condensed water and microdroplets, containing volatile and non-volatile compounds representing ASL from the lower respiratory tract [32
], may not mimic all aspects of ASL, it goes some way towards assessing the stability of small nucleic acid in the human lung. Importantly, stability of asODNs and siRNAs in EBC from inflamed lungs of CF patients was similar to non-CF patients. These data support previously published stability data for second-generation asODN, [where one of the non-bridging oxygen atoms in the phosphodiester bond is replaced with sulphur (phosporothioation)], and the addition of 2' O-methylated RNA residues increase nuclease resistance, and also support the apparent nuclease stability of siRNA even in the absence of chemical modification. We did not assess stability of GL67/siRNA or asODN complexes, because it has previously been shown that lipid complexing further increases nuclease stability of oligonucleotides [34