DNA Internalized via Caveolae Requires Microtubule-dependent, Rab7-independent Transport to the Late Endocytic Pathway for Delivery to the Nucleus*

Using cationic liposomes to mediate gene delivery by transfection has the advantages of improved safety and simplicity of use over viral gene therapy. Understanding the mechanism by which cationic liposome:DNA complexes are internalized and delivered to the nucleus should help identify which transport steps might be manipulated in order to improve transfection efficiencies. We therefore examined the endocytosis and trafficking of two cationic liposomes, DMRIE-C and Lipofectamine LTX, in CHO cells. We found that DMRIE-C-transfected DNA is internalized via caveolae, while LTX-transfected DNA is internalized by clathrin-mediated endocytosis, with both pathways converging at the late endosome or lysosome. Inhibition of microtubule-dependent transport with nocodazole revealed that DMRIE-C:DNA complexes cannot enter the cytosol directly from caveosomes. Lysosomal degradation of transfected DNA has been proposed to be a major reason for poor transfection efficiency. However, in our system dominant negatives of both Rab7 and its effector RILP inhibited late endosome to lysosome transport of DNA complexes and LDL, but did not affect DNA delivery to the nucleus. This suggests that DNA is able to escape from late endosomes without traversing lysosomes and that caveosome to late endosome transport does not require Rab7 function. Lysosomal inhibition with chloroquine likewise had no effect on transfection product titers. These data suggest that DMRIE-C and LTX transfection complexes are endocytosed by separate pathways that converge at the late endosome or lysosome, but that blocking lysosomal traffic does not improve transfection product yields, identifying late endosome/lysosome to nuclear delivery as a step for future study.

Gene therapy strategies are categorized into two areas: nonviral and viral. Non-viral gene therapies, which include the use of cationic liposomes or polymers, have the advantages of greater safety and ease of use compared with viral gene therapies. Cationic liposomes are amphiphilic molecules comprised of a hydrophobic chain and a polar headgroup whose positive charges associate with DNA through electrostatic interactions.
Cationic polymers, also known as polyplexes, such as polyethylenimine (PEI), 3 are more positively charged and lack a hydrophobic chain. The relatively low efficiency of both cationic liposome-and polymer-mediated gene delivery limits their utilization in the clinic (1)(2)(3). Understanding how DNA enters the cell when complexed with either of these classes of reagents may enable more successful strategies to improve transfection efficiencies.
There are four main potential barriers to successful DNA transfection, or more precisely the translocation of exogenous DNA from outside the cell into the nucleus. First the transfected DNA must be able to associate with the cell surface and be internalized into the cell. Next the DNA must not only successfully navigate the endosomal pathway(s), but also evade lysosomal degradation, which has been demonstrated to inhibit DNA transfection (4,5). Should the DNA escape from the endosomal pathway, it must also avoid degradation by nucleases in the cytoplasm. Finally, one of the major obstacles for gene delivery is nuclear entry. As transfected DNA is too large to be passively transported, the DNA must either be actively transported through nuclear pores or diffuse into the nucleus while the nuclear membrane is disassembled during cell division. Active nuclear import of DNA requires the presence of a nuclear localization signal (NLS) or association of DNA with proteins that shuttle from the cytoplasm to the nucleus (6). Some studies report increased transfection efficiency with plasmids that contain binding sequences for nuclear proteins or noncovalently linked NLS-containing peptides (7,8). These potential barriers to gene delivery may not apply to all forms of non-viral gene therapy, as multiple mechanisms of internalization have been suggested (1,3).
Whereas some reports suggest that non-viral gene delivery utilizes a non-endocytic mechanism (9,10), the majority of reports suggest that endocytosis is in fact the preferred route of cell entry (1,3,(11)(12)(13). Several forms of endocytosis have been demonstrated to be involved in DNA uptake: clathrin-mediated, caveolae-mediated, and macropinocytosis. Clathrin-mediated endocytosis, which entails the formation of clathrincoated pits that pinch off to form intracellular vesicles that subsequently uncoat and fuse with early endosomes (EE), is the * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. □ S The on-line version of this article (available at http://www.jbc.org) contains supplemental Figs. S1-S4. 1 To whom correspondence should be addressed. Tel.: 650-467-4767; E-mail: wong.athena@gene.com. 2 Both authors contributed equally to this work and should be considered co-senior authors. 3 The abbreviations used are: PEI, polyethylenimine; NLS, nuclear localization signal; EE, early endosomes; LE, late endosomes; MAb, monoclonal antibody; WT, wild type; GFP, green fluorescent protein; DN, dominant negative; PBS, phosphate-buffered saline; BSA, bovine serum albumin; DAPI, 4Ј,6-diamidino-2-phenylindole; ER, endoplasmic reticulum; FACS, fluorescent-activated cell sorting.
best characterized and has been shown to be responsible for various types of non-viral gene delivery such as SAINT-2/ DOPE-and DOTAP-mediated transfections (11,12). Fusion with EE is regulated by the EE-localized small GTPase Rab5 (39,40), and subsequent trafficking to late endosomes (LE) and lysosomes is similarly regulated by Rab7, which resides on LE (14,15)). Caveolae are cholesterol-rich domains within the plasma membrane that contain the caveolin protein and have been shown to internalize various pathogens (16 -19) in addition to basic peptides (20) and DNA complexed with PEI (12,21). Caveolae pinch off to form pH neutral, non-degradative endocytic organelles known as caveosomes (22). The caveolar pathway has been proposed to be advantageous over the clathrin-mediated pathway for transfection of DNA due to avoidance of lysosomal degradation (12,23). Macropinocytosis occurs when actin-driven invaginations of the membrane pinch off to form large endocytic vesicles (24). It has been demonstrated to be important for transfection of large, but not small, octa-arginine peptide:DNA complexes (25), as well as naked DNA in keratinocytes (26). Thus the mechanism of DNA internalization varies widely according to cell type, transfection reagent, and DNA complex size.
In this study we focused on DP12 cells, a derivative of the CHO-K1 cell line, because DP12 cells are used to manufacture large quantities of therapeutic proteins for clinical use (27). Because lysosomal degradation of transfected DNA has been proposed as a major reason for low transfection efficiencies (4,5), we aimed to first determine if the DNA trafficked to lysosomes, and, if so, whether inhibiting its transport to that compartment would improve transfection product yields. We compared gene delivery by two different transfection reagents, DMRIE-C and Lipofectamine LTX and found that they internalize via caveolar-and clathrin-mediated endocytosis, respectively. Additionally, we show for the first time that transfected DMRIE-C:DNA complexes (henceforth abbreviated to DMRIE-C:DNA) traffic to the late endosomes (LE) or lysosomes following caveolae-mediated internalization in a step that requires microtubules, suggesting that DNA may not escape lysosomal degradation simply by being internalized by caveolae, as previously proposed (12). Furthermore, unlike caveosome to EE transport requiring the activity of the early endosomal small GTPase Rab5 (28), cavesome to LE transport does not appear to similarly require the action of the late endosomal Rab7 GTPase. Inhibitors of lysosomal trafficking and function do not affect DNA transfection titers, suggesting that the DNA is able to escape to the cytosol directly from LE without obligatorily traversing lysosomes for subsequent transport to the nucleus.
Cell Culture and Transfection-The CHO-DP12 cell line is a derivative of the CHO-K1 cell line (ATCC number CCL-61) and has a reduced requirement for insulin. Cells were grown in Dulbecco's modified Eagle's medium Ham's F12-based media supplemented with low IgG serum.
For all transfections, CHO cells were seeded at 1.5 ϫ 10 6 cells/well in 2 ml of complete media on the day of transfection in non-tissue culture-treated 6-well plates. For each well of a DMRIE-C transfection, 4 g of DNA and 4 l of DMRIE-C (Invitrogen, catalog number 10459-014) were incubated in serum-free media for 30 min before being added dropwise to cells. For each well of a Lipofectamine LTX transfection, 2.5 g of DNA was combined with 1.25 l of Plus reagent (Invitrogen, catalog number 11514-015) in 500 l of serum-free media for 5 min. 5 l of Lipofectamine LTX (Invitrogen, catalog number 15338-100) was then added, mixed, and incubated for an additional 25 min before being added dropwise to cells. Cells were incubated with agitation at 37°C, 5% CO 2 for up to 24 h. For assays in which titers were collected at 48 h, the temperature was shifted to 33°C after the first 24 h of transfection.
In Fig. 2B, cells were pretreated with endocytic inhibitors and transfected as above for 6 h then washed, replated in fresh media, and returned to the incubator for 24 h at 37°C followed by another 24 h at 33°C, 5% CO 2 with agitation. 48 h supernatants were collected from the cultures for ELISA analysis and normalized to untreated cells as 100%. In Fig.  1C, cells were instead harvested after 24 h chase and assayed for viability using the ViaCount assay and a GUAVA personal cell analyzer (PCA, Guava Technologies) according to the manufacturer's instructions.
Immunofluorescence-Immunofluorescence staining was performed based on methods previously described (33). CHO cells were transfected with YOYO-1-labeled DNA using Lipofectamine LTX or DMRIE-C as above. At the indicated times, cells were washed in PBS and fixed with 3% paraformaldehyde in 5% serum-containing media for 20 min, washed twice with PBS/3% BSA, quenched with 50 mM NH 4 Cl for 10 min and washed again. For GM130, calnexin, EEA1 and caveolin staining, cells were permeabilized with 0.1% Triton X-100 for 3 min. For LAMP2 staining, cells were permeabilized with saponin (0.4% saponin/1% BSA/2% normal goat serum in PBS) for 20 min. Cells were washed twice with PBS/3%BSA and incubated with primary antibody for 20 min at a 1:200 dilution in PBS/3% BSA (or saponin buffer for LAMP2), washed twice, then incubated with goat anti-mouse Alexa555 or goat anti-rabbit Alexa555 at 1:1000 in the same respective buffers for 20 min. After two washes, cells were allowed to adhere to poly-L-lysinecoated 8-well slides and mounted under coverslips using DAPI-containing Vectashield (Vector Labs, Burlingame, CA). Images were analyzed by fluorescence microscopy using a DeltaVision deconvolution microscope (Applied Precision) with the 100ϫ objective and Soft-Worx (version 3.4.4) deconvolution software, then processed using Adobe Photoshop CS. Quantitation of cells was performed by manual analysis of colocalization from deconvolved z-stacks.
Microtubule Inhibition-CHO cells were transfected as above for 2 h, washed twice in PBS, and replated in fresh media containing 100 ng/ml nocodazole in Me 2 SO (Sigma) or Me 2 SO alone. At the indicated time points cells were fixed and quenched as above, then directly mounted in DAPI-containing Vectashield. 200 cells were counted at each time point, and the percentage of cells with YOYO-1labeled DNA in the nucleus or cytoplasmic organelles was scored.

Lipofectamine LTX and DMRIE-C Complexed with DNA Are
Internalized via Clathrin-mediated and Caveolae-mediated Endocytosis, Respectively-To assess the intracellular distribution of DNA transfected using liposomes, CHO cells were transfected with YOYO-1-labeled DNA and fixed at various time points throughout a 28-h period. After fixation, cells were stained with DAPI to visualize the nucleus and examined by fluorescent microscopy. Two different cationic liposomes, Lipofectamine LTX (LTX) and DMRIE-C, were utilized to compare potential differences in transfection. With both transfection reagents, the maximum percentage of cells containing transfected DNA in cytoplasmic organelles (ϳ98%) was observed by 2 h and in the nucleus by 18 h. However at the times of maximal nuclear DNA, only ϳ50% of cells transfected using DMRIE-C had DNA in the nucleus (Fig. 1A) compared with ϳ70% with LTX ( Fig. 1C). After 8 -10 h, the proportion of cells with DNA remaining in the cytoplasm started to decrease, more so in cells transfected with LTX than with DMRIE-C. Taken together, this suggests that the LTX reagent may be more efficient at promoting nuclear delivery of DNA. Indeed, transfection titers were ϳ5ϫ higher with LTX than DMRIE-C (data not shown).
To identify the route of uptake of a YOYO-1 labeled DNA construct, we pretreated CHO cells for 30 min with various endocytic inhibitors prior to transfection with LTX or DMRIE-C. After 45 min of transfection, cells were washed, fixed, and subjected to FACS analysis to determine the percentage of YOYO-1-positive cells. Approximately 98% of control LTX and DMRIE-C-transfected cells were YOYO-1 positive (data not shown). Filipin and methyl-␤-cyclodextrin (MBC), which are widely used to block caveolae-mediated endocytosis by binding and extraction of cholesterol (reviewed in Ref. 34), greatly inhibited the uptake of DMRIE-C: DNA ( Fig. 2A, left panel). This was unlikely due to disruption of the cholesterol moiety of DMRIE-C per se because the cholesterol-independent tyrosine kinase inhibitor, genistein, also an inhibitor of caveolar uptake (12,13), had a similar inhibitory effect. By contrast, these three caveolar inhibitors had no effect on the uptake of LTX:DNA. Conversely, chlorpromazine, an inhibitor of clathrin-mediated endocytosis (12,35), inhibited uptake of LTX:DNA and had no effect on DMRIE-C:DNA uptake. Transfection with both reagents was unaffected by amiloride, an inhibitor of macropinocytosis (36), at concentrations ranging from 1 to 5 mM (only 3 mM is shown in Fig. 2A).
Although the differential effects of these chemical inhibitors of endocytosis on uptake by the two transfection complexes served as an internal control for their specificity, we decided to confirm these results with biological inhibitors of endocytosis. Dynamin is a GTPase implicated in the fission of both clathrin-coated vesicles and caveolae (37)(38)(39), and its function can be inhibited by overexpression of DN dynamin-1 K44A (40). As expected, DN-dynamin inhibited internalization of both LTXand DMRIE-C-transfected DNA. (Fig. 2A, right panel). The small GTPase Rab5 regulates fusion with early endosomes (EE) (41) at a step following the initial internalization of cargo into clathrin-coated vesicles (42). As expected, its inactive form, DN-Rab5 (S34N), did not affect the initial internalization of DNA by either transfection reagent ( Fig. 2A, right panel). Rab5 function in EE fusion has been proposed to act at least partially via EE recruitment to and motility along microtubules (43). We therefore reasoned that overexpression of DN-Rab5 should inhibit transport of LTX:DNA, but not DMRIE-C:DNA. Indeed LTX:DNA remained colocalized with GFP-DN-Rab5 labeled EEs for up to 4 h, whereas it became less colocalized over time (between 1 and 4 h) in Rab5-WT cells (supplemental Fig. S1A). By contrast, there was very little overlap of DMRIE- To assess the effect of endocytic inhibitors on the transfected human ECD-Fc gene product, cells were treated with endocytic inhibitors for 30 min prior to transfection, and transfection titers assayed by ELISA. Titers from DMRIE-C-transfected cells were suppressed to ϳ30% of control with the three caveolar inhibitors, whereas LTX titers were diminished to 10% in the presence of the clathrin inhibitor (Fig. 2C). Cell viability was not significantly compromised 24 h after washout of these endocytic inhibitors (Fig. 2D), thus the observed differences in uptake and titers were not due to nonspecific toxicity. These data suggest that DMRIE-C and LTX utilize different endocytosis pathways for internalization of transfected DNA, with DMRIE-C relying on caveolae-mediated endocytosis and LTX requiring clathrin-mediated uptake.
Because caution is required in the use of endocytic inhibitors and compensatory uptake mechanisms can come into play (44), we wanted to further confirm the different endocytic pathways of the two transfection reagents by immunofluorescence microscopy. After 2 h of YOYO-1 labeled (green) DNA uptake cells were washed, fixed, permeabilized and stained with antibodies against caveolin (a marker of caveolae and caveosomes (22)); or a marker for the clathrin-mediated pathway, EEA1, an EE-associated protein (45); or the lysosomal-associated membrane protein LAMP2, which localizes to LE as well as lysosomes (46). In agreement with our endocytic inhibition data, LTX: DNA partially colocalized with EEA1 and LAMP2 (Fig. 3, A and B), but not with caveolin (Fig. 3C). Conversely, DMRIE-C:DNA colocalized well with caveolin, but not the clathrin pathway markers EEA1 or LAMP2, also as expected. This colocalization data thus confirms that DMRIE-C:DNA is internalized through caveolae-mediated endocytosis and LTX:DNA is taken up by clathrin-mediated endocytosis.
Post-caveosomal Trafficking of DMRIE-C:DNA Complexes-Although gene transfer by PEI has previously been shown to be caveolae-dependent in some cell lines (12), the post-caveosomal trafficking steps of DNA complexed with PEI or DMRIE-C are unknown. Caveolar-internalized cholera toxin translocates to the Golgi then the endoplasmic reticulum (ER) (35), while the SV40 virus translocates from caveosomes directly to the ER (22). Therefore we examined whether DMRIE-C:DNA similarly trafficked to these organelles at later time points. Cells transfected with YOYO-1labeled DMRIE-C:DNA (or LTX:DNA for comparison) were fixed and stained with antibodies against the ER marker, calnexin or the cis-Golgi marker, GM130. There was no detectable    Fig. 4), nor was there any significant colocalization with EE or LE and lyso-somes, as judged by lack of colocalization with EEA1 and LAMP2. As expected, LTX:DNA endocytosed through the clathrin pathway did not colocalize with the Golgi or ER at any time, remaining partially colocalized with EEA1 and LAMP2 after 5 h of continuous uptake. The clathrin-mediated pathway of transport from the plasma membrane to EE, EE to LE, and LE to lysosomes has been well-studied, with the latter two steps established to be dependent on microtubules (47). Because LTX:DNA follows this pathway, it is almost certain to be microtubule-dependent, and so we did not investigate this directly. Transport out of caveosomes has likewise been demonstrated to be microtubule-dependent (22,35), prompting us to test whether the microtubule-disrupting agent nocodazole had any effect on DMRIE-C:DNA transport. Because caveolae-endocytosed DNA did not colocalize with any of the tested organelles, we asked whether the transfected DNA could escape directly from the caveosomes into the cytoplasm en route to the nucleus, which should not require microtubules. If, however, additional transport steps are required, then microtubule inhibition should decrease the amount of nuclear transfected DNA. To distinguish between these two possibilities, we transfected cells with labeled DNA for 2 h (1-and 2-h post-transfection times shown in Fig. 5), washed away extracellular DNA, then treated cells with nocodazole to depolymerize microtubules, fixing after various times and analyzing the intracellular distribution of DNA by fluorescence microscopy. Under normal conditions (Me 2 SO control), DNA was found in cytoplasmic organelles in 80 -90% of the cells within 2 h of transfection (0 h chase), with significant amounts appearing in the nucleus between 4 and 28 h (Fig. 5A), similar to the time course of nuclear accumulation during continuous DNA uptake (Fig. 1A). Nocodazole treatment prevented this nuclear accumulation, keeping DMRIE-C:DNA in cytoplasmic organelles for up to 28-h post-transfection (26 h chase) (Fig. 5B), while control cells gradually lost their cytoplasmic organelle staining starting at 8 -10 h (6 -8 h chase) (Fig.  5A). (Note that this differs from the continuous uptake localization profile in that the number of cells with cytoplasmic DNA did not decrease until 18 h (Fig. 1A)). Furthermore, the DNA in nocodazole-treated cells remained colocalized with caveolin for up to at least 18 h (16 h chase) (Fig. 5C), showing that in the absence of microtubules, DNA remains trapped in the caveosomes. Although 80% of control cells still had their DNA in caveosomes after 3 h of transfection (1 h chase), this dropped to less than 5% by 18 h (16 h chase). This suggests that DNA does not escape directly from the caveosomes to enter the cytoplasm/nucleus, but requires additional microtubuledependent transport step(s).
Caveosomes containing cholera toxin or the SV40 virus have been shown to communicate with EE en route to the Golgi and/or ER in caveolin-GFP expressing HeLa cells, with ϳ15% of

Rab7-independent Caveolae to Late Endosome Transport
the caveolin-GFP signal localizing to EE at steady state (28). To determine the microtubule-dependent post-caveosomal destination of DMRIE-C:DNA, we re-examined its colocalization with various organelles under chase conditions (without nocodazole). We reasoned that because caveolae and caveosomes are long-lived compared with endosomes (22), perhaps the DNA spends relatively more time in these organelles than in the subsequent compartment(s) and may have been difficult to detect in the latter under conditions of continuous caveolar uptake. Cells were therefore pulsed (or transfected) for 2 h, then washed, re-plated in fresh media and chased for up to an additional 8 h (only up to 2 h shown). At the indicated time points (which reflect chase times), cells were fixed and stained for EEA1 or LAMP2 (Fig. 6A). At no time did we detect colocalization of YOYO-1-labeled DMRIE-C:DNA with EEA1, but we did observe partial colocalization with LAMP2 beginning at 30 min of chase (2.5 h total transfection time, Fig. 6A). Furthermore, we were unable to detect any colocalization of endogenous caveolin with EE-localized Rab5 in either WT, DN (S34N), or DA (Q67L) transfected cells, either with or without DMRIE-C: DNA cargo (supplemental Fig. S2A and data not shown). This is in contrast to previous observations of extensive caveolin-GFP colocalization with EE in DA-Rab5-transfected HeLa cells and moderate caveolin-GFP colocalization with EE in WT-Rab5 cells (28). However, we cannot exclude the possibility that caveolin-GFP is more readily detected than endogenous caveolin, or that there are differences in caveolin distribution between cell types. Taken together, this suggests that DMRIE-C:DNA is trafficked directly from caveosomes to LE or lysosomes, bypassing EE, unlike SV40 and cholera toxin (28). That the LAMP2 colocalization with DMRIE-C:DNA was only par-tial may be attributed to the need for a long pulse (2 h) to detect enough DNA and the long caveolar residence time resulting in poor synchronization of trafficking. DMRIE-C:DNA colocalized with LAMP2 in the LE/lysosomes after 3 h chase, but still not with the EE, Golgi, or ER (Fig. 6B). In cells doubly stained for LAMP2 and caveolin, we did not observe any colocalization of DNA with both of these markers simultaneously (data not shown), nor did we detect colocalization of caveolin with LElocalized WT or DA-Rab7 (Q67L mutant) in DMRIE-C:DNA-transfected cells (supplemental Fig. S2B). This suggests that DNA transfer from caveosomes to the LAMP2positive compartment may involve transport vesicles devoid of caveolin, unlike what was previously found for SV40 transport (28). However, we cannot exclude the possibility that the fusion of a caveolin-positive vesicle was too transient to detect under our conditions. These data indicate that DMRIE-C:DNA is internalized by caveolae, but that the DNA is subsequently, and most likely transiently, transferred to LE and/or lysosomes.
As observed with the continuous uptake, clathrin-mediated endocytosis of LTX:DNA showed some colocalization of DNA with EE and LE/lysosomes, but not the Golgi or ER (Fig. 6B). As expected, neither LTX:DNA nor DMRIE-C:DNA colocalized significantly with recycling endosomes identified with transferrin or GFP-Rab11 (data not shown). Thus, although DMRIE-C:DNA and LTX:DNA are initially internalized by different mechanisms, both pathways converge at the LE or lysosomes.
Biological and Chemical Inhibition of Lysosomal Traffic Does Not Affect Nuclear Localization of DMRIE-C:DNA or LTX:DNA-Because previous reports suggested that DNA degradation in lysosomes could be a major reason for ineffective nuclear delivery (1,4,5), we attempted to determine whether DNA present in the LAMP2 compartment has to traverse lysosomes to enter the cytoplasm, or if it could escape directly from LE (in which case blocking lysosomal activity might increase transfection product titers). 30 min prior to and during a 4-h transfection, cells were treated with chloroquine, a drug that compromises lysosomal integrity by preventing their acidification (23). Following a 48-h chase, supernatants were assayed by ELISA for transfected human Fc, and revealed no difference after chloroquine treatment whether transfected with LTX or DMRIE-C (Fig. 7), suggesting that lysosomal function is not required for nuclear delivery of DNA. This is not without precedent, as chloroquine treatment of COS-7 cells transfected using a DMRIE/DOPE (dioleoylphosphatidylethanolamine) formulation also did not affect reporter gene expres-  AUGUST 3, 2007 • VOLUME 282 • NUMBER 31 sion (48). Presumably DMRIE-C and LTX transfected DNA are able to escape from the LE to reach the nucleus under conditions of lysosomal disruption. Indeed, cationic liposomes have been demonstrated to destabilize endosomal membranes, allowing the transfected DNA to leave the LE or lysosome (26).

Rab7-independent Caveolae to Late Endosome Transport
Because lysosomotropic agents such as chloroquine can sometimes render lysosomes more permeable (potentially aiding release of DNA complexes) (49), we took an independent approach to test whether transfected DNA must traffic to the lysosome by examining the nuclear localization of transfected DNA in cells overexpressing DN-Rab7. The precise role of the small GTPase Rab7 is controversial, as it has been reported to regulate traffic from EE to LE (14,50), as well as from LE to lysosomes (29,51,52) and LE back to EE. Two DN mutants have been demonstrated to inhibit these processes, N125I and T22N (15,29). First we sought to determine which endosomal transport step Rab7 controls in our system by analyzing diI-LDL uptake in CHO cells expressing Rab7-GFP-WT or DN. Following a 10-min pulse, LDL transiently colocalized with WT-Rab7 (itself a LE marker) (14,29,52) between 30 and 45 min chase and showed less overlap by 70 min of chase, consistent with previous time courses of LE and lysosomal delivery in CHO cells (53). By contrast, LDL remained colocalized with Rab7 for over 70 min in DN-Rab7 (T22N)-expressing cells, implying that LE-lysosome transport is the most likely step regulated by Rab7 in this system (supplemental Fig. S3A).
We then examined the effect of blocking LE-lysosomal transport with DN-Rab7 on LTX:DNA and DMRIE-C:DNA localization. After 2 h of transfection with LOLO-1 (red) labeled DNA, cells were washed, and chased for up to 2 h, then stained for EEA1 or LAMP2 (Fig. 8A, supplemental Fig. S4, panels b, d, f, h correspond to LAMP2 panels in Fig. 8A). To more clearly illustrate DNA-Rab7 colocalization, EEA1/LAMP2 staining was omitted in Fig. 8A and displayed in supplemental Fig. S4. LTX:DNA was partially colocalized with both WT-and DN-Rab7 (Fig. 8A) and EEA1 and LAMP2 at all time points (supplemental Fig. S4); and both WT-and DN-Rab7 partially colocal-ized with LAMP2, but not EEA1, confirming their expected LE localization. We did not observe any difference in the extent of LTX:DNA colocalization with EEA1 in WT versus DN-Rab7 cells (Fig. 8A, graphs in right panels), in agreement with our LDL data (supplemental Fig. S3A) suggesting that Rab7 does not affect EE to LE transport of LTX:DNA. More importantly, we did not detect any difference in the colocalization of LTX:DNA with LAMP2-or WT-and DN-Rab7, in agreement with our chloroquine data that DNA must be able to escape from the LE. DMRIE-C:DNA did not colocalize with EEA1 at any time point, as expected, but did with LAMP2 and Rab7 after 30 min of chase in both WT-and DN-Rab7 cells to a similar extent (Fig.  8A), again implying DMRIE-C:DNA can escape from LE when lysosomal traffic is inhibited (and possibly also under normal conditions). Furthermore, this demonstrates for the first time that caveosome to LE traffic does not require Rab7 activity, in contrast to the involvement of Rab5 in caveosome to EE transport (28). To directly confirm that DNA can escape the LE and translocate to the nucleus, we examined the nuclear versus cytoplasmic distribution of DNA at later time points (18-h chase). Both N125I and T22N GFP-DN-Rab7 and Q67L GFP-DA-Rab7 constructs had no effect on the nuclear localization of DNA transfected with either reagent compared with WT-Rab7 or GFP control expressing cells (Fig. 8B), supporting the ability of DNA to escape from LE. By contrast, the diminished nuclear localization of LTX:DNA in DN-Rab5 cells (Fig. 2B) suggests that LTX:DNA is not similarly able to escape from EE, while the lack of DN-Rab5 effect on DMRIE-C:DNA supports the bypassing of the EE by this transfection reagent.
To further confirm these results, we analyzed the effect of inhibiting the Rab7 effector, RILP (Rab-interacting lysosomal protein), which induces recruitment of the dynein-dynactin motor complex to regulate late endocytic traffic. Over-expression of the C-terminal portion of RILP (RILP-C33) acts as a dominant negative, causing lysosomal fragmentation and inhibition of late endosomal traffic similar to DN-Rab7 (30,31). We first verified that GFP-tagged full length RILP and RILP-C33 were functional by showing that RILP-C33, but not full length RILP, inhibited the degradation of diI-LDL as measured by FACS analysis (supplemental Fig. S3B). Confirming the DN-Rab7 results, overexpression of RILP-C33 had no effect on the nuclear delivery of DNA transfected with either reagent (Fig.  8C). Thus, both chemical and biological inhibition of lysosomes does not increase transfection efficiency, suggesting that DMRIE-C:DNA and LTX:DNA do not require trafficking from the LE to the lysosome to localize to the nucleus. Even under conditions of enhanced lysosomal trafficking, generated by overexpression of DA-Rab7, DNA transport to the nucleus was unperturbed.

DISCUSSION
The mechanism of DNA internalization varies widely among cell types and transfection reagents. We aimed to further understand the endocytic pathway mediated by liposome-DNA complexes in DP12 CHO cells. Our data show that two commercial liposomes, DMRIE-C and Lipofectamine LTX, promote initial DNA internalization by separate endocytic path-  ways in these cells, with DMRIE-C:DNA complexes depending on caveolae-mediated internalization and LTX:DNA complexes relying on clathrin-mediated endocytosis. DMRIE-C: DNA is transported from caveosomes to LE or lysosomes in a microtubule-dependent manner that does not require Rab7 activity and bypasses EE. While we cannot formally exclude the possibility that DMRIE-C:DNA had already traversed EE during the first 2 h of transfection, we think this unlikely because of the high degree of caveolin colocalization at this time point and more importantly the lack of effect of DN-Rab5 on nuclear localization of DMRIE-C:DNA. Although the lack of caveolin colocalization with WT, DN-, or DA-Rab5 further suggests that DMRIE-C:DNA bypasses the EE (supplemental Fig. S2 and data not shown), we cannot eliminate the possibility that this colocalization was undetectable in our cells with these staining conditions. The only other cargo, to the best of our knowledge, which bypasses EE en route from caveolae or caveosomes to LE is the EGF receptor, under conditions of recombinant decorin treatment (54). We never observed simultaneous colocalization of DMRIE-C:DNA with caveolin and LAMP2 (supplemental Fig. S2), suggesting that DNA transfer from caveosomes to the LAMP2-positive compartment either involves transport vesicles devoid of caveolin, or that caveolin-positive vesicles fuse only transiently with LE, similar to the fusion of caveolin-positive vesicles with endosomal compartments seen during SV40 transport (28). However, in contrast to the Rab5 involvement in caveolin-positive vesicle-EE fusion in SV40 and cholera toxin transport (28), our caveosome-LE transport step does not appear to similarly utilize Rab7. The exact transport step to or from the LE that requires Rab7 remains a topic of debate. Because we did not observe any accumulation of LDL or LTX:DNA in the EE of DN-Rab7 cells, and DMRIE-C:DNA similarly accumulates in LE even though it bypasses EE, our data strongly suggest that Rab7 regulates LE to lysosome traffic in our cells. The maturation of EE to LE is marked by the exchange of Rab5 for Rab7 (55,56) and Rab5 was found to be required for exchange of material between caveosomes and EE (28). However, our observations that transfection efficiency was unaffected by DN-Rab7, DA-Rab7, or RILP-C33 suggest that traffic from caveosomes to LE does not similarly depend on Rab7 or its microtubule motor-recruiting effector RILP, despite being microtubule-dependent.

Rab7-independent Caveolae to Late Endosome Transport
The most plausible explanation for the lack of difference in transfection efficiency seen in DN-Rab7 and DA-Rab7 cells is that the DNA escapes directly from the LE. However, we cannot exclude the possibility that under conditions where lysosomal transport is not inhibited, DNA:liposome complexes could escape from lysosomes instead of, or in addition to, LE, because LAMP2 labels both compartments. LE/lysosomal escape may occur through a mechanism in which cationic liposomes mediate fusion of the transfection complexes with the endosomal membrane, as has been visualized by electron microscopy (57,58). Liposomes have been shown to destabilize the membrane in a detergent-like manner, allowing the DNA to be released into the cytosol (59, 60). In our system it is unclear whether it is free DNA molecules or DNA:liposome complexes that escape from the LE/lysosome into the cytosol, where they would presumably await cell division to diffuse into the nucleus (1). The proximity of lysosomes to the nucleus means that DNA would be released into the perinuclear region of the cytoplasm, perhaps facilitating its nuclear entry during cell division when the nuclear membrane is disassembled (3).
As the structure and composition of these cationic liposomes are not disclosed by the manufacturer, one can only speculate as to why they would be internalized by alternative endocytic pathways only to converge at a later step. One possibility is that the uptake mechanism simply depends on the size of the DNA: transfection reagent complex, because it was previously found that the uptake of latex beads smaller than 200 nm occurs via clathrin-coated vesicles, while larger ones were internalized by caveolae (61,62). However, others have demonstrated that large complexes of cationic polymers and DNA (Ͼ200 nm) are internalized by macropinocytosis, intermediate complexes (100 -200 nm) by clathrin coated vesicles and small complexes (Յ100 nm) were internalized through caveolae (63). Our preliminary results suggest that LTX:DNA complexes are smaller than DMRIE-C:DNA complexes (data not shown), in agreement with the former hypothesis. Alternatively, the DMRIE-C reagent contains a cholesterol component, which may allow it to associate with the cholesterol-rich regions of the plasma membrane, such as lipid rafts and caveolae, thus potentially explaining its caveolar uptake.
Interestingly, the percentage of cells with transfected DNA localized to the nucleus is higher in cells transfected using LTX than DMRIE-C (Fig. 1), perhaps explaining why the raw values from the Fc ELISA were ϳ5ϫ higher with the LTX reagent (data not shown). As both cationic liposomes appear to mediate delivery to the LE/lysosomes, the differences in nuclear localization are likely a result of the liposomes' ability to promote delivery from this organelle to the nucleus. Because of their large size, liposome:DNA complexes require active transport to enter the nucleus under non-mitotic conditions and it is possible that the enhanced nuclear localization observed in LTX transfections is based on its superior ability to promote active transport. For example, the LTX might associate with an NLScontaining protein that allows for active transport of the cargo. Alternatively, the LTX may exhibit an enhanced ability to diffuse to the nucleus during cell division because it has a particular affinity for chromatin or another nuclear component. The precise mechanism by which the cationic liposome:DNA complexes translocate into the nucleus is unknown and remains an important area of investigation for future research efforts to improve non-viral gene therapy. Because our data suggest that lysosomal degradation is not a major barrier for gene delivery to the nucleus using LTX or DMRIE-C, we propose that future efforts to improve transfection efficiency with cationic liposomes should focus not on blocking LE-lysosome transport or compromising lysosomal integrity, but rather on improving transport from LE/lysosome to the nucleus. Although these results were disappointing in that we were unable to increase transfection titers by inhibiting lysosomal degradation of DNA, they are interesting from a membrane trafficking perspective in that not only do we show that a cargo (DMRIE-C:DNA) is transported directly from cavesomes to LE, bypassing EE, but that this step is microtubule-dependent and Rab7-independent.