MicroRNAs (miRNAs) are single strand RNA molecules ~ 21-23 nucleotides long that play important roles in many biological processes through regulating gene expression 1. In animal cells, miRNAs act primarily by inhibiting mRNA translation and/or stability through a process involving partial complementary base-pairing with sequences at the 3'-untranslated region (3' UTR). Numerous miRNAs have been identified. To study their functions, antisense reagents against miRNAs have been developed as a reverse genetics tool. Synthetic oligonucleotide analogues, including 2'-O-methyl oligoribonucleotides 2, locked nucleic acids 3, 2'-O-methoxyethyl oligoribonucleotides 4, and morpholinos 5, have been tested. These antisense nucleotide analogues have been used to knock down miRNAs in cultured cells 234 and in live animals including zebrafish 5, D. melanogaster 6 and mice 7.

Caenorhabditis elegans has long been used as a model organism for studying the regulation and function of small non-coding RNA molecules, and yet no antisense reagents are available to reliably inhibit miRNAs in worms. Such a technique would be very useful for studying functions of miRNAs whose deletion strains are difficult to generate; for example, mutations causing lethality or sterility 8. In addition, to dissect functions of individual miRNAs that are clustered together, or to block intronic miRNAs 89 without perturbing the function of the corresponding protein-coding genes, antisense reagents would offer a convenient approach to circumvent the limitation of using deletion strains.

Previous studies by Hutvagner et al. showed that antisense 2'-O-methyl oligoribonucleotide injected into C. elegans larvae could inhibit functions of a miRNA 2. However, injection of worm larvae is technically very demanding, and so the larval injection of anti-microRNA oligonucleotides has not been employed for C. elegans except the original report 2. A much more straightforward approach would be to inject the antisense compound into the syncitial gonad of hermaphrodites using the standard injection methods employed for the DNA transformation of C. elegans 10 so that the reagent would inhibit a targeted microRNA during the embryonic and larval development of the injected hermaphrodite's progeny. A previous attempt at this approach using 2'-O-methyl oligoribonucleotides failed to produce the expected phenotype 2. We suspect that the cellular uptake, retention or distribution properties of unconjugated 2'-O-methyl oligoribonucleotides might not be optimal for an efficient inhibition of miRNA in worms. To develop a robust antisense technique to inhibit miRNAs in C. elegans, we explored modifying 2'-O-methyl oligoribonucleotides through conjugation with dextran, a polysaccharide that is nontoxic, inert and soluble in aqueous solutions. Dextrans are well retained in cells over a long period of time and dextran-dye conjugates have been widely used for cell labelling and cell lineage tracing 11. To conjugate dextran with 2'-O-methyl oligoribonucleotides, we first reacted dextran amines (molecular weight 40 KDa, ~ 8 amines/dextran) with a water-soluble bifunctional linker, MAL-dPEG4-NHS ester in order to produce dextran- [(PEG)₄-MAL]₈ (Figure 1a). This thiol-reactive intermediate was then conjugated with eight equivalents of antisense 2'-O-methyl oligoribonucleotides containing a 5'-thiol group. The resulting product, dextran-(^as-2'OMelin-4)₈, or D-(^as-2'OMelin-4)₈, contains, on average, eight copies of 2'-O-methyl oligoribonucleotides complementary to lin-4, the founding miRNA first identified in C. elegans 1213.

To apply dextran-(^as-2'OMelin-4)₈ to inhibit lin-4 in vivo, we injected the compound into gonads of adult hermaphrodites. Dextran-rhodamine (40 KDa) was coinjected as a fluorescent marker. About 16 h after injection, we collected rhodamine-labeled embryos (n = 50) under a fluorescence dissection scope. When these embryos reached adulthood, we scored for the egg laying defective (Egl) phenotype. In C. elegans, lin-4 is required during larval development to control the timing and pattern of cell division in the hypodermis of larva stage 1 (L1) and stage 2 (L2). lin-4 loss of function mutants (lin-4(lf)) display inappropriate reiterations of early fates at late developmental stages and show a retarded heterochronic phenotype in adults in the form of the absence of adult structures (such as vulva) and the failure of egg-laying 1213.

When using an injection with a concentration of 50 μM (all concentrations refer to the total concentration of 2'-O-methyl oligoribonucleotides in the sample as determined from the ultraviolet (UV) absorption at 260 nm) dextran-(^as-2'OMelin-4)₈ was effective in inhibiting lin-4 and caused Egl in about 70% of worms (Figure 1b-d). Raising the injection concentration to 100 μM or above increased Egl to over 90% in the labelled worms. In contrast, antisense 2'-O-methyl oligoribonucleotides that is not conjugated to dextran only had a small effect, even at 200 μM (Figure 1b). In order to examine the specificity of dextran-(^as-2'OMelin-4)₈ in inhibiting lin-4, we prepared two control dextran conjugates, dextran-(^as-2'OMemiR-237)₈ and dextran-(^s-2'OMelin-4)₈. Dextran-(^as-2'OMemiR-237)₈ contains 2'-O-methyl oligoribonucleotides complementary to miR-237, a miRNA of the lin-4 family with similar, but not identical, sequence as lin-4. Dextran-(^s-2'OMelin-4)₈ contains lin-4 sequence (sense). We did not observe Egl phenotypes, or other abnormalities, in worms labelled with either of these two control oligonucleotides (Figure 1b) which confirmed that dextran-(^as-2'OMelin-4)₈ inhibits lin-4 in a sequence specific manner, also suggesting that worms tolerate dextran conjugates of 2'-O-methyl oligoribonucleotides fairly well.

In dextran-(^as-2'OMelin-4)₈, each dextran molecule is conjugated to eight copies of antisense 2'-O-methyl oligoribonucleotides. Having a high density of oligonucleotides on the surface of dextran molecule might increase the steric hindrance and compromise the hybridization efficiency of the antisense oligonucleotide to its target miRNA. In order to test whether we could improve the potency of these dextran conjugated antisense reagents by varying the coupling stoichiometry, we decreased the amount of 2'-O-methyl oligoribonucleotides used for conjugation. In addition, we also linked a fluorescent label (rhodamine B isothiocyanate) with dextran so that we could visualize the distribution of these antisense reagents directly. We prepared two rhodamine-dextran (^Rhdextran) conjugates of 2'-O-methyl oligoribonucleotides, ^Rhdextran-(^as-2'OMelin-4)₄ and ^Rhdextran-(^as-2'OMelin-4)₁, by varying the equivalents of 2'-O-methyl oligoribonucleotides added to the conjugation reaction (Figure 2). Each ^Rhdextran-(^as-2'OMelin-4)₄ or ^Rhdextran-(^as-2'OMelin-4)₁ on average contains four or one copy of lin-4 antisense 2'-O-methyl oligoribonucleotide, respectively (Figures 2 and 3a). These two dextran conjugates were comparably efficient in inhibiting lin-4, yet both were much more potent than dextran-(^as-2'OMelin-4)₈. At 20 μM or above, both ^Rhdextran-(^as-2'OMelin-4)₄and ^Rhdextran-(^as-2'OMelin-4)₁ caused Egl in nearly 100% of labelled worms (Figure 3b). In contrast, dextran-(^as-2'OMelin-4)₈ was completely ineffective at 20 μM. In addition, fluorescence imaging of ^Rhdextran-(^as-2'OMelin-4)₁ confirmed that the conjugate was localized fairly evenly in the cytosol after being taken up by cells.

In order to confirm that these antisense reagents act specifically by inhibiting lin-4, we examined several molecular and cellular markers to characterize the development of animals labelled with ^Rhdextran-(^as-2'OMelin-4)₁: (1) the formation of vulval structure; (2) adult specific alae formation and col-19 expression; and (3) the stage-specific seam cell division programmes.

The egg-laying defect of lin-4(lf) was due to inappropriate vulva development. In control animals, vulva morphogenesis is evident by early L4. By mid-L4 stage, the developing vulva displays a characteristic structure reminiscent of a Christmas tree (Figure 4a). In contrast, the vulva structure was missing in a lin-4(0) mutant, lin-4(e912) (Figure 4b). Wild type animals labelled with ^Rhdextran-(^as-2'OMelin-4)₁ showed the similar vulvaless anatomy as lin-4(e912) (Figure 4c), consistent with lin-4 inhibition by ^Rhdextran-(^as-2'OMelin-4)₁.

Besides affecting the development of vulval cells, lin-4 also controls stage-specific lateral hypodermal cell fates. At the L4 molt, lateral seam cells exit cell cycle and form alae. At about the same time, hypodermal cells start to express an adult specific green fluorescence protein (GFP) reporter, col-19::GFP (Figure 5a, b) 1415. Animals labelled with ^Rhdextran-(^as-2'OMelin-4)₁ (50 μM injection concentration) failed to form alae at the L4 molt (100%, n = 20) and showed no col-19::GFP expression (100%, n = 56; Figure 5c, d). Interestingly, 10 h later, both col-19::GFP (100%, n = 56) and alae (100%, n = 19) were observed in these animals (Figure 5e, f). Increasing the dose of ^Rhdextran-(^as-2'OMelin-4)₁ (150 μM injection concentration) prolongs the duration of lin-4 inhibition, such that at 10 h post-L4 molt, only 1.8% of labelled worms showed col19::GFP (n = 55); about 60% of these animals displayed col19::GFP 15 h later (or 25 h post-L4 molt). These results suggest that lin-4 knockdown with ^Rhdextran-(^as-2'OMelin-4)₁ caused these animals to go through extra larval stages (L5 and L6) before they entered adult. This is consistent with a reduction in (but not complete elimination of) lin-4 function.

During larval development, the lateral hypodermal seam cells follow a characteristic division pattern. After each division, one of the daughter cells fuses with the hypodermis (hyp7) and the other daughter cells divides again at a later stage until it terminally differentiates at the L4 molt (Figure 6a) 16. Seam cells of lin-4(e912) animals repeat the L1 fates in hypodermal cell lineages and are unable to exit the cell cycle at the L4 molt. In order to determine how seam cell development is altered in lin-4 knockdown animals, we followed the seam cell division pattern by observing GFP-labelled seam cells (wIs51, strain JR667). Both L1 and L2 divisions in lin-4 knockdown appeared to be normal, with each V cell (including V1 - V4 and V6) divided once in L1 (100%, n = 10, data not shown) and twice in L2 (100%, n = 10; Figure 6b, c). However, in early L3, V cells in lin-4 knockdown repeated the L2 division pattern, so that each V cell gave rise to twice as many daughter cells as the corresponding seam cells of the control animals (100%, n = 6; Figure 6d, e). This pattern of reiterated L2 seam cell division is consistent with a reduced, but not absent, activity of lin-4, possibly resulting in an intermediate level of lin-14 over expression 17.

Finally, since lin-4 functions through lin-14, mutations in lin-14 should suppress the phenotype of lin-4 knockdown. Indeed, at 20°C, inhibition of lin-4 in lin-14(n179), a lin-14 nonnull mutant, with ^Rhdextran-(^as-2'OMelin-4)₁ (50 μM injection concentration) only caused Egl in 2.6% of labelled worms (n = 190) and all the examined young adult worms displayed normal alae (n = 30).

Together, these data showed that ^Rhdextran-(^as-2'OMelin-4)₁ caused developmental retardation consistent with lin-4 knockdown, confirming its efficacy and specificity in inhibiting lin-4 during development.

In order to test whether these conjugated antisense agents can be used to inhibit other miRNAs in worms, we prepared ^Rhdextran-(^as-2'OMelsy-6)₁ and ^Rhdextran-(^as-2'OMelet-7)₁ using the same procedure as for making ^Rhdextran-(^as-2'OMelin-4)₁. These two dextran conjugates were designed to block lsy-6 and let-7, respectively, two miRNAs of known functions in C. elegans.

The lsy-6 microRNA regulates left-right asymmetry of ASE neurons, a pair of chemosensory neurons that share many bilaterally symmetrical features, yet differ in their ability to discriminate different ions by expressing distinct sets of chemoreceptors of the gcy gene family 18. lsy-6 is only present in ASEL (left ASE) of adult worms. It promotes gcy-7 expression through repression of cog-1, a transcription factor that negatively regulates gcy-7 expression 18. In order to score the inhibition of lsy-6 activity, we used two reporter strains expressing GFP in either ASEL (gcy-7^prom::gfp, strain OH3191) or ASER (right ASE; gcy-5^prom::gfp, strain OH3192). ^Rhdextran-(^as-2'OMelsy-6)₁ was highly effective in inhibiting lsy-6. At the injection concentration of only 3 μM, the compound inhibited lsy-6 with nearly 100% penetrance, assayed by the repression of gcy-7^prom::gfp expression in ASEL (Figure 7a), and by the induction of ectopic gcy-5^prom::gfp expression in ASEL (Figure 7b).

Another miRNA, let-7, controls the larval-to-adult transition by repressing the translation of lin-41 and hbl-1. Mutations in let-7 lead to a retarded terminal differentiation of seam cells, which results in the elimination of alae and a bursting of the animal at the vulva as the animal undergoes the fourth molt 19. We observed a dose-dependent bursting or no-alae phenotype, consistent with inhibition of let-7 activity in the progeny of hermaphrodites injected with ^Rhdextran-(^as-2'OMelet-7)₁ (Figure 8). When ^Rhdextran-(^as-2'OMelet-7)₁ was injected at 20 μM, it caused bursting vulva in about 40% of labelled worms and eliminated the formation of alae in nearly 80% of worms. Raising the concentration of ^Rhdextran-(^as-2'OMelet-7)₁ to 50 μM or above increased the penetrance, and none of the labelled animals showed alae when they reached young adults. In contrast, an injection of ^Rhdextran-(^as-2'OMemir-84)₁ directed against another let-7 family microRNA caused no observable phenotype. Since mir-84 mutations do not cause visible phenotypes 20, this result is consistent with these antisense reagents inhibiting the targeted miRNA with high specificity. Further, in a lin-41 nonnull mutant, lin-41(ma104), ^Rhdextran-(^as-2'OMelet-7)₁ (50 μM) failed to produce the bursting vulva or alae defect in labelled worms (n = 12). This again suggested that these antisense reagents acted specifically by blocking their corresponding microRNAs.

In order to confirm that this class of antisense reagents is also effective in inhibiting microRNAs during embryo development, we prepared ^Rhdextran-(^as-2'OMemir-42)₁ against mir-42, a member of the mir-35 family which consists of eight microRNA genes of similar sequences. This family of microRNAs is expressed only during embryogenesis 21 and functions redundantly to control embryonic development 822. Deletion of the seven microRNAs (mir35 - 41, strain MT14119) leads to a temperature-sensitive late embryonic or L1 lethal phenotype. At 15°C, about 10% of the embryos displayed this phenotype, while the remaining 90% developed normally (Figure 9). In this mir35-41 null background, inhibition of mir-42 with ^Rhdextran-(^as-2'OMemir-42)₁ dramatically increased embryonic or L1 lethality: at the injection concentration of 3 μM, 77% of labelled worms were embryonic or L1 lethal. In contrast, inhibition of mir-42 had little effect in N2 strain (Figure 9).

Among four miRNAs tested (lin-4, lsy-6, let-7 and mir-42), the dose of antisense reagents required for the effective inhibition of individual miRNA varied from as low as three micromolar (lsy-6 and mir-42) to as high as tens of micromolar (let-7). Two factors may account for the difference in the apparent potency of these antisense reagents. First, the cellular expression level of different miRNAs varies over a wide range 23, so a higher concentration of antisense reagents is needed to block more abundant miRNAs. Second, to inhibit a miRNA that is expressed late in larval development, it would require a higher concentration of reagent in the zygote to compensate for the dilution effect of cell division and larval growth. let-7, for example, is not expressed until the third larval stage and functions in the fourth stage 19. By that time, the antisense reagent would be further diluted as animals grow and expand in size. In principle, a higher concentration of antisense reagents can be used to overcome the dilution effect. However, it should be cautioned that, as we raised the dose, we also noticed that an increasing number of embryos failed to hatch at concentrations above 50 μM. The extent to which these antisense reagents affect embryo development seemed to vary. Among the reagents tested, antisense reagents (antimirs) against lin-4 and mir-237 were best tolerated, with nearly 100% of the embryos hatched normally at 100 μM. By comparison, the antimir against let-7 was least tolerated, with close to 40% of embryos hatched at 100 μM (Figure 10). We have not yet explored the causes for this differential effect, although we anticipate that the future development of other bioconjugates employing different classes of oligonucleotides, for example locked nucleic acids or 2'-O-methoxyethyl oligoribonucleotides, may offer an improvement on the potency with concomitant reduction of the perturbation on embryo development.

In order to test whether these antisense reagents can be used combinatorially to inhibit more than one miRNAs at a time, we coinjected ^Rhdextran-(^as-2'OMelsy-6)₁ and ^Rhdextran-(^as-2'OMelin-4)₁ into gonads of OH3192 strain (gcy-5^prom::gfp). Together, these two reagents caused Egl in all labelled worms and they induced ectopic gcy-5^prom::gfp expression in ASEL (Figure 11). In contrast, ^Rhdextran-(^as-2'OMelin-4)₁ alone did not alter the expression pattern of gcy-5^prom::gfp, and ^Rhdextran-(^as-2'OMelsy-6)₁ by itself failed to cause Egl (Figure 11). The results again confirmed the specificity of these dextran-conjugated antisense reagents and suggested that multiple antisense reagents can be used in combination to block more than one miRNAs in C. elegans. As numerous deletion strains of single miRNA genes appear to be grossly normal, it has been suggested that redundancy might mask their functions 8. Mixing these antisense reagents would allow the study of the combinatorial effects of multiple miRNAs on gene expression and should facilitate screening genetic interactions using mutants or reporter strains.

We have developed a new class of antisense reagents that potently and selectively inhibit miRNAs in C. elegans. This offers an experimental approach complementary to mutational strategy for the study of the functions of miRNA in vivo.

ASEL: left ASE; ASER: right ASE; Egl: egg laying defective; GFP: green fluorescence protein; MWCO: molecular weight cut off; TCEP: tris(2-carboxyethyl) phosphine; UV: ultraviolet; 3'UTR: 3'-untranslated region.

The authors declare that they have no competing interests.

GHZ and WHL designed the conjugates and syntheses. VA and WHL designed biological assays. GHZ performed all the experiments. WHL conceived the project. WHL, VA and GHZ analyzed the data and wrote the manuscript.