Introduction

The blood–brain barrier is a major obstacle to the delivery of therapeutics into the central nervous system (CNS).¹ Long-term drug delivery into the CNS by physical means such as indwelling catheters is associated with many risks and complications.² An alternative method for continuous delivery of therapeutic proteins into the CNS is gene therapy. Long-term protein expression through transgene delivery to the CNS in humans may have fewer complications and higher efficacy than the currently available physical protein delivery techniques. A minimally invasive route for direct gene delivery to the CNS is through lumbar puncture into the cerebrospinal fluid (CSF)-filled space that surrounds the spinal cord [intrathecal (i.t.) space]. Most nonsecreted transgene expression resulting from i.t. gene vector injection is localized to cells of the meninges.^3,4,5 However, proteins and molecules in the CSF of the i.t. space can move along perivascular spaces and axon tracts into the spinal cord.⁶ In this manner, transgene expression in the cells of the meninges may affect cells in the brain and spinal cord parenchyma.

Delivery of viral vectors into the lumbar and cisterna magna i.t. space has been used for treating hearing loss,⁷ neuropathic pain,⁸ multiple sclerosis,⁹ disseminated brain tumor,¹⁰ and ischemic spinal cord injury¹¹ in rodents. However, the use of viral vectors is associated with a number of safety issues.¹² Nonviral vector i.t. lumbar gene delivery has been used for regenerating peripheral nerves after transection,¹³ and for treating various animal models of neuropathic pain.^{14,15,16,17,18,19,20} Two of these reports included evidence of significant therapeutic benefit and transgene expression at 5 and 6 days after i.t. injection of a naked DNA vector with no subsequent physical manipulation such as electroporation.^15,20 Injection of naked DNA has been most successful in muscle²¹ and liver. Therapeutic levels of human factor IX (up to18 g/ml)²² and human -1-antitrypsin (>1 mg/ml)²³ were produced for 7 months and 3 months, respectively, following hydrodynamic plasmid delivery to the liver. Intrathecal injection of naked DNA produced ~5–50% of the expression of i.t. cationic lipid/DNA,^5,15,19 PEI/DNA,^3,4 PPE-EA/DNA,⁴ PEG-PEI/DNA³ or ultrasound contrast reagent/DNA or followed by ultrasound treatment.^24,25 Secreted transgene production in CSF has been measured in a small portion of i.t. gene therapy studies, with expression being observed for 1 month (lentivirus),²⁶ 14 days (electroporation),¹⁴ or 6 days (naked DNA)¹⁵ after gene delivery.

The main advantage of using naked DNA as a gene therapy vector is that it is simple to produce and safe to use. Safety is derived from achievable purity, an extremely low frequency of recombination into the genome, and low immunogenicity. Naked DNA has been injected repeatedly in nonhuman primates without inducing anti-DNA antibodies.²⁷ No serious adverse events have been reported in the >200 naked DNA gene therapy clinical trials to date.²⁸ Ten percent of all trials between 1995 and 1999 used naked DNA while 25% did so during the 5 year period ending in 2007.²⁹ The apparent safety of using naked DNA may be a factor behind this increase in its use.

Together, these reports suggest that i.t. injection of naked DNA would be a safe method for gene delivery to the CNS, and could be capable of producing detectable quantities of secreted proteins in the CSF and therapeutic effects in animal models. The work reported here extends previous observations of expression after i.t. delivery of naked DNA and, for the first time, characterizes the distribution of the secreted transgene product in the CSF.

Results

SEAP expression in vivo after i.t. injection of naked DNA

To determine whether long-term transgene expression can be achieved through i.t. delivery of naked DNA in rats, lumbosacral CSF (CSF between the DNA delivery site and 3.5 cm caudal to it) was sampled at 7, 14, 92, and 140 days after injection of 25 g of plasmid pTR-secreted alkaline phosphatase (SEAP). The pTR-SEAP plasmid contains adeno-associated virus serotype 2 inverted terminal repeats flanking the expression cassette. This feature has been shown to improve transgene expression in several systems including rat brain³⁰ and zebrafish.³¹ pTR-SEAP also contains a constitutive mammalian promoter (chicken -actin/CMV enhancer) which was shown to produce luciferase expression for 2 months after i.t. delivery of PEI/DNA complexes.³ Human SEAP was detected (greater than naive CSF background + 3 standard deviations; n = 15) in five of six rats at 7 days and 14 days, but in only one of the six rats at 92 days and 140 days after the injection (data not shown).

Next, we wanted to determine whether we could obtain more long-lasting or greater transgene expression by utilizing a different expression vector incorporating two elements that are reported to improve transgene expression. This plasmid, pCpGfree-SEAP, has matrix attachment regions (MARs) flanking the expression cassette, and totally lacks CpG dinucleotides outside of the SEAP gene. MARs are thought to inhibit silencing and/or allow episomal maintenance of plasmids in mammalian cells.³² CpG motifs increase plasmid immunogenicity,³³ and in vitro methylation of plasmid CpG dinucleotides decreases subsequent in vivo expression.³⁴ Immune response to plasmid CpG motifs and de novo in vivo plasmid CpG methylation are thought to limit transgene expression.^34,35 However, a recent report does not support de novo in vivo CpG methylation-induced silencing.³⁶ We compared the pTR-SEAP and pCpGfree-SEAP plasmids in vivo along with a third plasmid that encodes human interleukin-10 (pCpGfree-hIL10). This third plasmid was a negative control which would not produce any SEAP activity. A diagram of the time course of this experiment is shown in Figure 1. Rats (n = 24) were injected i.t. with 100 g of one of the three plasmids. One day later, lumbosacral CSF was collected from 14 of the 24 rats. Two days after the injection, lumbosacral CSF was collected from the 10 other rats. An earlier report had shown that the efficacy of plasmid therapy in a neuropathic pain model was prolonged by a repeat administration of plasmid 2 days later.²⁰ Therefore, to determine whether two 100 g injections spaced 2 days apart could produce higher expression than a single 100 g injection, these 10 rats were reinjected with an additional 100 g of the same plasmid that they had originally received. This second injection did not increase transgene expression for either of the plasmids tested (dotted lines, Figure 2a). Overall, between 0 and 116 days after the injection, rats injected with pCpGfree-SEAP, pTR-SEAP, or the control (pCpGfree-hIL10) had significantly different SEAP expression [P < 0.01, generalized estimating equation (GEE)]. The SEAP concentrations in CSF, collected either 1 day ("diamond" and "triangle" groups) or 2 days ("square" and "crossout" groups) after a 100 g injection of either plasmid were not significantly different for the two plasmids [Figure 2a, P > 0.05, generalized linear model (GLZ)]. The nine rats from the two groups that had received injections of pCpGfree-SEAP ("diamond" and "square", Figure 2a) had the highest SEAP levels (between 250 and 1,000 pmol/l at 1 or 2 days after the injection), while those injected with pTR-SEAP ("triangle" and "crossout", Figure 2a) had SEAP levels between 6 and 165 pmol/l, and the values in the control group (pCpGfree-hIL10, n = 6) ranged from 1 to 3 pmol/l ("circle," Figure 2a). Expression in individual rats across time was consistent. There was a significant correlation between expression at 1 or 2 days and at 14 days after the injection (Spearman's = 0.56, P < 0.05, n = 18) and between all other subsequent adjacent time points in SEAP transgene-injected rats (Spearman's 0.82, P < 0.01, n = 18).

Figure 1.

Experimental design for Figures 2 and 4.

Full figure (12K)

Figure 2.

Long-term SEAP expression in lumbosacral CSF after i.t. injection of naked DNA. (a) All the rats were injected with 100 g of the indicated plasmid at the first time point (day 0, arrow). The rats in the "square" and "crossout" groups (broken lines, each n = 5) and two of the six rats in the "circle" group were then reinjected with 100 g of plasmid two days later (day 2, arrow) immediately after CSF collection. Lumbosacral CSF samples were collected from each rat at the indicated time points (0, 1 or 2, 14, 28, 56, 84, and 116) for up to 7 samples per rat. n = 4 for the "diamond" and "triangle" groups. (b) SEAP expression in the rats belonging to the groups displayed in panel a, after an additional plasmid injection. After CSF collection on day 116, the rats in all the groups were injected with 100 g of the indicated plasmids (day 116, arrow). The animals in the "diamond" and "triangle" groups (thick solid lines, n = 4 in each group) were injected with pCpGfree-SEAP, while the ones in the "square" and "crossout" groups (broken lines, n = 5 in each group), and two of the six rats in the "circle" group were injected with pCpGfree-GFP while the other four animals in the "circle" group were injected with pCpGfree-hIL10 (thin solid line). CSF was then collected from the animals in all the groups at day 117 and at day 130. Error bars are mean values SEM. CSF, cerebrospinal fluid; SEAP, secreted alkaline phosphatase.

Full figure and legend (27K)

A second injection at day 116 increases transgene expression

Some applications of gene therapy may require more long-lasting or higher expression than can be achieved by using one injection, and this makes reinjection efficacy important. Therefore, we next determined whether reinjection of pCpGfree-SEAP at ~4 months after the initial injection would increase transgene expression. Immediately after CSF collection at 116 days after the initial injection, groups of rats that had received only one previous injection with a SEAP plasmid ("diamond" and "triangle" Figure 2a) were injected again with pCpGfree-SEAP. Similarly, the rats that had been injected once with pCpGfree-hIL10 were reinjected. The rest of the rats were injected with 100 g of pCpGfree-GFP to control for the effect of plasmid injection on ongoing SEAP expression, and CSF was collected 1 and 14 days later. As expected, SEAP concentrations in the lumbosacral CSF of rats previously injected with pCpGfree-hIL10 did not increase after injection with either pCpGfree-GFP or pCpGfree-hIL10 ("circle," Figure 2b). There was a transient SEAP increase 1 day after injection of pCpGfree-GFP in rats previously injected with SEAP-encoding plasmids. Despite this transient increase, rats injected with pCpGfree-SEAP ("diamond" and "triangle" Figure 2b) had higher SEAP than their respective pCpGfree-GFP-injected animals in the control groups (each n = 5, broken lines Figure 2b, GEE, P < 0.05).

Although pCpGfree-SEAP produced higher expression than pTR-SEAP, expression from both plasmids decreased over time to a similar extent (Figure 2a). Chastain et al. found that mice that initially expressed the highest levels of SEAP at 3 or 4 days after injection demonstrated a fall in SEAP to background levels 8 weeks after intramuscular injection of an SEAP-encoding plasmid. This preferential shutoff of initial high expressers correlated with serum anti-SEAP antibody titers.³⁷ To determine whether this mechanism was also responsible for the loss of SEAP expression observed in the experiment, the relationship between expression at day 1 or 2 and at 8 weeks was analyzed. There was no consistent shutoff of initial high expressers at 8 weeks after injection (Figure 3). In a separate experiment, 8 rats were injected with 100 g of pCpGfree-SEAP each, and 28 days later sera and lumbosacral CSF were collected and analyzed by means of ELISA for the presence of anti-SEAP antibodies (see Supplementary Materials and Methods). There was no significant difference detected in either CSF or serum between naive and pCpGfree-SEAP-injected rats (Supplementary Table S1, P > 0.05, GLZ).

Figure 3.

Relationship between expression at day 1 or 2 and expression at 8 weeks from Figure 2. The rats in all the groups were injected with a total of 100 or 200 g of the indicated plasmid. The broken line represents the background activity (mean value + 3 standard deviations of pCpGfree-hIL10 injected controls 8 weeks after the injection, n = 5).

Full figure and legend (11K)

Human IL-10 expression after pCpGfree-hIL10 injection

To determine whether the results described above are unique to the SEAP transgene, aliquots of CSF samples obtained in the experiment described earlier (Figure 2) were then assayed for hIL-10 content using ELISA. The two available samples taken from pCpGfree-hIL10 injected rats at day 1 after the injection measured 79 and 51 pmol/l hIL-10, respectively, while the highest value of hIL-10 in the SEAP plasmid-injected rats was only 0.8 pmol/l (n = 6, Figure 4a). Significant hIL-10 expression was detected at 84 to 116 days after the injection in the combined pCpGfree-hIL10-injected groups of rats (n = 6) as compared to pCpGfree-SEAP-injected animals (n = 7 GEE, P < 0.01 Figure 4a). Furthermore, there was a significant correlation between hIL-10 concentration in lumbosacral CSF across time in individual rats (day 84 vs. day 116 Spearman's = 0.83, P < 0.05, N = 6). Reinjection of pCpGfree-hIL10 (n = 4) at day 116 increased the hIL-10 concentration from day 117 to day 130 as compared to the SEAP-injected controls (n = 6, GEE, P < 0.05 Figure 4b). That is, as with SEAP, hIL-10 expression can be detected for close to 4 months after the injection, this expression decreases over time, and reinjection of the hIL-10 transgene at ~4 months after the initial injection again increases expression.

Figure 4.

Long-term hIL-10 expression in lumbosacral CSF after i.t. injection of naked DNA. Aliquots of the samples from the experiment described in Figure 2 were assayed for hIL-10 using ELISA. Arrows indicate the time points at which injections were administered to at least one group of rats. (a) Human IL-10 expression in lumbosacral CSF after a single pCpGfree-hIL10 100 g injection (first arrow, "diamond" group) or two 100 g injections separated by an interval of two days (arrows, "square" group). (b) Human IL-10 concentrations in lumbosacral CSF after reinjection (arrow) of pCpGfree-hIL10 ("diamond" group) or pCpGfree-GFP ("square" group) 116 days after the initial pCpGfree-hIL10 injection. The animals in the control group ("circle") were injected with 100 g of either pCpGfree-GFP or pCpGfree-SEAP (arrow). The CSF samples available for hIL-10 analysis varied greatly, and therefore the data points represent varying numbers of samples. For the animals in the "diamond" group: day 1 (n = 1), day 14 (n = 2), day 84 (n = 4), day 116 (n = 4), day 117 (n = 4), and day 130 (n = 4) after the initial injection. For the animals in the "square" group: for day 1n = 1, and for all other time points n = 2. For the animals in the control group ("circle"): n > 6. Preinjection samples from the animals in all the groups were combined to create the day 0 data point (n = 5). Error bars are mean values SEM.CSF, cerebrospinal fluid; hIL-10, human interleukin-10; SEAP, secreted alkaline phosphatase.

Full figure and legend (20K)

Half-life of SEAP protein in the i.t. space

To determine how well lumbosacral SEAP concentration reflects ongoing expression, the half-life of SEAP protein in lumbosacral CSF was measured. Sixteen rats were injected i.t. with 25 ng of purified SEAP protein. Immediately before the SEAP protein injection, lumbosacral CSF was collected (day 0, Figure 5). CSF was sampled two more times in each rat, once at a time point of 0.5, or 3, or 24, or 48 hours, and then once more (for all the rats) at 5 days after injection. Serum and cisterna magna CSF were also collected as indicated (Figure 5). There was a significant rise in serum SEAP activity at 3 and 24 hours after the injection as compared to preinjection values (Figure 5, GLZ, P < 0.01), and SEAP concentration in the cisterna magna at 24 hours was significantly higher than at 48 hours after the injection (Figure 5, GLZ, P < 0.05). It appears, therefore, that SEAP protein can move from the CNS to the serum and from the delivery site to the cisterna magna. SEAP activity is cleared quickly from the lumbosacral CSF with a half-life of 7 1 hours and is undetectable 5 days after the injection. This indicates that plasmid-induced SEAP activity is the result of ongoing expression rather than enduring levels of SEAP produced transiently after plasmid injection.

Figure 5.

SEAP is rapidly cleared from lumbosacral CSF. SEAP protein (25 ng) was injected i.t. (arrow) and CSF was extracted at 0.5, 3, 24, and 48 hours after the injection from 4 separate groups of rats (for each group, n = 4). Lumbosacral CSF was then drawn for a second time from all of these rats at 120 hours after the injection. Serum and cervical CSF were also drawn at the time points indicated. Error bars are mean values SEM. CSF, cerebrospinal fluid; SEAP, secreted alkaline phosphatase.

Full figure and legend (15K)

Distribution of plasmid-produced SEAP

Cisterna magna i.t. injection of naked DNA results in luciferase expression localized to the brain stem.⁴ Lumbar i.t. injection results in luciferase expression localized to the lumbosacral-thoracic area of the spinal cord³ and not to the muscle or sciatic nerve.⁵ The distribution of a secreted protein in the CSF after i.t. plasmid gene delivery has not been elucidated previously. We therefore characterized SEAP distribution in CSF and serum in several experiments. In one group of experiments, CSF was collected at the DNA delivery site (~vertebral level T13) and 3.5 cm caudal to it (~vertebral level L5) at 7, 14, or 92 days after injection of pTR-SEAP (25 g). SEAP concentrations at these two sites correlated well (Spearman's = 0.97, n = 60, P < 0.01, Figure 6). The best line fit to the data, has a slope of 1.09, indicating that expression is very similar at the two sites. In another set of experiments, cisterna magna CSF (~8.5 cm rostral of the DNA delivery site), serum, and lumbosacral CSF were collected from the animals in the groups described in Figure 2 at 84 and 130 days after the injection. There was a significant correlation between SEAP expression in cisterna magna CSF and lumbosacral CSF (Spearman's = 0.92, P < 0.01, n = 13) and also between SEAP concentration in serum and lumbosacral CSF (Spearman's = 0.75, P < 0.05, n = 10). Cisterna magna CSF and serum were also collected simultaneously along with lumbosacral CSF in other experiments. A summary of the data relating to distribution of SEAP activity in the CSF and serum is shown in Table 1. SEAP concentration is highest and fairly uniform near the site of DNA delivery, is present at lower concentrations in the CSF far from the injection site (cisterna magna), and is detectable but greatly reduced in the serum. These data are consistent with the conclusion that most of the SEAP is produced in cells near the site of DNA delivery in the CNS.

Figure 6.

SEAP concentration is similar at the DNA delivery site and at ~3.5 cm caudal to the plasmid delivery site. An amount of ~15 l of CSF was first drawn at a site 3.5 cm caudal to the DNA delivery site by collecting natural outflow from a needle inserted between L5 and L6. A catheter was then threaded through the needle 3.5 cm up to the point of DNA delivery at ~L1, and another 15 l of CSF was collected through the catheter. The 60 data points in this graph relate to animals from several treatment groups (see methods) at several time points (1 week, 2 weeks, and 3 months after the injection). None of these cofactors had a significant effect on SEAP distribution between the two points, and therefore the data were combined for this analysis. One outlying data point (71, 37) was not included in this analysis. CSF, cerebrospinal fluid; SEAP, secreted alkaline phosphatase.

Full figure and legend (15K)

Table 1 - SEAP distribution.

Full table (29K)

Rat CSF contains less DNAse activity than serum does

Nucleases appear to affect in vivo transfection efficiency as nuclease inhibition increases transgene delivery/expression in the lungs,³⁸ skin,³⁹ and salivary glands.⁴⁰ Cationic lipids protect plasmids from nuclease degradation.⁴¹ We hypothesized that one explanation for the finding that naked DNA complexed with cationic lipid enhances expression 1,000 times in intravenous delivery⁴² but only a factor of 2 when delivered i.t.,¹⁵ is that serum contains more nuclease activity than CSF does. One report showed that phosphodiester oligonucleotides are stable in human CSF.⁴³ We decided to test whether plasmid DNA at a concentration similar to that present in vivo 30 minutes after a 100 g injection (400 g/ml, see Materials and Methods) would be stable in freshly isolated rat CSF. Plasmids incubated in CSF at 37 °C for 30 minutes containing between 0.3 and 25% serum were degraded, while plasmids incubated in either PBS or CSF with 0.003% serum were not (Figure 7). These data indicate that degradation of plasmid DNA in CSF is minimal, and this may increase the efficacy of i.t. gene delivery of naked DNA.

Figure 7.

CSF contains less DNAse activity than serum does. Plasmid DNA (20 g in 50 l CSF) was incubated at room temperature for 12 minutes and then at 37 °C for 18 minutes in freshly isolated rat CSF containing various levels of rat serum, ranging in concentration from 0.003% serum in the two rightmost lanes to 25% serum in the lane marked 25%. The arrows indicate degradation bands that are either not present or are much lighter in the DNA incubated in either phosphate-buffered saline (PBS) or CSF containing very little (0.003%) serum. The lane marked L contains the same plasmid DNA cut with the restriction enzyme XbaI which linearizes the plasmid. M lanes contain the size marking ladder. Some repetitive intervening lanes between 0.3 and 0.003% were removed to compact the image. CSF, cerebrospinal fluid.

Full figure and legend (44K)

Discussion

The longest reported periods of transgene expression in the CNS after i.t. gene delivery using nonviral and naked DNA is 56 days (PEI/DNA, luciferase)³ and 28 days (naked DNA, luciferase).⁴ Our work adds to these reports relating to duration of intracellular gene expression by demonstrating that i.t. delivery of naked DNA can produce significant amounts of secreted proteins in the CSF for ~4 months (Figures 2 and 4). We found that, at only 1 day after plasmid injection, transgene expression appeared maximal and then decreased thereafter. The levels of expression measured after a single 100 g injection of pCpGfree-SEAP, from 1 day after the injection (552 90 pmol/l, n = 4) to 4 months after the injection (10 4 pmol/l, n = 4), or of pCpGfree-hIL10 from 1 day (65 14 pmol/l, n = 2) to 3 months (3.2 0.7 pmol/l, n = 4), are physiologically relevant for some proteins. For example, the half-maximal effect of several cytokines, including hIL-10 (ref. 44), IL-2 (ref. 45), and IL-5 (ref. 46) is observed at concentrations of <10 pmol/l. Also, ~50 pmol/l -endorphin was therapeutically effective in a rat neuropathic pain model.^14,17

Plasmid reinjection

To examine the ability to reintroduce transgenes, a SEAP plasmid injection was administered 2 days or ~4 months after an initial SEAP transgene injection. Reinjection at 2 days did not significantly increase transgene expression over the time course (Figure 2a). In contrast, reinjection at 116 days significantly increased transgene expression as compared to animals in a control group that had initially been injected with an SEAP transgene and then, ~4 months later, with pCpGfree-GFP (Figure 2b). There was a transient increase in SEAP expression in animals in this control group, perhaps because of increased expression from the SEAP-encoding plasmid already present,⁴⁷ or to decreased elimination of SEAP. Lumbar puncture produces spontaneous CSF outflow from rats that have received no lumbar puncture within the previous one week; however, this outflow is absent in the context of a recent (within ~7 days) DNA injection. This altered CSF flow could alter elimination of SEAP from lumbosacral CSF.

The pCpGfree plasmid enhances expression but does not eliminate loss of expression

We found that the pCpGfree-SEAP plasmid produces ~ 10 times more SEAP than pTR-SEAP does in vivo (Figure 2a). This difference was not the result of a generally weaker expression from pTR-SEAP, given that it produced ~50% more expression than pCpGfree-SEAP in HEK-293 cells (data not shown). The presence of MARs,⁴⁸ lack of CpG dinucleotides,^34,36 or promoter differences could allow higher expression from pCpGfree-SEAP in vivo. While the pCpGfree plasmid resulted in approximately tenfold higher transgene expression than the pTR plasmid, the expressions from each of the plasmids decreased over time to similar extents (Figure 2a). Chen et al. found that the expression from a CpG-depleted plasmid (similar to pCpGfree) fell by a factor of 15 (hFIX) or 32 (hAAT) at 90 days after hydrodynamic delivery to the liver, apparently because of transcriptional silencing.³⁶ This finding is similar to the 22-fold (hIL-10) or 28-fold (SEAP) decrease over 84 days observed with the pCpGfree plasmid and the 34-fold decrease (SEAP) seen with the pTR plasmid over the same time period. These decreases are much less than the 59-fold (hFIX) and 500-fold (hAAT) decreases seen with CpG-replete plasmids over the same time period, and therefore it is possible that some aspects of pTR and pCpGfree plasmid may be limiting the expression loss. In this same report, it was found that removal of the plasmid backbone limited the expression loss to only threefold (hFIX) or fourfold (hAAT).³⁶ Removal of the transcriptionally silent areas of pCpGfree-SEAP and pTR-SEAP or use of the full-length genomic gene, (which greatly reduces expression loss over time in the liver),²³ could limit time-dependent losses.

We found that repeated sampling did not increase in interleukin 1 (IL-1) levels in CSF nor did it cause weight loss in the rats (see methods). Further, the lack of a steep SEAP concentration gradient caudal to the injection site allows accurate quantification of expression from CSF samples collected in slightly different locations. These findings indicate that repeated lumbar puncture sampling of a secreted transgene product in CSF is an effective method of monitoring transgene expression over the long term, lending itself well to optimization of plasmids for long-term i.t. expression.

In summary, this study provides evidence that i.t. naked DNA can produce at least an ~4-month transgene expression of a secreted product, uniformly distributed over a span of 3.5 cm in the lumbosacral CSF, but that this expression is reduced in the cisterna magna and even more so in the serum. In addition, MARs and lack of CpG sequences did not abolish, but may have limited, expression loss and/or increased expression. Finally we found that CSF contains little DNAse activity, and this may contribute to the efficacy of i.t. naked DNA gene delivery.

Materials and Methods

Plasmids. pTR-SEAP (7246 bp, 32 RRCpGYY motifs) contains two inverted terminal repeats flanking the expression cassette which consists of a CMV enhancer, chicken -actin promoter, intron, SEAP cDNA with ~400 nucleotides of the natural SEAP 3' UTR, and an SV40 polyadenylation sequence. pTR-hIL10 was constructed by replacing the SEAP cDNA in pTR-SEAP with the human IL-10 open reading frame. CpGvitro-neo-LacZ (Invivogen, San Diego, CA) entirely lacks CpG dinucleotides. pCpGfree-SEAP (6976 bp) consists of the SEAP open reading frame from pTR-SEAP (without the 3' UTR portion) inserted between the NcoI and NheI sites of pCpGvitro-neo-LacZ, replacing LacZ. The pCpGfree-hIL10 and pCpGfree-GFP plasmids were constructed in the same manner. The pCpGfree-SEAP, pCpGfree-hIL10, and pCpGfree-GFP plasmids contain 6, 1, and 1 RRCpGYY motifs, respectively, all of which are found in their respective transgenes. The plasmids were purified using the Qiagen endofree plasmid giga kit (Qiagen, Valencia, CA) and resuspended in PBS containing 3% sucrose. Most of the plasmid preparations were run through a single column; however, at times two columns were used so as to remove excessive lipopolysacharide (LPS), using a Qiagen protocol (http://www1.qiagen.com/literature/protocols/pdf/QP12.pdf). Four plasmid preps (pTR-SEAP, 19 5 pg LPS/100 g DNA, 6 g/l; pCpGfree-SEAP, 49 6 pg LPS/100 g DNA 6 g/l; pCpGfree-hIL10, <8 pg LPS/100 g DNA, 8.3 g/l) were used in all the experiments described in this paper, except those presented in Figure 6, for which a pTR-SEAP preparation containing ~2 ng LPS/100 g DNA (6 g/l) was used. LPS content was quantified with HEK-blue-4 cells and an LPS standard curve using ultrapure E. coli K-12 LPS (Invivogen, San Diego, CA). Using the endpoint chromogenic Limulus Amebocyte Lysate test, QCL-1000 (Cambrex, Walkersville, MD), and by fitting data points from 4 LPS dilutions (in triplicate) with a line through the origin, we found that the K-12 LPS contained 2.9 EU/ng.

Animals. Adult male Sprague-Dawley rats (Harlan Labs, Madison, WI), weighing 325–350 g at the time of arrival at the experiment site, were used in all the experiments. They were kept on a 12-hour light/dark cycle with food and water provided ad libitum. All experimental procedures on the rats were approved by the Institutional Animal Care and Use Committee at the University of Colorado at Boulder.

Intrathecal injection, CSF and serum collection. Injections were administered as reported earlier⁴⁹ with the following modifications: 4 l of 0.9% saline and then up to 21 l of plasmid DNA were drawn into the catheter, which was then threaded into the i.t. space. The injectant was delivered in <5 seconds, and there was a wait of 5 seconds prior to catheter removal. Lumbosacral CSF collection at the site of DNA delivery was carried out in the same way except that, instead of delivering injectant, ~20 l of CSF was drawn (with minimal pressure) through a syringe into the PE-10 catheter. However, in most instances, lumbosacral CSF was collected ~3–3.5 cm caudal to the DNA delivery site by inserting an 18 gauge needle (with hub removed) between vertebrae L4 and L5, or between L5 and L6. Lumbosacral CSF samples drawn at time points close (<7 days) to a previous lumbar puncture were collected using the catheter method, because natural outflow was diminished or absent at these time points. Cisterna magna CSF was collected by inserting a 26-gauge 3/8 inch needle (without syringe) at midline just rostral to vertebra C1. As the needle was inserted past the dura mater, CSF would spontaneously flow into the needle hub. If it did not do so, no further attempt was made to extract the CSF. All CSF was spun (<2 minutes) in a minicentrifuge # 05-090-128 (Fisher Scientific, Pittsburgh, PA), and the supernatant was transferred to a new tube and placed on dry ice. Almost all cisterna magna CSF and most of the lumbosacral CSF samples were completely clear or very light pink prior to spinning, indicating very little contamination with blood. Blood was collected through a tail nick into a microvette CB 300 capillary tube #16.440 (Sarstedt, Nümbrecht, Germany), kept at room temperature for at least 20 minutes, and then spun for 5 minutes.

Repeated CSF collection. Repeated CSF collection had no apparent impact on rat health. At day 116, the average weight of 22 of the rats used in the experiment described in Figure 2 was 440 3 g (mean value SEM), which was not significantly different from the average weight of two age-matched naive rats (439 9 g, mean value SEM) kept in the same facility for the same duration of time. Furthermore, despite the earlier administration of two DNA injections, the concentration of IL-1 in the lumbosacral CSF was undetectable (<0.3 pmol/l, n = 16) at 56 days after the time point at which the last of 4 CSF collections was made, and no higher than that in the CSF of naive rats (0.8 0.25 pmol/l, mean value SEM, n = 4).

SEAP distribution in CSF. The samples used in the analysis of the distribution of SEAP expression between the DNA delivery site and 3.5 cm caudal to that site (Figure 6) came from experiments designed to test the effect of prior immune stimulation on transgene expression (Hughes et al., manuscript in preparation). Therefore, two days prior to injecting pTR-SEAP, these rats were either injected with noncoding plasmid, or vehicle, or an immunostimulatory oligonucleotide, or a control oligonucleotide, or LPS, or were left uninjected. Some of the rats had chronic sciatic nerve constriction injury. None of these factors significantly affected distribution. Samples in which SEAP concentrations from both sites were less than three standard deviations above their respective average preinjection or control values were excluded from this analysis. SEAP protein half-life in lumbosacral CSF was calculated using WinNonLin (Pharsight, Mountain View, CA) with a two-compartment i.v. bolus model for distribution.

Measurement of DNAse activity in CSF. CSF was harvested and placed on ice, and contaminating serum was quantified by counting red blood cells in the CSF. The CSF samples were then centrifuged, and 50 l was transferred to a 96-well tissue culture plate. Plasmid DNA was added at a final concentration of 400 g/ml, which is a concentration similar to one that could be expected 30 minutes after a 100 g injection in vivo (see later text). All the samples were incubated at room temperature for 12 minutes, then at 37 °C for 18 minutes, and purified over a PCR purification column (Qiagen, Valencia, CA) and run on an agarose gel. A lower bound for plasmid concentration in CSF 30 minutes after the injection was estimated on the assumption that SEAP protein occupies at least 0.131 ml at 30 minutes after the injection (25 ng/191 ng/ml, Figure 5). It is assumed the plasmid (>3 MDa) diffusion rate is less than that of SEAP (53 kDa). Therefore, at 30 minutes after the injection the plasmid occupies 0.131 ml at most, and has a concentration of at least 764 g/ml (100 g/0.131 ml).

SEAP, hIL-10, and rat IL-1 quantification. SEAP activity in CSF and serum was quantified using a chemiluminescent assay # 11779842001 (Roche Applied Science, Indianapolis, IN) in accordance with the manufacturer's instructions, with the following exceptions: 45 l of dilution buffer was added to each well of a white, flat-bottom 96-well plate. Sample (15 l) and standard were then added, and the plate was sealed and floated in a 65 °C water bath for 30 minutes. Thereafter, the rest of the instructions were followed and luminescence was quantified on a Dynex luminometer (Chantilly, VA). Human IL-10 and rat IL-1 were quantified using colorimetric sandwich ELISAs (R&D systems, Minneapolis, MN).

Statistical analysis. All errors reported in the graphs and in the text refer to standard errors. All statistical analysis was carried out using SPSS v16 for Windows (SPSS, Chicago, IL). Fitting lines to the data was carried out on Excel (Microsoft, Redmond, WA). SEAP expression was clearly heteroscedastic, with a significant correlation between the mean value of the expression 14 days after the injection and the standard deviation (stdev = 0.68 mean, R² = 0.95, n = 18 sample groups, 4–8 rats per group). The data were assumed to fit a gamma distribution. Where negative values of either SEAP or hIL-10 concentration were obtained, the negative value (x) was made positive by adding the absolute value of x + 0.01 to all values obtained from the rats in that group and to the rats from all the comparison groups. The largest negative values for SEAP and hIL-10 were -0.6 and -0.7 pmol/l, respectively. Comparison between groups over a time course of length two or greater was achieved using a GEE with a log link. Results obtained from samples from the same rat across time were assumed to be interdependent, and this dependency was modeled as a first order autoregressive (AR1) relationship. All other data was analyzed using a GLZ, with a log link. Post hoc comparisons were carried out using estimated marginal means with sequential Bonferroni correction. Results were considered significantly different at the 5% level, and only two-sided hypotheses were tested.

Supplementary Material
Table S1. Absorbance data for the anti-SEAP antibody ELISA.
Materials and Methods.

References

REFERENCES

1. Persidsky, Y, Ramirez, SH, Haorah, J and Kanmogne, GD (2006). Blood-brain barrier: structural components and function under physiologic and pathologic conditions. J Neuroimmune Pharmacol 1: 223–236. | Article | PubMed |
2. Follett, KA and Naumann, CP (2000). A prospective study of catheter-related complications of intrathecal drug delivery systems. J Pain Symptom Manage 19: 209–215. | Article | PubMed | ChemPort |
3. Shi, L, Tang, GP, Gao, SJ, Ma, YX, Liu, BH, Li, Y et al. (2003). Repeated intrathecal administration of plasmid DNA complexed with polyethylene glycol-grafted polyethylenimine led to prolonged transgene expression in the spinal cord. Gene Ther 10: 1179–1188. | Article | PubMed | ISI | ChemPort |
4. Li, Y, Wang, J, Lee, CG, Wang, CY, Gao, SJ, Tang, GP et al. (2004). CNS gene transfer mediated by a novel controlled release system based on DNA complexes of degradable polycation PPE-EA: a comparison with polyethylenimine/DNA complexes. Gene Ther 11: 109–114. | Article | PubMed | ChemPort |
5. Meuli-Simmen, C, Liu, Y, Yeo, TT, Liggitt, D, Tu, G, Yang, T et al. (1999). Gene expression along the cerebral-spinal axis after regional gene delivery. Hum Gene Ther 10: 2689–2700. | Article | PubMed | ChemPort |
6. Abbott, NJ (2004). Evidence for bulk flow of brain interstitial fluid: significance for physiology and pathology. Neurochem Int 45: 545–552. | Article | PubMed | ISI | ChemPort |
7. Oshima, K, Shimamura, M, Mizuno, S, Tamai, K, Doi, K, Morishita, R et al. (2004). Intrathecal injection of HVJ-E containing HGF gene to cerebrospinal fluid can prevent and ameliorate hearing impairment in rats. Faseb J 18: 212–214. | PubMed | ChemPort |
8. Milligan, ED, Langer, SJ, Sloane, EM, He, L, Wieseler-Frank, J, O'Connor, K et al. (2005). Controlling pathological pain by adenovirally driven spinal production of the anti-inflammatory cytokine, interleukin-10. Eur J Neurosci 21: 2136–2148. | Article | PubMed |
9. Butti, E, Bergami, A, Recchia, A, Brambilla, E, Del Carro, U, Amadio, S et al. (2008). IL4 gene delivery to the CNS recruits regulatory T cells and induces clinical recovery in mouse models of multiple sclerosis. Gene Ther 15: 504–515. | Article | PubMed | ChemPort |
10. Kramm, CM, Rainov, NG, Sena-Esteves, M, Barnett, FH, Chase, M, Herrlinger, U et al. (1996). Long-term survival in a rodent model of disseminated brain tumors by combined intrathecal delivery of herpes vectors and ganciclovir treatment. Hum Gene Ther 7: 1989–1994. | Article | PubMed | ISI | ChemPort |
11. Chou, AK, Yang, LC, Wu, PC, Wong, WT, Liu, GS, Chen, JT et al. (2005). Intrathecal gene delivery of glial cell line-derived neurotrophic factor ameliorated paraplegia in rats after spinal ischemia. Brain Res Mol Brain Res 133: 198–207. | Article | PubMed | ChemPort |
12. Thomas, CE, Ehrhardt, A and Kay, MA (2003). Progress and problems with the use of viral vectors for gene therapy. Nat Rev Genet 4: 346–358. | Article | PubMed | ISI | ChemPort |
13. Wang, X, Wang, C, Zeng, J, Xu, X, Hwang, PY, Yee, WC et al. (2005). Gene transfer to dorsal root ganglia by intrathecal injection: effects on regeneration of peripheral nerves. Mol Ther 12: 314–320. | Article | PubMed | ISI | ChemPort |
14. Wu, CM, Lin, MW, Cheng, JT, Wang, YM, Huang, YW, Sun, WZ et al. (2004). Regulated, electroporation-mediated delivery of pro-opiomelanocortin gene suppresses chronic constriction injury-induced neuropathic pain in rats. Gene Ther 11: 933–940. | Article | PubMed | ISI | ChemPort |
15. Yao, MZ, Gu, JF, Wang, JH, Sun, LY, Lang, MF, Liu, J et al. (2002). Interleukin-2 gene therapy of chronic neuropathic pain. Neuroscience 112: 409–416. | Article | PubMed | ChemPort |
16. Lee, TH, Yang, LC, Chou, AK, Wu, PC, Lin, CR, Wang, CH et al. (2003). In vivo electroporation of proopiomelanocortin induces analgesia in a formalin-injection pain model in rats. Pain 104: 159–167. | Article | PubMed | ChemPort |
17. Chen, KH, Wu, CH, Tseng, CC, Shiau, JM, Lee, CT and Lin, CR (2008). Intrathecal coelectrotransfer of a tetracycline-inducible, three-plasmid-based system to achieve tightly regulated antinociceptive gene therapy for mononeuropathic rats. J Gene Med 10: 208–216. | Article | PubMed | ChemPort |
18. Lin, CR, Yang, LC, Lee, TH, Lee, CT, Huang, HT, Sun, WZ et al. (2002). Electroporation-mediated pain-killer gene therapy for mononeuropathic rats. Gene Ther 9: 1247–1253. | Article | PubMed | ChemPort |
19. Yao, MZ, Wang, JH, Gu, JF, Sun, LY, Liu, H, Zhao, ZQ et al. (2002). Interleukin-2 gene has superior antinociceptive effects when delivered intrathecally. Neuroreport 13: 791–794. | Article | PubMed | ChemPort |
20. Milligan, ED, Sloane, EM, Langer, SJ, Hughes, TS, Jekich, BM, Frank, MG et al. (2006). Repeated intrathecal injections of plasmid DNA encoding interleukin-10 produce prolonged reversal of neuropathic pain. Pain 126: 294–308. | Article | PubMed | ChemPort |
21. Zhang, G, Budker, V, Williams, P, Subbotin, V and Wolff, JA (2001). Efficient expression of naked dna delivered intraarterially to limb muscles of nonhuman primates. Hum Gene Ther 12: 427–438. | Article | PubMed | ISI | ChemPort |
22. Miao, CH, Ohashi, K, Patijn, GA, Meuse, L, Ye, X, Thompson, AR et al. (2000). Inclusion of the hepatic locus control region, an intron, and untranslated region increases and stabilizes hepatic factor IX gene expression in vivo but not in vitro. Mol Ther 1: 522–532. | Article | PubMed | ISI | ChemPort |
23. Alino, SF, Crespo, A and Dasi, F (2003). Long-term therapeutic levels of human alpha-1 antitrypsin in plasma after hydrodynamic injection of nonviral DNA. Gene Ther 10: 1672–1679. | Article | PubMed | ISI | ChemPort |
24. Takahashi, M, Kido, K, Aoi, A, Furukawa, H, Ono, M and Kodama, T (2007). Spinal gene transfer using ultrasound and microbubbles. J Control Release 117: 267–272. | Article | PubMed | ChemPort |
25. Shimamura, M, Sato, N, Taniyama, Y, Yamamoto, S, Endoh, M, Kurinami, H et al. (2004). Development of efficient plasmid DNA transfer into adult rat central nervous system using microbubble-enhanced ultrasound. Gene Ther 11: 1532–1539. | Article | PubMed | ChemPort |
26. Fedorova, E, Battini, L, Prakash-Cheng, A, Marras, D and Gusella, GL (2006). Lentiviral gene delivery to CNS by spinal intrathecal administration to neonatal mice. J Gene Med 8: 414–424. | Article | PubMed | ChemPort |
27. Jiao, S, Williams, P, Berg, RK, Hodgeman, BA, Liu, L, Repetto, G et al. (1992). Direct gene transfer into nonhuman primate myofibers in vivo. Hum Gene Ther 3: 21–33. | Article | PubMed | ISI | ChemPort |
28. Edelstein, ML, Abedi, MR and Wixon, J (2007). Gene therapy clinical trials worldwide to 2007--an update. J Gene Med 9: 833–842. | Article | PubMed |
29. The Journal of Gene Medicine Clinical Trial site. <http://www.wiley.co.uk/genetherapy/clinical/>.
30. Wang, CY and Wang, S (2006). Astrocytic expression of transgene in the rat brain mediated by baculovirus vectors containing an astrocyte-specific promoter. Gene Ther 13: 1447–1456. | Article | PubMed | ChemPort |
31. Hsiao, CD, Hsieh, FJ and Tsai, HJ (2001). Enhanced expression and stable transmission of transgenes flanked by inverted terminal repeats from adeno-associated virus in zebrafish. Dev Dyn 220: 323–336. | Article | PubMed | ChemPort |
32. Conese, M, Auriche, C and Ascenzioni, F (2004). Gene therapy progress and prospects: episomally maintained self-replicating systems. Gene Ther 11: 1735–1741. | Article | PubMed | ChemPort |
33. Zelenay, S, Elias, F and Flo, J (2003). Immunostimulatory effects of plasmid DNA and synthetic oligodeoxynucleotides. Eur J Immunol 33: 1382–1392. | Article | PubMed | ChemPort |
34. Hodges, BL, Taylor, KM, Joseph, MF, Bourgeois, SA and Scheule, RK (2004). Long-term transgene expression from plasmid DNA gene therapy vectors is negatively affected by CpG dinucleotides. Mol Ther 10: 269–278. | Article | PubMed | ISI | ChemPort |
35. Jenke, AC, Scinteie, MF, Stehle, IM and Lipps, HJ (2004). Expression of a transgene encoded on a non-viral episomal vector is not subject to epigenetic silencing by cytosine methylation. Mol Biol Rep 31: 85–90. | Article | PubMed | ISI | ChemPort |
36. Chen, ZY, Riu, E, He, CY, Xu, H and Kay, MA (2008). Silencing of episomal transgene expression in liver by plasmid bacterial backbone DNA is independent of CpG methylation. Mol Ther 16: 548–556. | Article | PubMed | ChemPort |
37. Chastain, M, Simon, AJ, Soper, KA, Holder, DJ, Montgomery, DL, Sagar, SL et al. (2001). Antigen levels and antibody titers after DNA vaccination. J Pharm Sci 90: 474–484. | Article | PubMed | ISI | ChemPort |
38. Glasspool-Malone, J and Malone, RW (1999). Marked enhancement of direct respiratory tissue transfection by aurintricarboxylic acid. Hum Gene Ther 10: 1703–1713. | Article | PubMed | ChemPort |
39. Glasspool-Malone, J, Somiari, S, Drabick, JJ and Malone, RW (2000). Efficient nonviral cutaneous transfection. Mol Ther 2: 140–146. | Article | PubMed | ISI | ChemPort |
40. Niedzinski, EJ, Chen, YJ, Olson, DC, Parker, EA, Park, H, Udove, JA et al. (2003). Enhanced systemic transgene expression after nonviral salivary gland transfection using a novel endonuclease inhibitor/DNA formulation. Gene Ther 10: 2133–2138. | Article | PubMed | ChemPort |
41. Elouahabi, A and Ruysschaert, JM (2005). Formation and intracellular trafficking of lipoplexes and polyplexes. Mol Ther 11: 336–347. | Article | PubMed | ChemPort |
42. Tan, Y, Liu, F, Li, Z, Li, S and Huang, L (2001). Sequential injection of cationic liposome and plasmid DNA effectively transfects the lung with minimal inflammatory toxicity. Mol Ther 3: 673–682. | Article | PubMed | ChemPort |
43. Whitesell, L, Geselowitz, D, Chavany, C, Fahmy, B, Walbridge, S, Alger, JR et al. (1993). Stability, clearance, and disposition of intraventricularly administered oligodeoxynucleotides: implications for therapeutic application within the central nervous system. Proc Natl Acad Sci USA 90: 4665–4669. | Article | PubMed | ChemPort |
44. Sozzani, S, Ghezzi, S, Iannolo, G, Luini, W, Borsatti, A, Polentarutti, N et al. (1998). Interleukin 10 increases CCR5 expression and HIV infection in human monocytes. J Exp Med 187: 439–444. | Article | PubMed | ISI | ChemPort |
45. Melder, RJ, Osborn, BL, Riccobene, T, Kanakaraj, P, Wei, P, Chen, G et al. (2005). Pharmacokinetics and in vitro and in vivo anti-tumor response of an interleukin-2- human serum albumin fusion protein in mice. Cancer Immunol Immunother 54: 535–547. | Article | PubMed | ChemPort |
46. Dickason, RR and Huston, DP (1996). Creation of a biologically active interleukin-5 monomer. Nature 379: 652–655. | Article | PubMed | ChemPort |
47. Loser, P, Jennings, GS, Strauss, M and Sandig, V (1998). Reactivation of the previously silenced cytomegalovirus major immediate-early promoter in the mouse liver: involvement of NFB. J Virol 72: 180–190. | PubMed | ISI | ChemPort |
48. Ehrhardt, A, Peng, PD, Xu, H, Meuse, L and Kay, MA (2003). Optimization of cis-acting elements for gene expression from nonviral vectors in vivo. Hum Gene Ther 14: 215–225. | Article | PubMed | ChemPort |
49. Milligan, ED, Sloane, EM, Langer, SJ, Cruz, PE, Chacur, M, Spataro, L et al. (2005). Controlling neuropathic pain by adeno-associated virus driven production of the anti-inflammatory cytokine, interleukin-10. Mol Pain 1: 9. | Article | PubMed | ChemPort |

Acknowledgments

This work was funded by NIH grants DA018156, DA056542, R0150560, and Avigen Inc.