Short hairpin RNA–expressing bacteria elicit RNAinterference in mammalsShuanglin Xiang1,2, Johannes Fruehauf1,2& Chiang J Li1RNA-interference (RNAi) is a potent mechanism, conservedfrom plants to humans for specific silencing of genes, whichholds promise for functional genomics and gene-targetedtherapies. Here we show that bacteria engineered to producea short hairpin RNA (shRNA) targeting a mammaliangene induce trans-kingdom RNAi in vitro and in vivo.Nonpathogenic Escherichia coli were engineered to transcribeshRNAs from a plasmid containing the invasin gene Inv andthe listeriolysin O gene HlyA, which encode two bacterialfactors needed for successful transfer of the shRNAs intomammalian cells. Upon oral or intravenous administration,E. coli encoding shRNA against CTNNB1 (catenin b-1) inducesignificant gene silencing in the intestinal epithelium and inhuman colon cancer xenografts in mice. These results providean example of trans-kingdom RNAi in higher organisms andsuggest the potential of bacteria-mediated RNAi for functionalgenomics, therapeutic target validation and development ofclinically compatible RNAi-based therapies.Several years ago, it was demonstrated that systemic gene silencingcould be attained in the nematode Caenorhabditis elegans when itingested E. coli engineered to produce interfering RNAs, suggestingthat RNAi-mediated information transfer between species or king-doms might be possible1,2. This phenomenon has offered a practicalsolution for the in vivo application of RNAi to high-throughputfunctional genomic analyses of C. elegans3–6.IntheabsenceofRNAispreading2, however, it is unknown whether and how mammalsrespond to bacterial interfering RNA under commensal, pathologicalor other conditions. Given the significance of bacteria in biology andmedicine and their versatility as gene vectors7, we decided to explorethe possibility of targeting genes in mammalian cells with engineeredbacteria that produce interfering RNAs. Bacteria-mediated RNAi inhigher organisms may hold potential for functional genomics inmammalian systems, as well as for other in vivo applications.To investigate this possibility, we first constructed the bacterialplasmid pT7RNAi-Hly-Inv,termedTRIP(trans-kingdomRNAiplas-mid) (Supplementary Fig. 1 online). In this plasmid, the expressionof shRNA was controlled by the bacteriophage T7 promoter8,9ratherthan by a mammalian promoter or enhancer. shRNA was detected inthe bacterial system (Fig. 1a). The TRIP vector contains the Inv locusthat encodes invasin10, which permits the noninvasive E. coli to enterb1-integrin-positive mammalian cells11. The TRIP vector also containsHlyA whose product, listeriolysin O, permits genetic materials toescape from entry vesicles12,13.TRIPconstructswereintroducedinto a competent strain of nonpathogenic E. coli, BL21DE3, whichcontains the T7 RNA polymerase needed for expression of shRNA. Asa proof of concept, we constructed a TRIP against the cancer geneCTNNB1. Activation of the CTNNB1 pathway from overexpression oroncogenic mutation of CTNNB1 is responsible for the initiation ofthe vast majority of colon cancers and is involved in a variety ofother cancer types14.DespitethepotentialofCTNNB1 as a cancertherapeutic target, the CTNNB1 pathway has been recalcitrant toinhibition by small molecules. CTNNB1 is a preferred choice inproof-of-concept experiments for testing the potency of new RNAiapproaches because it is commonly stabilized in cancer cells.Theoretically, TRIP could be modified to enable bacteria to expressinterfering RNA against various genes of interest.To determine if gene silencing can be achieved through this trans-kingdom system, we cocultured human colon cancer cells (SW480)in vitro with E. coli for 2 h (Fig. 1b–e), then treated them withantibiotics to remove extracellular bacteria. Cells were further culturedfor 16–48 h before harvest for analysis of gene silencing. CTNNB1 waspotently and specifically silenced at the mRNA and protein level(Fig. 1b,c), whereas b-actin, KRAS and glyceraldehyde-3-phosphatedehydrogenase (GAPDH) were not affected (Fig. 1b–f and Supple-mentary Fig. 2 online). Direct introduction of the TRIP plasmids intoSW480 cells by transfection induced no gene silencing, furtherconfirming that active shRNA was made by the bacteria but not bymammalian cells (Supplementar y Fig. 3 online). As an irrelevantshRNA control, E. coli containing the TRIP against wild-type KRASexerted no gene-silencing effect on mutated KRAS in SW480 cells(Fig. 1d). As another control for the specificity of the trans-kingdomRNAi, E. coli containing the TRIP against mutant KRAS (GGT-GTTat codon 12) silenced KRAS expression in SW480 cells containing thesame codon 12 mutation, but not in DLD1 cells with a mutation in adifferent codon of KRAS (GGC-GAC at codon 13, Fig. 1e). Theseresults suggest that the trans-kingdom RNAi is gene-specific andsufficient to discriminate a point mutation.To investigate the variables that affect the potency of the trans-kingdom RNAi system, cells were incubated for 2 h with E. coli atdifferent multiplicities of infection (MOIs). The potency of genesilencing was dependent on MOI, with near complete gene silencingReceived 15 February; accepted 10 April; published online 14 May 2006; doi:10.1038/nbt12111Division of Gastroenterology, Beth Israel Deaconess Medical Center, Harvard Medical School, 330 Brookline Ave., Boston, Massachusetts 02215, USA.2These authorscontributed equally to this work. Correspondence should be addressed to C.J.L. ([email protected]).NATURE B IOTECHNO LOGY VOLUME 24 NUMBER 6 JUNE 2006 697LETTERS© 2006 Nature Publishing Group http://www.nature.com/naturebiotechnologyat an MOI of 1:1,000 (Fig. 1c). To determine the effect of coculturetime on gene silencing, we incubated cells with the E. coli at an MOI of1:500 for different times. Gene-silencing potency increased withincubation times up to 2 h (Fig. 1f). The dependency of gene silencingon MOI and coculture time provides controllable flexibility for genesilencing in various applications.To further confirm that the CTNNB1 gene silencing is mediatedspecifically by shRNA, we attempted to identify the specific cleavagefragment of CTNNB1 mRNA by using the 5¢-RACE (rapid amplifica-tion of cDNA ends) PCR technique15.AspecifichallmarkofRNAi-mediated gene silencing is the cleavage of target mRNA at the specificsites of the mRNA as predicted from the shRNA sequence. Based onthe time course of CTNNB1 silencing (Fig.