|Home | About | Journals | Submit | Contact Us | Français|
The completion of the human and mouse genomes has identified at least 20 connexin isomers in this family of intercellular channel proteins. However, there are no specific gap junction blockers or channel-blocking mimetic peptides available for the study of specific connexins. We designed antisense oligodeoxynucleotides that functionally reduce targeted connexin protein expression and can be used to reveal the biological function of individual connexins in vivo. Connexin mRNA was firstly exposed in vitro to deoxyribozymes complementing the sense coding sequence. Those that cleaved the target connexin mRNA in defined regions were used as the basis to design oligodeoxynucleotides to the accessible sites, thus taking into account tertiary mRNA configurations rather than relying on computed predictions. Antisense oligodeoxynucleotides designed to bind to accessible mRNA sites selectively reduced connexin26 and −43 mRNA expression in a corneal epithelium ex vivo model. Connexin43 protein levels were reduced correlating with the knockdown in mRNA and the protein’s rapid turnover; protein levels of connexin26 did not alter, supporting lower turnover rates reported for that protein. We show, for the first time, an inexpensive and empirical approach to the preparation of specific and functional antisense oligodeoxynucleotides against known gene targets in the post-genomic era.
Gap junction channels are transmembrane channels that couple cells both electrically and metabolically. Each hemi-channel consists of six connexin proteins forming a hexameric connexon that is inserted into the membrane of the cell. Each connexon docks with a connexon in the opposing membrane to form a single gap junction channel.1 Gap junctions can be found throughout the body. A tissue such as the corneal epithelium, for example, has six to eight cell layers, yet expresses different gap junction channels in different layers, with connexin43 in the basal layer and connexin26 from the basal to middle wing cell layers.2 Sequencing the human and mouse genome has identified at least 20 connexin isomers in this family of intercellular channel proteins.1 However, it has proved difficult to dissect out the function of individual gap junction proteins due to a lack of specific channel-closing chemicals, connexinspecific channel-blocking mimetic peptides, or antibodies able to access extracellular protein domains.
Antisense oligonucleotides represent a potentially powerful and specific molecular biology technique to knock down protein expression and thus study the function of genes.3 The ability of short synthetic antisense oligonucleotides to block protein translation was first demonstrated 25 years ago by Paterson and colleagues in a cell-free system,4 and a year later in cultures.5 The first published account of connexin43 antisense oligonucleotides affecting channel conductance was not until 1994, in an embryonic rat aortic cell line.6 These authors reported functional reduction of connexin43-related channel conductance, but could not demonstrate any effect at the protein or mRNA levels. A subsequent paper7 reported reduced dye coupling and protein levels in BALB/c 3T3 cells treated with phos-phothrioate connexin43 antisense oligonucleotides. Since then, there have only been two connexin43 antisense oligonucleotide sequences in the published scientific literature. The most frequently used is an antisense sequence covering the start codon.6–11 A separate connexin43-specific antisense oligonucleotide, DB1, has significant biological effects on the developing chicken embryo in ovo12–15 and in tissue response to damage.16–19
Antisense technology, however, is still not widely accepted due at least in part to difficulties in identification of functional antisense sequences and problems with delivery of antisense oligonucleotides into cells. In this study, we address the issue of antisense design. We propose that the folding and tertiary structure of mRNA may impede the binding of some antisense oligonucleotides to the target sequence, and have designed an autocatalytic deoxyribozyme library specifically against the connexin43 and −26 mRNAs to map out regions accessible to antisense oligonucleotides. We have tested the ability of these antisense sequences to reduce mRNA levels and reduce protein translation in whole eye corneal organ cultures, and demonstrated that our protocol can differentiate between antisense sequences that have functional capabilities and those that are likely to be non-functional in vivo.
Rat connexin43 plasmid, T7291, was a generous gift from N. Kumar (University of Illinois at Chicago), and rat connexin26 plasmid was from S. Casalotti (University College London). Mouse connexin43 and connexin26 plasmids were purchased from Invivogen (San Diego, CA). Mouse anti-rat connexin43 antibody was our own,20 and rabbit anti-rat connexin26 antibody was obtained from Zymed (51–2800; South San Francisco, CA). Goat anti-mouse Alexa488 and goat anti-rabbit Alexa 568 secondary antibodies were obtained from Molecular Probes (Eugene OR). Nuclei were stained using Hoechst 33258 dye (Sigma). All deoxyribozymes and oligodeoxynucleotides were purchased from Sigma Genosys, Australia, as desalted oligonucleotides. TaqMan labeled oligonucleotides were purchased from Applied Biosystems (Foster City, CA). All oligodeoxynucleotides were purchased as unmodified phosphodiester oligodeoxynucleotides.
Deoxyribozyme design and testing was similar to that described in previous studies.21,22 In brief, all AU and GU sites in the mRNA sequence of the target connexin were selected with eight or nine nucleotides on each side of the A or G. The deoxyribozymes are the complement of this sense coding sequence with the A or G replaced with the 10–23 catalytic core ggctagctacaacga. Control deoxyribozymes had a defective catalytic core of ggctaActacaacga, with a single-point mutation (g→A).23 We also designed GC- and AC-specific deoxyribozymes to cover gaps left by AU and GU deoxyribozymes not meeting the three requirements below. Each deoxyribozyme was named according to the position of A or G nucleotides from the start ATG codon. Those deoxyribozymes selected for in vitro assay had to fulfill three requirements:
Mouse connexin43 and connexin26 cDNAs were excised from the pORF vector (Invivogen) using NcoI and NheI, and subcloned into pGEM-T (Promega) prior to in vitro transcription. Both the full-length 2.4-kb rat connexin43 cDNA and the full coding 1.4-kb rat connexin43 cDNA including 200 nucleotides of 5′-untranslated regions were used for in vitro transcription. Full-length mRNA was transcribed from linearized plasmid DNA using a Promega Riboprobe Kit. The resulting mRNA was purified with a PCR spin column (Qiagen). Concentration was determined by spectrophotometer reading of OD at λ = 260 nm. Deoxyribozymes (40 μM final concentration) and mRNA (0.01 to 0.05 μg/μL total mRNA) were then separately pre-equilibrated with a 2X cleavage buffer (100 mM Tris 7.5; 20 mM MgCl2; 300 mM NaCl; 0.02% SDS) for 5–10 min at 37°C. mRNA and deoxyribozyme mix were then incubated for 1 h at 37°C, following which 10X Bluejuice (Invitrogen) was added to stop the cleavage reaction and the mixture kept on ice. The reaction mixture was then loaded onto a pre-run 4% polyacrylamide gel (19:1 acryl: bis ratio, BioRad) in 1X TBE buffer and 7 M urea, and run for up to 2 h. Gels were stained with a 1:10,000 dilution of SYBR green II (Molecular Probes, Invitrogen) in TBE buffer and imaged using a BioRad Chemi Doc system.
Antisense sequences were chosen based on the nucleotide sequences of the deoxyribozyme binding arms that were successful in cleaving the mRNA in vitro. Selected sequences were chosen for use in the design of 30-mer oligos.24,25 In brief, sequence-related side effects such as partial sequence homology of 8–10 CG base pairings to unrelated genes, GGGG, and CpG motifs were avoided. Antisense sequences with the 3′-end ending with a thymidine or more than three C or Gs in the last five nucleotides are also avoided if possible to prevent miss-priming. Oligonucleotides that form stable secondary structures such as homodimers, palindrome motifs, or secondary hairpin structures will impede oligonucleotides binding to the target mRNA. Control oligonucleotides, including sense, scrambled, reverse, and mismatch oligonucleotides, were also designed to assess possible chemistry-related side effects due to cross-hybridization, nonspecific protein binding, and toxicity.24
Approval for all animal work was obtained from the University of Auckland Ethics Committee. Thirty to 34-dold Wistar rats were euthanized with carbon dioxide and whole rat eyes dissected. The ocular surface was dissected, disinfected with 0.1 mg/mL penicillin-streptomycin for 5 min and rinsed in sterile PBS. The whole eye was then transferred onto a sterile holder in a 60-mm culture dish with the cornea facing up. The eyes were mounted with the corneal epithelium exposed at the air-medium interface and cultured at 34°C in a humidified 5% CO2 incubator in serum-free medium (Opti-MEM, Invitrogen) for up to 48 h. One-hundred microliters of medium was added drop wise to the surface every 8 to 12 h to moisten the epithelium. Medium levels were maintained to the level of the limbal conjunctiva.
Antisense oligonucleotides were mixed with 30% (w/v) Pluronic F127 gel (Sigma) on ice to a final 2 μM concentration, and 10 μL applied onto the corneas as previously described.18,26 Each treatment had a sample size of three to four corneas per experiment. Preliminary experiments showed that double treatments of our positive control, DB1, for 8 h had little effect on connexin43 protein expression in our corneal culture. Corneas were therefore cultured for 24 h and connexin43-specific oligonucleotides applied every 8 h. However, we found that endogenous connexin26 expression is affected if the culture was maintained for 24 h. Hence, we reduced the culture period for corneas treated with connexin26-specific oligonucleotides to 12 h, with application of antisense oligonucleotides every 4 h. Medium was changed 10 min prior to every repeat application of antisense or control oligonucleotides. At defined times, corneas were rinsed with PBS, immersed in OCT (Tissue Tek, Japan), and snap-frozen in liquid nitrogen. Twenty-five-μm cryosections were subsequently cut with a Leica cryostat (CM3050s) and mounted on SuperFrost Plus slides (Menzel, Germany). For analysis of both Cx43 and Cx26 mRNA levels, corneas were collected 8 h after a single antisense treatment.
Total RNA was extracted from isolated rat corneas using TRIzol reagent (GIBCO, Invitrogen) according to the manufacturer’s protocols. The quality of RNA samples was assessed by electrophoresis through ethidium bromide–stained agarose gels and the 18S and 28S rRNA bands visualized under UV illumination. The extraction yield was quantified spectrophotometrically at λ = 260 nm. For real-time PCR, cDNA was prepared from 5 μg of total RNA by using oligo dT and superscript II Rnase H-reverse transcriptase (Life Technologies, Invitrogen) in a final reaction volume of 20 μL. Quantitative PCR reaction was carried out in 96-well optical reaction plates using a cDNA equivalent of 100 ng total RNA for each sample in a volume of 50 μL using the TaqMan Universal PCR Master Mix (Applied Biosystems) according to the manufacturer’s instructions. PCR was developed on the ABI Prism 7700 Sequence Detection system instrument (Applied Biosystems). The thermal cycling conditions comprised an initial denaturation step at 95°C for 10 min and 50 cycles of two-step PCR, including 15 sec of denaturation at 95°C and 1 min of annealing-elongation at 60°C, using the standard protocol of the manufacturer. All experiments were repeated in triplicate. The monitoring of negative control for each target showed an absence of carryover.
To minimize the errors arising from the variation in the amount of starting RNA among samples, amplification of 18S rRNA was performed as an internal reference against which other RNA values can be normalized. If the efficiencies of the target and 18S rRNA amplifications were approximately equal, then the formula 2−ΔΔCt was used to calculate relative levels of mRNA without the need for a standard curve. If the efficiency of amplification of the target and 18S rRNA were significantly different, a relative standard curve method was used to calculate absolute quantities of mRNA and 18S rRNA for each experiment from the measured Ct, and then the relative mRNA levels of the target gene compared with control quantified after normalization to 18S rRNA.
All calculations were performed with Prism 3.02 software (GraphPad, San Diego, CA). Statistical difference between groups was determined by using the Student’s t-test. Comparisons among several groups were performed by ANOVA, and significance was calculated using Dunnett’s multiple comparison test.
Cy3 and TaqMan (Fam, Tamra) labelled oligonucleotides were used to assess penetration and stability. Cy3-labeled oligonucleotides (Sigma Genosys) and TaqMan (FAM, TAMRA) labeled oligonucleotides (Applied Biosystems) were applied with Pluronic F-127 gel to measure both the stability and the penetration of oligonucleotides into the corneal epithelium. The treated corneas were fixed in 4% paraformaldehyde for 20 min, mounted in 1% agar, and viewed under a 40× water immersion lens as whole mount. The depth of oligonucleotide penetration was measured using the Z-scan option on a Leica SP2 confocal microscope, and plots of intensity versus z-distance measured. The breakdown of TaqMan oligonucleotides was measured using the Lamdba scan option on the confocal. Fluorescence resonance energy transfer (FRET) between the FAM (donor) and TAMRA (receptor) molecule occurs in intact 30mer oligonucleotides. When the oligonucleotide is broken down, FAM and TAMRA are no longer in close proximity and FRET no longer occurs. For details, see reference 27 .
Immunolabeling of connexins on corneal sections was performed as previously described.2 In brief, sections were blocked in 10% goat serum and incubated with primary antibody at 1:250 (mouse anti-rat connexin43) or 1:500 (rabbit anti-rat connexin26) at 4°C overnight. The sections were then washed with PBS, incubated with 1:400 dilution of Alexa488-labeled secondary antibody at room temperature for 2 h, and then fixed in 4% paraformaldehyde and counterstained with 0.2 μM propidium iodide or a 1:50 dilution of Hoechst 33258 for 10 min. Sections were mounted in Citifluor antifade medium (Agarscientific, UK). All images were collected using either a Leica TCS-4D or Leica SP2 confocal laser scanning microscope and stored as TIF files. All images were collected using consistent voltage (520–540 V) and offset (−2) settings. The voltage and offset were set using the glow-over-under display option to maximize the gray scale for images of control tissue. The same settings were then used for all samples within the same experiment.
For quantification, four optical slices through three micrometers were processed into a single extended-focus optical image by using the center-of-mass topographic projection option on the TCS-4D. Spots of connexin label were counted using NIH ImageJ after thresholding at 90- to 100-pixel intensity on the 256 grayscale image. The area of corneal epithelium was also measured and a connexin density per unit area was calculated. An average of four extended-focus images were used to calculate the absolute connexin density of each cornea. This number was then normalized with the medium connexin density of either sense control–treated or gel-treated corneas. We have represented the data as percentage knockdown when comparing different treatments.
Sixty-six deoxyribozymes were designed specifically against rodent connexin43 mRNA. Twenty-two of these deoxyribozymes were designed to recognize both mouse and rat connexin43 mRNA. We also purchased two defective deoxyribozymes with a single-point mutation in the 10–23 catalytic core as negative controls. The deoxyribozyme cleavage results were similar for the rat connexin43 mRNA 1.1 Kb in length (not shown) and the rat connexin43 mRNA 2.4 Kb in length (Figure 1A1A).). Both rat (Figure 1A1A)) and mouse (Figure 1B1B)) connexin43 mRNA appear to have similar regions accessible to the deoxyribozymes. The results indicate four regions on the rodent connexin43 mRNA that are exposed and available for deoxyribozyme cleavage. These regions are around 367–466, 526–622, 783–885, and 1007– 1076 bases from the start ATG codon. The two defective deoxyribozymes, a1df605 and a1df783, showed no cleavage of rodent connexin43 mRNA. Deoxyribozymes designed against the 200-bp 5′ untranslated region of rat connexin43 mRNA also did not show any cleavage activity (data not shown).
We tested 44 deoxyribozymes designed specifically against rodent connexin26 mRNA, of which 17 match both mouse and rat connexin26 mRNA. The rat connexin26 mRNA appeared as a double band on the gel, owing to the presence of two T7 RNA polymerase promoters on the cloning plasmid (personal communication, Stefano Casolotti, UCL). The cleavage results show that connexin26 mRNA has at least two regions accessible to deoxyribozymes, in the 318–379 and 493–567 base regions (Figure 2A, 2B2B).). These figures show that most consistently cleaving deoxyribozyme is the cx26dz330, which cleaves both species of mRNA within 1 h. The two defective deoxyribozymes (b2df351 and b2df379) showed no cleavage of rodent connexin26 mRNA. The deoxyribozymes cx26dz341, dz351, dz375, and dz379 consistently cleave rat connexin26 mRNA at a higher rate compared to mouse connexin26 mRNA. On the other hand, mcx26dz153 and dz567 appear to be superior connexin26 deoxyribozymes in mouse when compared to rat.
Tables 11 and 22 show selected ribozyme sequences, and defective controls, upon which antisense sequences were designed for connexins 43 and 26 respectively. These tables also show the results of in vivo analysis (mRNA and/or protein levels) with those antisense sequences selected on the basis of the in vitro ribozyme cleavage studies.
Rat corneas maintain expression of both connexin43 and connexin26 in organ culture and are easily accessible to the delivery of antisense oligonucleotides by 30% Pluronic F-127 gel. The rat cornea organ culture was therefore selected as the model system to test the effectiveness of the antisense oligodeoxynucleotides designs derived from the in vitro model. We cultured rat corneas for 24 h and found that the endothelium remains intact. Cy3-labeled oligonucleotides were used to determine the extent of penetration of the oligonucleotide into the cultured cornea. Confocal optical slices down through the intact cornea show that fluorescent signal is present with CY3-labeled oligonucleotides (Figure 3A3A shows fluorescent signal 10 μm deep in the cultured cornea, 1 h after initial application). TaqMan probes conjugated to oligodeoxynucleotides were used to measure and demonstrate the delivery of intact oligodeoxynucleotide with 30% Pluronic gel into corneal epithelium. A signficant proportion of oligonucleotide remained intact 1 h after treatment (Figure 3B, 3C3C).). The punctate signal of intact oligonucleotides (FRET occurring in Figure 3C3C)) can be seen as the red wavelength, while signal from degraded oligonucleotides (no FRET) appears in the green emission spectrum (Figure 3B3B).
In a preliminary experiment, we treated rat corneas with a single application of our positive control, DB1, and found no significant changes in connexin43 protein expression after 8 h. Clear protein knockdown at 24 h was seen after three applications at eight-hourly intervals. Based in part on results from the deoxyribozyme cleavage assay, we tested certain antisense oligonucleotides in vivo (DB1, r43as605, r43as783, r43as885, r43as953, and r43as1076), as well as antisense oligonucleotides that were predicted to be non-functional (r43as14, r43as769, and r43as892), and a negative control (DB1 sense). We found knockdown of connexin43 protein levels after 24 h of treatment compared to controls (Figure 4A4A)) with all of the antisense oligonucleotides that we had determined should be positive (Figure 4C, 4E, 4G4G).). All three of those predicted to be negative, and the negative control oligonucleotide, did not affect connexin43 expression (Figure 4B, 4D, 4F, 4H4H).). DB1, a 30-mer version of as885, showed a similar percentage knockdown to the shorter as885 (just under 50% knockdown). One of the best antisense oligonucleotides appeared to be as605, with a 64% reduction in protein level. A summary of these quantified results is presented in Figure 55.
To test the technique for other connexins, further oligodeoxynucleotides were designed and tested for connexin26. Two 30-mer antisense oligodeoxynucleotides, designated as r26as330N and 375N, together with their appropriate reverse control oligodeoxynucleotides, were designed against connexin26 based on regions within the cleavage areas of b2dz330 and b2dz375. However, we found that these antisense oligodeoxynucleotides (as330N and as375N) did not lead to a significant difference in protein expression levels within the 12-h time period for these experiments when antisense oligonucleotide–treated cultures were compared with the reverse control–treated corneas.
Real-time PCR was used to determine the effect of antisense oligodeoxynucleotides on mRNA levels. It confirmed that antisense oligodeoxynucleotides that knock down connexin43 protein expression (as605, as885, DB1) also have lower connexin43 mRNA levels compared to control corneas within 8 h after treatment (Figure 66).). The percentage reduction in relative levels of connexin43 mRNA correlated well with the level of reduction of connexin43 protein. The negative antisense oligonucleotide (as769) and negative controls (DB1 sense, gel only) have unchanged levels of connexin43 mRNA compared to control corneas.
Connexin26 mRNA expression was also significantly reduced by as330N and as375N within 8 h of antisense treatment (Figure 77).). The reverse sequence control for as330N and a gel-only control had no effect on mRNA levels.
There are many published approaches relating to selection of optimal antisense oligodeoxynucleotide (or oligonucleotide) sequences.28 Some of these approaches, however, require significant investment in oligonucleotide synthesizing equipment to make either a scanning array29 or a randomized nucleotide library.30 Also, computer-based secondary structure prediction methods still require experimental validation.31,32
We used a combinational approach in which a selected deoxyribozyme library was generated using computer software predictions, followed by empirical experimentation to validate chosen sequences in vitro and in vivo. We discovered that the folding and tertiary structure of mRNA could impede antisense oligonucleotide binding to some of the target sequences, at least in part accounting for unpredictability in existing antisense design protocols. Using our protocol we demonstrated a strong correlation between those antisense oligonucleotides predicted to work by deoxyribozyme assay and their effect on mRNA levels in organ culture of rat cornea.
This study concentrated on the design of “10–23” deoxyribozymes,21,33 and unmodified phosphodiester antisense oligonucleotides. The “10–23” deoxyribozyme has a small catalytic domain of 15 nucleotides flanked by two-substrate recognition domains, each 7–9 nucleotides long.22 These deoxyribozymes are small in size (only around 30–35 nucleotides in length), have intrinsic catalytic activity (and therefore are not dependant on RNase H), are inexpensive to produce and are more resistant to degradation than RNA based ribozymes. These “10–23” deoxyribozymes cleave any mRNA containing a purine (U or C)-pyrimidine (A or G) junction, with GU and AU preferred, using Mn++, Mg++ or Ca++ cofactors.23
There were similar overall patterns of deoxyribozyme cleavage for both rat and mouse mRNA. We found deoxyribozyme accessible regions around 526–605, 783–885, and 1007–1076 nucleotides for connexin43 mRNA, and 318– 379 nucleotide region for connexin26 mRNA. However, the rate of cleavage for deoxyribozymes varied between the two mRNA species, those such as b2dz375 and b2dz379 show greater than 90% cleavage in rat connexin26 mRNA compared to only 50% cleavage in mouse connexin26 mRNA. Both rat and mouse connexin coding mRNAs have 95% homology, and the 5% difference does not appear to significantly influence the tertiary structures of the connexin mRNA. Some contradictions, such as the better cleavage of m26dz153 and m26dz567 compared to similar rat deoxyribozymes do exist for specific deoxyribozymes, but this appears to happen only within restricted stretches of the mRNA.
Similar levels of deoxyribozyme cleavage were seen in rat connexin43 mRNA transcripts that varied in length from 1100 to 2400 nucleotides long. The 3′ untranslated region of connexin43 mRNA plays a role in half-life and translational efficiency of the connexin mRNA.34,35 However, our results reveal that there is no masking of accessible sites by the tertiary folding of 3′ untranslated regions, or, alternatively, that the interaction is weak and can be easily overcome by the deoxyribozyme. There is less certainty about the effects of 5′- untranslated regions on deoxyribozyme mapping of tertiary structures because the rat connexin43 mRNA used in this study contained only 200 nucleotides of the native translation initiation region. We believe that antisense oligonucleotides targeted against this region may gain access only after binding of the ribosomal initiation complex, hence blocking the translation of connexin proteins. Further variations in the length of 5′- untranslated regions will need to be undertaken to analyze its contribution to the tertiary folding of connexin mRNA. An apparent difference between the accessibility of deoxyribozymes to the 367–466 regions could theoretically be attributed to the 5′- untranslated region. Rat connexin43 mRNA showed less cleavage, as compared with the mouse connexin43 mRNA, perhaps due to the lack of 5′- untranslated region in the mouse construct.
We did not apply deoxyribozymes to our corneal organ cultures because of the sensitivity of unmodified deoxyribozymes to enzymatic degradation. Their short binding arm length renders them functionless in a shorter time period. We chose instead to concentrate on developing functional antisense oligonucleotides based upon the tertiary structure mapping data. All of our connexin43 specific antisense oligonucleotides are 20 nucleotides long except for DB1 and DB1 sense (which are 30-mers). We initially tried in vitro testing of our antisense oligonucleotides in the presence of RNase H, but all of the antisense oligonucleotides appeared to bind to the target mRNA and were cleaved by RNase H. This cleavage appeared specific because sense control oligonucleotides did not show similar results. One possible explanation for this in vitro result is that RNase H may unwind mRNA and lead to cleavage of any DNA-RNA duplets.36 We subsequently tested all our connexin43 antisense oligonucleotides using the corneal organ culture model. Results of connexin43 protein knockdown were found to be specific and agreed with the “accessible site” data mapping obtained using our in vitro deoxyribozyme tertiary mapping protocol.
We have developed ways to improve the efficacy of antisense oligonucleotides in organ systems, overcoming problems associated with antisense oligonucleotides delivery into cell cultures. Three major considerations to consider are the delivery protocol, sequence effects, and target protein turnover. We can overcome antisense oligonucleotide breakdown by utilizing Pluronic gel as a non-specific method for sustained oligonucleotide delivery. Pluronic F-127 gel is a weak surfactant and at 30% (w/v) sets as hydroscopic gel at physiological temperature. This allows a reservoir of oligonucleotides to be continuously released onto the corneas as long as it sits higher than the subconjunctiva above the medium. Pluronic gel dissolves quickly if placed into medium itself. We have shown here that within 1 h after treatment with Pluronic gel and oligonucleotides, intact oligonucleotides had penetrated at least 10 μm into the corneal epithelium.
Synthetic oligonucleotides can cause a number of different and unexpected biological effects inside cells. Carefully designed control oligonucleotides and experiments are helpful to identify the specific effect of target protein knockdown and to separate it from nonspecific side effects. More importantly, we have shown that designing an antisense sequence to match accessible regions of the mRNA is a significant determinant in the preparation of a successful oligonucleotide. We have shown more than one antisense oligonucleotide to be effective in knocking down connexin43. No compensation by the closely related connexin26 protein was seen. We conclude that the biological effects seen with our various antisense oligonucleotides are due to knockdown of specific gap junction channel communication. We have not performed functional dye-transfer studies between epithelial cells because the corneal epithelium expresses at least four connexins2 and we do not yet have dyes that can differentiate between these different overlapping connexins.
An antisense oligonucleotide approach may be preferred when there is faster protein turnover—for example, during wound healing, where connexin protein expression changes rapidly.19 The half-life of connexin43 protein in adult heart and in primary cell culture is 1.5 to 2 h,37 but protein turnover rate varies for different tissues and under different situations. In the corneal cultures, we find that a 64% reduction in protein expression appears to be the maximum achieved. This may be due to a slower protein turnover rate in the stable corneal epithelium compared with heart. The different phosphorylation states of connexin43 in different tissues could also influence protein turnover rate.38 Connexin26 appears to be the only connexin that is not phosphorylated, and it has a longer half-life of 5–19 h in liver.39,40 Although connexin26 mRNA was reduced in our studies, this may be the reason that no differences in protein expression could be observed (although the corneas were also treated for a shorter time than for the connexin43- specific antisense oligdeoxynucleotides, as described in Materials and Methods).
Deoxyribozyme design has to satisfy many design parameters, especially the sequence specificity and length of the binding arms. It was not feasible to design deoxyribozymes for all the possible AU/GU target sites on the mRNA, because many deoxyribozyme sequences were predicted to form stable internal structures, such as hairpin loops, which would impede binding to target sites. Other potential target sites could not be used because they shared conserved sequences with homologues of the same gene family or other known rodent genes. Overall, only one-third of all possible mouse connexin43 AU/GU sites generated useful deoxyribozymes, with some regions showing a higher density of potential targets than others. Deoxyribozyme mapping of target mRNA can therefore provide only a limited representation of the tertiary accessibility of the mRNA.
Mis-priming can also be a problem. The optimum length of deoxyribozyme binding arms is between 7 and 10 nucleotides and depends on the nucleotide composition of the binding arm, with CG-rich binding arms tending to be shorter. The biological specificity of binding arms, however, is a function of length and affinity ΔG. A longer binding arm may be more specific but might lead to mis-priming. We observed such nonspecific priming with some of our deoxyribozymes, specific examples being m43dz1007 and 1028 (excluding the original double bands of rat cx26 mRNA). Confirmation of biological activity inside cells is therefore important for any antisense prediction from this protocol.
Another deterrent is the amount of work associated with the initial designing of a deoxyribozyme library, given that this process must be repeated for every gene of interest. However, we believe the designing of the deoxyribozyme library could be performed in the future by computer software and that only the in vitro testing will have to be done by hand. Unlike other methods requiring assess to an oligonucleotide synthesizer to make scanning arrays or random libraries, our method is affordable. The number of oligonucleotides required is less than that required for antisense walking or even a shotgun approach. Our approach also allows future applications for modified deoxyribozymes in studies or treatments of whole organ systems.
In conclusion, we suggest that antisense oligonucleotides provide a low-cost, reversible, and highly specific approach to the regulation and study of gap junction channels in vertebrates. Our study demonstrates, for the first time, an accurate, predictable, and affordable protocol for the design of potent antisense oligonucleotides against connexin-mediated gap junction communication. The approach is likely to have more widespread application than antisense oligonucleotide design, and although our focus has been on the connexin gene family, it can be used for any mRNA binding application, such as RNAi.
This work was funded by The New Zealand Marsden Fund and Catalyst Biomedica (The Wellcome Trust, UK). Imaging was carried out in the Biomedical Imaging Research Unit at the University of Auckland.