Construction and characterization of the genetic circuits targeting TNF-α
We evaluated the therapeutic effects of the genetic circuit designed to specifically target TNF-α in colonic macrophages for the treatment of UC. We constructed a genetic circuit consisting of two functional modules: the promoter module drives the transcription of siRNA and directs the package of siRNA into exosomes, while the siRNA expression cassette module maximizes the expression of the siRNA guide strand and minimizes the expression of undesired passenger strand. Based on our previous study, the cytomegalovirus (CMV) promoter was selected as the promoter module, and the pre-miR-155 backbone was selected as the optimal siRNA expression cassette to produce siRNA.18 By inserting the TNF-α siRNA sequence into the 5’ arm of the pre-miR-155 hairpin, a CMV-directed genetic circuit (in the form of a DNA plasmid) carrying a TNF-α siRNA expression cassette was constructed (hereafter, the CMV-siRTNF-α circuit) (Fig. 1a).
Next, we tested whether the CMV-siRTNF-α circuit has the ability to synthesize functional TNF-α siRNA in vitro. ANA-1 cells were transfected with three CMV-siRTNF-α circuits designed to target different sites of the coding sequence (CDS) of TNF-α and then treated with LPS to stimulate an inflammatory response. A CMV-directed genetic circuit encoding a scrambled RNA (CMV-scrR circuit) was transfected as a negative control. ANA-1 cells exhibited a remarkable increase in TNF-α mRNA expression with LPS stimulation; however, this effect was considerably reduced in ANA-1 cells transfected with the three CMV-siRTNF-α circuits (Supplementary Fig. 1a). Likewise, while LPS dramatically stimulated the secretion of TNF-α into the cell culture medium, the three CMV-siRTNF-α circuits significantly suppressed the amounts of secreted TNF-α (Supplementary Fig. 1b). The circuit with the greatest interference efficiency, CMV-siRTNF-α-1, was selected for further evaluation.
Evaluation of the self-assembly of TNF-α siRNA-encapsulating exosomes in an ex vivo model
Subsequently, we established an ex vivo model to examine the self-assembly and secretion of TNF-α siRNA-encapsulating exosomes. An acute UC model was induced in male BALB/c mice by replacing their drinking water with a 2.5% DSS solution for 7 days; the CMV-scrR or CMV-siRTNF-α circuit was intravenously injected into DSS mice 7 times, and then, the exosomes were purified from mouse plasma and incubated with in vitro cultured macrophages (Fig. 2a). First, we confirmed the proper enrichment of exosomes from mouse plasma by nanoparticle tracking analysis (NTA) and transmission electron microscopy (TEM) (Supplementary Fig. 2). Second, we investigated whether the CMV-siRTNF-α circuit could trigger the efficient packaging of TNF-α siRNA into exosomes. A time-dependent accumulation of TNF-α siRNA was observed in the plasma derived from mice injected with the CMV-siRTNF-α circuit (peaking at 12 hours and decreasing to a background level at 48 hours) (Fig. 2b). Interestingly, most TNF-α siRNA was detected in the plasma exosomes rather than exosome-free plasma of CMV-siRTNF-α-injected mice (Fig. 2c). These results suggested that TNF-α siRNA was released into the blood circulation in an exosome-dependent manner. Third, we evaluated whether the formation of TNF-α siRNA-encapsulating plasma exosomes could reduce TNF-α expression in macrophages. When macrophages were incubated with plasma exosomes derived from mice injected with the CMV-siRTNF-α circuit and then stimulated with LPS, a significant decrease in TNF-α mRNA expression was observed (Fig. 2d), which was accompanied by a reduced amount of TNF-α secreted into the cell culture medium (Fig. 2e).
Tracking of the delivery of self-assembled TNF-α siRNA to immune cells in inflamed mucosa
We investigated whether exosome-enclosed TNF-α siRNA was successfully delivered to mononuclear phagocytes, particularly to colonic macrophages, via the exosome-circulating system. The CMV-siRTNF-α circuit was intravenously injected into DSS-induced acute UC mice, and then the biodistribution of TNF-α siRNA in various tissues was determined. Time-dependent accumulation and clearance of TNF-α siRNA was observed in the livers of CMV-siRTNF-α-injected mice (peaking at 12 hours and decreasing to a background level at 48 hours) (Fig. 2f). Similarly, direct tracking of TNF-α siRNA in mouse liver by fluorescence in situ hybridization (FISH) also revealed a time-dependent change in TNF-α siRNA levels in the liver (Fig. 2g and Supplementary Fig. 3a). These results are consistent with the findings of previous studies and support the idea that the liver can express transgenes introduced by intravenously injected genetic circuits (naked DNA plasmids).21 In addition to the liver, a time-dependent accumulation of TNF-α siRNA was also observed in the colon, spleen and kidney of CMV-siRTNF-α-injected mice (peaking at 12-24 hours and decreasing significantly after 48 hours) (Fig. 2h). Likewise, when the distribution of TNF-α siRNA to the colon was examined by FISH, hybridization signals of TNF-α siRNA were detected, with a similar time trend, in the colon sections of CMV-siRTNF-α-injected mice (Fig. 2i and Supplementary Fig. 3b). To further quantify the cellular uptake of TNF-α siRNA, immune cells (monocytes, macrophages and CD4+ T cells) were specifically isolated from the colonic lamina propria, peripheral blood and spleen of CMV-siRTNF-α-treated mice. A time-dependent increase in TNF-α siRNA levels was detected in these isolated immune cells (Fig. 2j and 2k). Thus, TNF-α siRNA-encapsulating exosomes possessed the capacity to access immune cells in inflamed mucosa, blood and spleen, resulting in efficient delivery of TNF-α siRNA to the sites of desire.
In vivo therapeutic efficacy of self-assembled TNF-α siRNA in an acute UC model
Next, we evaluated the therapeutic effects of self-assembled TNF-α siRNA on mice suffering from acute UC. An acute UC model was induced in male BALB/c mice by replacing their drinking water with a 2.5% DSS solution for 7 days. Starting on day 3, DSS mice were intravenously injected with CMV-scrR or three dosages of the CMV-siRTNF-α circuit or infliximab once a day for seven times until the final analysis on day 10 (Fig. 3a). CMV-scrR circuit-treated DSS mice experienced apparent body weight loss compared with normal control mice; however, treatment with a high dose of the CMV-siRTNF-α circuit significantly alleviated body weight loss in DSS mice (Fig. 3b). The disease activity index (DAI), a composite score reflecting the severity of the disease characteristics, was dramatically elevated in CMV-scrR circuit-treated DSS mice, but this score was significantly alleviated by treatment with a high dose of the CMV-siRTNF-α circuit (Fig. 3c). Furthermore, CMV-scrR circuit-treated DSS mice had distinctly shorter colon lengths than normal mice, but treatment with a high dose of the CMV-siRTNF-α circuit caused a significant recovery of colon length (Fig. 3d and Supplementary Fig. 4a). In terms of the expression levels of TNF-α, while a remarkable increase in colonic TNF-α mRNA and protein levels was observed in the DSS mice treated with the CMV-scrR circuit, a dose-dependent reduction in TNF-α mRNA and protein levels was observed in the colon of DSS mice injected with the CMV-siRTNF-α circuit (Fig. 3e and 3f). Immunofluorescence staining of TNF-α also confirmed a dose-dependent decline in TNF-α protein levels in the colonic lamina propria derived from CMV-siRTNF-α circuit-treated mice (Fig. 3g and Supplementary Fig. 4b). Accordingly, while the production of proinflammatory cytokines (IL-6, IL-12p70, IL-17A and IL-23) that are mechanistically associated with TNF-α was evoked in the colon of CMV-scrR circuit-treated DSS mice, all of these cytokines were markedly reduced by a high dose of the CMV-siRTNF-α circuit (Supplementary Fig. 4c). Likewise, myeloperoxidase (MPO) activity, a direct indicator of the infiltration of neutrophils into colonic mucosa, was alleviated by a high dose of the CMV-siRTNF-α circuit (Supplementary Fig. 4d). Finally, many cardinal histological signs of UC, including mucosal damage, ulceration, neutrophil infiltration, crypt abscesses and muscular layer thickness, were clearly manifested in the colonic sections derived from CMV-scrR circuit-treated mice, but these signs were significantly improved after treatment with the CMV-siRTNF-α circuit, and the colonic sections derived from CMV-siRTNF-α circuit-treated mice had relatively intact epithelia, well-defined crypt structures and relatively low levels of neutrophil infiltration (Fig. 3h and Supplementary Fig. 4e). Accordingly, the colonic histological score was higher in CMV-scrR circuit-treated mice, but this score was dose-dependently reduced by treatment with the CMV-siRTNF-α circuit (Supplementary Fig. 4f). Overall, these results demonstrated that the self-assembled TNF-α siRNA induced by the CMV-siRTNF-α circuit could alleviate the characteristics of DSS-induced inflammation and achieve mucosal healing. Conversely, infliximab also improved many critical signs of UC in DSS mice, but the therapeutic effect was somehow inferior to that of the CMV-siRTNF-α circuit (Fig. 3b-3h and Supplementary Fig. 4). Similarly, in an acute colitis model established by trinitrobenzene sulfonic acid (TNBS), the characteristic symptoms of colitis were significantly alleviated by treatment with the CMV-siRTNF-α circuit, to a similar, if not better, extent as treatment with infliximab (Supplementary Fig. 5).
In vivo therapeutic efficacy of self-assembled TNF-α siRNA in a chronic UC model
Because UC is a chronic condition characterized by recurrent episodes of intestinal inflammation, we established a DSS-induced chronic UC model (Supplementary Fig. 6a). Similar to the therapeutic efficacy acquired in acute model, CMV-siRTNF-α circuit-treated mice experienced milder symptoms of UC, including lower body weight loss, a decreased DAI score, a longer colon length, reduced colonic proinflammatory cytokines, reduced MPO activity and an improved histological appearance of colonic sections, compared with the control mice treated with the CMV-scrR circuit (Supplementary Fig. 6b-6g). At the molecular level, treatment with the CMV-siRTNF-α circuit significantly reduced the expression levels of TNF-α mRNA and protein in the colon of the chronic UC model (Supplementary Fig. 6j-6l). The CMV-siRTNF-α circuit performed slightly better than infliximab in ameliorating the clinical and histopathological severity of UC in the chronic model (Supplementary Fig. 6h-6i). Furthermore, the potential side effects and tissue toxicity of the genetic circuits were evaluated. Repeated injection of the CMV-siRTNF-α circuit posed negligible hepatic toxicity and renal toxicity and caused no overt tissue damage in the chronic UC model (Supplementary Fig. 7a). Likewise, representative serum biochemical indexes, including alanine aminotransferase (ALT), aspartate aminotransferase (AST), total bilirubin (TBIL), blood urea nitrogen (BUN), alkaline phosphatase (ALP) and creatinine (CREA), were unchanged after treatment with the CMV-siRTNF-α circuit (Supplementary Fig. 7b). In addition, the anti-TNF-α efficacy of the CMV-siRTNF-α circuit was validated in a TNBS-induced chronic colitis model (Supplementary Fig. 8). Overall, these results demonstrate the therapeutic potential of the self-assembled TNF-α siRNA to inhibit the progression of intestinal inflammation and to promote the recovery of colon tissue from UC.
Design of a multitargeted genetic circuit to simultaneously assemble multiple siRNAs for combination therapy of UC
Subsequently, we designed a combinatory multitargeted genetic circuit to simultaneously block multiple causal genes and pathways relevant to the onset and persistence of UC. Based on literature mining,22-26 we first set up a group of candidate target genes, including proinflammatory cytokines (IL-17A, INF-γ and IL-6),22-25 adhesion molecules involved in T cell trafficking to the gut (integrin α4 and ICAM-1)26,27 and molecules essential for T cell activation (CD3 and B7-1)8,9. Second, we constructed a library of genetic circuits among which each genetic circuit carried only one siRNA expression cassette directed against one of the candidate target genes (Supplementary Table. 2). After assessing the therapeutic efficacy of these genetic circuits individually, genetic circuits targeting B7-1, integrin α4 and ICAM-1 were shown to be more effective than circuits targeting IL-17A, INF-γ, IL-6 and CD3 in ameliorating the manifestations of DSS-induced UC, as assessed by the DAI score and colon length (Supplementary Fig. 9a-9c). Moreover, genetic circuits targeting B7-1 and integrin α4 were more efficient than circuits targeting IL-17A, INF-γ, IL-6, CD3 and ICAM-1 in knocking down their target genes (Supplementary Fig. 9d-9j). To view the effects as a whole, genetic circuits targeting B7-1 and integrin α4 were retained. Third, we constructed a multitargeted genetic circuit carrying three siRNA expression cassettes, which were organized into a head-to-tail tandem array under the control of a CMV promoter to simultaneously silence TNF-α, B7-1 and integrin α4 (hereafter, the CMV-siRT+B+I circuit) (Fig. 1b). As a basis of comparison, the CMV-siRTNF-α, CMV-siRB7-1 and CMV-siRIntegrin α4 circuits carrying a single siRNA expression cassette against their corresponding target genes were included as controls.
To validate that intravenous injection of the CMV-siRT+B+I circuit indeed induced the coassembly of three siRNAs into plasma exosomes and facilitated the delivery of three siRNAs to target cells, we assessed the in vivo distribution of the three siRNAs in a DSS-induced UC model. Regardless of injection with individual CMV-siRTNF-α circuit or multitargeted CMV-siRT+B+I circuit, a similar amount of TNF-α siRNA was detected in the plasma and colon of DSS mice (Supplementary Fig. 10a-10b). Direct tracking of TNF-α siRNA by FISH also revealed the clear presence of TNF-α siRNA in the colonic macrophages derived from the mice injected with the CMV-siRTNF-α or CMV-siRT+B+I circuit (Supplementary Fig. 10c). Simultaneously, a similar amount of B7-1 siRNA accumulated in the plasma, colon and colonic macrophages derived from the mice treated with individual CMV-siRB7-1 circuit or multitargeted CMV-siRT+B+I circuit (Supplementary Fig. 10d-10f). Moreover, compared with the DSS mice treated with the CMV-scrR circuit, extensive accumulation of integrin α4 siRNA was detected in the plasma, colon and colonic CD4+ T cells of DSS mice treated with the CMV-siRIntegrin α4 or CMV-siRT+B+I circuit (Supplementary Fig. 10g-10i).
Next, the therapeutic effects of the multitargeted CMV-siRT+B+I circuit were evaluated in a DSS-induced chronic UC model to investigate whether a synergistic advantage was achieved (Fig. 4a). While both multi- and single-targeted genetic circuits significantly alleviated body weight loss, promoted the recovery of colon length, reduced the DAI score and improved the histological appearance in the DSS-induced UC model, the CMV-siRT+B+I-treated group showed the least body weight loss, best colon length recovery, lowest DAI score and inflammatory cytokine levels, and minimal histological signs of UC among the three treatment groups (Fig. 4b-4e and Supplementary Fig. 11). In terms of body weight, colon length and inflammatory cytokine levels, treatment with the CMV-siRT+B+I circuit even alleviated these pathophysiological parameters to a similar level in normal control mice (Fig. 4b-4e and Supplementary Fig. 11). In particular, colon tissues from the CMV-siRT+B+I-treated group exhibited almost the same tissue morphology as that observed in the normal control group, especially with respect to the integration of the colonic epithelial layer and the infiltration of inflammatory cells (Fig. 4e and Supplementary Fig. 11c). Moreover, hepatic and renal toxicity and abnormal alterations in serum biochemical indexes were not observed after injection with the multitargeted CMV-siRT+B+I circuit (Supplementary Fig. 7). Overall, combination therapy with the multitargeted CMV-siRT+B+I circuit yielded the highest therapeutic efficacy among all the tested groups, indicating that the multitargeted genetic circuit could exert a synergistic therapeutic effect against DSS-induced UC.
At the molecular level, the knockdown efficiency of the corresponding target genes was evaluated. First, a remarkable reduction in TNF-α mRNA and protein levels was observed in the colon of mice injected with the CMV-siRTNF-α or CMV-siRT+B+I circuit (Fig. 4f-4g). To further characterize the colonic macrophages that internalized TNF-α siRNA, primary macrophages were specifically isolated from the colonic lamina propria and cultured in conditioned medium for 12 hours. Immunofluorescence staining of TNF-α in primary macrophages revealed that TNF-α protein levels were significantly decreased by treatment with CMV-siRTNF-α or CMV-siRT+B+I circuit (Supplementary Fig. 12a). Accordingly, the number of colon-infiltrating macrophages, as measured by flow cytometric analysis of F4/80+ cells in colonic lamina propria macrophages cells, was significantly decreased by treatment with the CMV-siRTNF-α or CMV-siRT+B+I circuit (Supplementary Fig. 12b-12c). Second, the levels of B7-1 mRNA were significantly reduced in the colon of mice injected with the CMV-siRB7-1 or CMV-siRT+B+I circuit (Fig. 4h). Flow cytometry also revealed that the colonic lamina propria mononuclear cells, peripheral blood mononuclear cells and splenic mononuclear cells isolated from DSS mice treated with the CMV-siRB7-1 or CMV-siRT+B+I circuit exhibited reduced amounts of B7-1 protein (Fig. 4i-4j and Supplementary Fig. 13a-13d). Immunofluorescence staining of B7-1 in primary cultured macrophages of DSS mice also confirmed an apparent reduction of B7-1 protein after treatment with the CMV-siRB7-1 or CMV-siRT+B+I circuit (Supplementary Fig. 13e). Furthermore, double staining of B7-1 and F4/80 in colon sections of DSS mice revealed that the number of double-positive B7-1+ F4/80+ cells increased remarkably in CMV-scrR circuit-treated DSS mice compared with normal mice, but this increase was abrogated by treatment with the CMV-siRB7-1 or CMV-siRT+B+I circuit, especially the latter (Fig. 4k and Supplementary Fig. 13f). These results suggest that B7-1-positive macrophages are increased at sites of intestinal inflammation, whereas intravenous injection of genetic circuits targeting B7-1 causes a synergistic decline in B7-1 protein and colon-infiltrating macrophages in inflamed mucosa. Since the B7-1 molecule on APCs provides costimulatory signals for T cell activation, we evaluated T cell activation using CD25 as a marker.28 Flow cytometry revealed that after CMV-siRB7-1 or CMV-siRT+B+I circuit treatment, the positive rate of CD25+ T cells was significantly reduced (Supplementary Fig. 13g-13h). Third, a significant decrease in integrin α4 mRNA was observed in the colon of DSS mice injected with CMV-siRIntegrin α4 or CMV-siRT+B+I circuit (Fig. 4l), which was accompanied by a significant decline in integrin α4 protein levels in the membrane surface of lymphocytes derived from the colonic lamina propria, peripheral blood and spleen of these mice (Fig. 4m-4n and Supplementary Fig. 14a-14d). In the immunofluorescence staining assay, double-positive signals of integrin α4+ CD4+ cells were readily detected in the colon sections of CMV-scrR circuit-treated DSS mice; however, the double-positive signals were remarkably diminished by treatment with the CMV-siRIntegrin α4 or CMV-siRT+B+I circuit (Fig. 4o and Supplementary Fig. 14e). To investigate the interaction of integrin α4β7 with their ligands, a solid-phase adhesion assay was performed with plates coated with E-cadherin.26 The amounts of adherent CD4+ T cells isolated from the peripheral blood of CMV-siRIntegrin α4 circuit- or CMV-siRT+B+I circuit-treated mice were significantly lower than those from CMV-scrR circuit-treated mice (Supplementary Fig. 14f-14g), indicating that homing of lymphocytes to the inflamed gut mucosa was abolished by treatment with the CMV-siRIntegrin α4 and CMV-siRT+B+I circuit.
Development of an AAV-based strategy for long-term self-assembly and delivery of exosome-enclosed siRNAs for the treatment of UC
Next, we focused on optimizing the delivery vehicles of genetic circuits to achieve a long-term therapeutic effect. Since adeno-associated virus (AAV) is clinically safe and capable of establishing long-term transgene expression, we sought to investigate whether AAV-based liver delivery and expression of a genetic circuit enabled long-term self-assembly of siRNAs in the liver and induced constant target gene silencing in vivo.29 We inserted the whole sequence of the CMV-siRTNF-α circuit into an AAV serotype 9 (AAV-9) vector (AAV-CMV-siRTNF-α) and evaluated the therapeutic effects in a chronic UC model (Supplementary Fig. 15a). Since the AAV-9 vector could coexpress TNF-α siRNA and a luciferase reporter, evaluation of luciferase activity might reflect TNF-α siRNA accumulation in vivo. AAV-mediated luciferase expression was dose-dependently increased from week 1, reached a peak at week 3 and decreased to the background level at week 10 (Supplementary Fig. 15b-15c). Compared with treatment with AAV-CMV-scrR, treatment with a high dose of AAV-CMV-siRTNF-α caused a significant recovery of body weight and colon length, a significant decline in the DAI score and proinflammatory cytokine levels and an apparent alleviation of the histological appearance in DSS mice (Supplementary Fig. 15d-15i). At the molecular level, treatment with a high dose of AAV-CMV-siRTNF-α resulted in a significant accumulation of TNF-α siRNA in plasma, and consequently, an apparent reduction in TNF-α mRNA and protein levels was observed in colon tissue (Supplementary Fig. 15j-15l). These results revealed that AAV-mediated liver expression of a genetic circuit provided a substantial and lasting therapeutic effect following a single administration.
Finally, the multitargeted CMV-siRT+B+I circuit was inserted into the AAV-9 vector (AAV-CMV-siRT+B+I) to induce long-term combination therapy in a chronic UC model (Fig. 5a). AAV-CMV-siRT+B+I-treated mice recovered body weight quite fast after each DSS administration, especially those in the high-dose group (Fig. 5b). Likewise, AAV-CMV-siRT+B+I-treated mice experienced a significant recovery in colon length, had a dramatic decrease in the DAI score and proinflammatory cytokine levels and displayed a striking improvement in histological signs compared with mice treated with AAV-CMV-scrR (Fig. 5c-5e and Supplementary Fig. 16a-16d). Remarkably, while colon tissues from the AAV-CMV-scrR-treated mice exhibited epithelial disruption, goblet cell depletion and significant infiltration of inflammatory cells into the mucosa, colons from the AAV-CMV-siRT+B+I-treated mice showed a relatively normal histology, with no clear signs of inflammation or disruption of tissue morphology (Fig. 5e and Supplementary Fig. 16c). At the molecular level, treatment with AAV-CMV-siRT+B+I resulted in a dose-dependent increase in TNF-α siRNA, B7-1 siRNA and integrin α4 siRNA in the liver, plasma and colon of DSS mice (Fig. 5f-5h and Supplementary Fig. 16e-16j), which was accompanied by a dose-dependent reduction in TNF-α, B7-1 and integrin α4 in colon tissues (Fig. 5i-5k).