Characterizations of sugar-alcohol-derived LNPs for mRNA delivery in ASCs
Dianhydrohexitols, including isosorbide, isoidide, and isomannide, are a series of six-carbon sugar-derived heterocyclic compounds consisting of two cis-fused tetrahydrofuran rings with two secondary hydroxyl groups situated at carbon 2 and 542. The different configurations of the two hydroxyl groups, designated as exo or endo, generate three isomers with distinct reaction activity and steric hindrance. Such structure diversity empowers dianhydrohexitols as versatile building blocks for the development of new chemicals in the medical and material sciences. Inspired by the unique properties of dianhydrohexitols, three sets of ionizable lipids (DIM, DIS, and LIS) were prepared with isomannide, isosorbide, and L-sorbitol as the starting materials, respectively (Supplementary Fig. 1). The synthetic route of DIM lipids was shown in Fig. 1b as a general synthetic approach to all sets of lipids. Commercially available isosorbide was treated with acrylonitrile and sodium hydroxide to give compound 3 via a double Michael addition reactions. Then, the free core amine 4 was obtained via the reduction of the nitrile groups with borane followed by treatment with hydrochloride and anionic ion exchange resin47. As the hydrophobic tails of ionizable lipids can greatly affect LNP formulations and their interaction with bio-membranes such as cell membranes and endosome membranes, we installed five types of hydrophobic tails to the core amine 4 to equip DIM lipids via epoxide ring-opening reaction with alkyl epoxide or reductive amination reaction with aldehydes. The resulting lipids contain hydrophobic domains incorporating five different functional groups, such as hydroxyl (DIM1), hydrocarbon (DIM2 to DIM6), ester (DIM7), carbonate ester (DIM8), and acetal (DIM9 and DIM10) (Fig. 1b; Supplementary Fig. 1). The synthesis of LIS lipids started with the treatment of L-sorbitol with sodium ethoxide and dimethyl carbonate, resulting in the enantiomer of isosorbide48. With the commercially available isosorbide (derived from D-sorbitol) and the enantiomer of isosorbide as the starting materials, the DIS and LIS lipids were synthesized following similar synthetic routes, respectively. 1H nuclear magnetic resonance and mass spectrometry were used to confirm the chemical structures of these ionizable lipids43.
As reported in our prior studies43,49, we formulated these dianhydrohexitols-derived ionizable lipids with firefly-luciferase (FLuc) mRNAs (FLuc-LNPs) and evaluated their physicochemical properties (Supplementary Fig. 2a–c). We then examined their mRNA delivery in ASCs based on luminescence intensity and profiled the structure-activity relationship. Generally, isomannide-based lipids (DIM) showed more efficient mRNA delivery than DIS and LIS lipids, the amine cores of which are a pair of enantiomers, indicating that the chirality of the amine cores can influence the delivery efficiency of the formulated LNPs in primary ASCs (Fig. 2a). Moreover, the lipids with hydroxyl tails, including DIM1, DIS1, and LIS1, yielded higher mRNA delivery capacity than those with other hydrophobic tails. In particular, the LNPs with hydrocarbon and carbonate ester groups in tails showed negligible mRNA delivery efficiency in ASCs. Among all these LNPs, DIM1 LNPs exhibited the highest capacity to deliver mRNA, which was over 70-fold more effective than Lipofectamine™ 3000 and electroporation at the same mRNA concentration (Fig. 2a). To optimize the formulation of DIM1 LNPs, we designed an L16 (4)4 orthogonal table for the determination of optimal molar ratios of each lipid (Supplementary Fig. 3a). Then, we varied the mass ratios of DIM1 lipid to mRNA in the orthogonal-predicted formulation to further increase the mRNA delivery efficiency of DIM1 LNPs (Fig. 2b, c). The optimal formulation DIM1T resulted in 1.5-fold and 2.7-fold higher luminescence intensity than the top orthogonal formulation DIM1E and the original formulation DIM1, respectively (Fig. 2d). Moreover, the delivery efficiency of DIM1T LNPs in ASCs was over 140-fold greater than those of ALC-0315 and MC3 LNPs, and 22-fold higher than that of SM102 LNPs (Fig. 2e). In addition, DIM1T LNPs encapsulating GFP mRNA yielded over 90% GFP positive ASCs, with the mean fluorescence intensity (MFI) notably superior to those FDA-approved lipid formulations (Supplementary Figs. 4a–c and 5a). The hydrodynamic diameter of DIM1T LNPs was around 110 nm with a polydispersity index (PDI) < 0.15 (Fig. 2f). Over 80% mRNA was encapsulated in DIM1T LNPs, and the resultant particles displayed a slightly positive charge and spherical morphology (Fig. 2g, h; Supplementary Fig. 3b). Additionally, DIM1T LNPs had an apparent pKa of 6.56 as analyzed by the 6-(p-toluidino)−2-naphthalenesulfonic acid (TNS) assay (Supplementary Fig. 3c)50. Accordingly, the DIM1T LNPs were chosen for ex vivo ASCs engineering in the following studies.
Delivery of saRNA and E3 mRNA complex (SEC) using DIM1T LNPs facilitates prolonged protein production in ASCs
Sustained protein production by ASCs is crucial for augmenting their therapeutic potency. In contrast to traditional IVT mRNAs, self-amplifying RNAs (saRNAs), which encode both genes of interest and viral replicase, can self-replicate within cells and continuously generate equivalent or superior expression of desired proteins at a lower dose51. Therefore, we speculate that the delivery of saRNA encoding specific therapeutic proteins to ASCs may facilitate topical protein secretion for a relatively long period of time compared with mRNA. To study the delivery feasibility of saRNA in ASCs, we quantified the luminescence intensity in the ASCs treated with DIM1T LNPs encapsulating FLuc-saRNA. Interestingly, the expression of FLuc-saRNA in ASCs plunged to a negligible level 48 h post-treatment (Fig. 3a). We reproduced this treatment in 293T cells, a human embryonic kidney cell line, and did not observe a similar sharp decrease in expression (Supplementary Fig. 6a). Such different expression dynamics indicated that certain inflammatory pathways in ASCs were triggered to shut down the translation of saRNA or even directly eliminate them since saRNA can generate double-stranded RNA (dsRNA) intermediates during replicative translation, which are natural ligands of cytoplasmic RNA sensors46. As previously reported, dsRNA intermediates may activate PKR and subsequently phosphorylate eIF2α, thereby blocking cap-dependent translation45. To explore the mechanism of saRNA translational shutdown in ASCs, we evaluated the cytosolic level of PKR, eIF2α, and their phosphorylated derivatives in the saRNA-treated ASCs. Delivery of FLuc-saRNA to ASCs sharply increased the intracellular expression of the phosphorylated PKR and eIF2α after 48 h compared with the WT ASCs and mRNA-treated ASCs, indicating that the dsRNA intermediates triggered the PKR/eIF2α-mediated translation inhibition in ASCs (Supplementary Figs. 7a–e and 8). To evade cellular immunity against saRNA translation, specialized immune evasion proteins developed by viruses can be implemented to either avoid activation of PKR or prevent eIF2α phosphorylation. It has been demonstrated that co-delivery of saRNA and mRNA encoding vaccinia virus (VACV) protein E3 could substantially increase and prolong the translation of saRNA in human foreskin fibroblasts (HFF)44. To evaluate whether E3 proteins can expedite the durable expression of genes transferred by saRNA in ASCs, we encapsulated FLuc-saRNA and E3-mRNA in DIM1T LNPs at different saRNA/mRNA mass ratios and treated ASCs with these LNPs at the same total RNA doses. The results showed that co-delivery of E3-mRNA significantly augmented luciferase expression by saRNA and the expression could remain at least 48 h post-treatment (Fig. 3b). Moreover, the DIM1T LNPs encapsulating FLuc-saRNA/E3 mRNA complex at a mass ratio of 0.5 induced the highest luminescence intensity among all the ratios tested (Fig. 3b). Delivery of FLuc saRNA/E3 mRNA complex (abbreviated as SEC) using DIM1T LNPs significantly decreased the expression level of phosphorylated PKR and eIF2α in ASCs compared with that of FLuc saRNA-treated ones (Supplementary Figs. 7a–e and 8). Collectively, the saRNA/mRNA mass ratio of 0.5 was applied in the following experiments.
To investigate the expression duration of FLuc-SEC in ASCs, we treated ASCs with FLuc-mRNA, FLuc-saRNA, and FLuc-SEC at the same total RNA doses, respectively. The DIM1T LNPs displayed similar particle sizes, encapsulation efficiency and zeta potential when formulating different RNA cargos (Fig. 3c, Supplementary Fig. 6b, c). Compound C16, a PKR inhibitor, was included as a positive control group. Significantly, co-delivery of FLuc-SEC induced a 1.48-fold, 1.54-fold, and 1.17-fold higher luminescence signal than FLuc-saRNA, FLuc-mRNA, and FLuc-saRNA/C16 group on day 1, respectively (Fig. 3d). Such leads were substantially expanded on day 2 as the translation of saRNA was blocked by ASCs without inhibiting PKR activation. Compared with the C16 compound, the delivery of E3-mRNA was able to make the inhibition more durable as only the FLuc-SEC group could still detect FLuc expression on day 5, which then persisted for another 4 days. Additionally, intracellular levels of E3 proteins in ASCs remained detectable on day 9 (Supplementary Fig. 9a). Despite the low levels, these levels were adequate to retain the expression of proteins of interest after 9 days. To evaluate whether the delivery of FLuc-SEC using DIM1T LNPs would alter the ex vivo characteristics of ASCs, we used flow cytometry analysis to compare surface marker expression between non-engineered ASCs and FLuc-SEC mRNA reprogrammed ones. Notably, we found both types of ASCs express similar levels of CD106, CD44, CD29, and SCA-1 in the absence of CD11b and CD45, indicating the treatment of FLuc-SEC mRNA delivered by DIM1T LNPs could induce and facilitate sustained protein production in ASCs without disrupting their phenotypes (Fig. 3e, f; Supplementary Fig. 5b). Moreover, both WT ASCs and their SEC-engineered counterparts retained the differentiation capacity as evidenced by the detection of FABP4+ and osteopontin+ cells after incubation in standard differentiation culture, respectively (Supplementary Fig. 9b). The DIM1T LNPs encapsulating FLuc-SEC caused minimal cytotoxicity to ASCs at various RNA doses, confirming the excellent biocompatibility of DIM1T lipids (Fig. 3g). To investigate cellular uptake pathways of DIM1T LNPs in ASCs, we incorporated an RNA labeled with Alexa-Fluor 647 in the formulation. Only when pre-incubating the cells with methyl-beta-cyclodextrin (MβCD), a caveolae-mediated endocytosis inhibitor, did we observe a dramatic decrease (97%) in uptake efficiency, suggesting that the cellular uptake of these DIM1T LNPs in ASCs is predominantly mediated by caveolae-mediated endocytosis (Fig. 3h). By treating ASCs with both DIM1T LNPs and calcein, a membrane-impermeable fluorescent dye, we observed the green fluorescence dots diffused in the cytoplasm of DIM1T LNPs-engineered ASCs, suggesting the endosome membranes were disrupted and DIM1T LNPs were released into the cytosol (Fig. 3i).
HGF DIM1T-SEC engineered ASCs (DS-ASCs) accelerate diabetic wound healing
To explore the therapeutic potential of ASCs engineered by DIM1T-SEC LNPs in diabetic cutaneous wounds, we constructed saRNAs encoding hepatocyte growth factor (HGF) to generate therapeutic protein-hypersecreting ASCs. Previous studies reported that HGF can regulate cellular migration, proliferation, and morphogenesis in many types of cells at wound sites, thereby promoting epithelial repair and neovascularization during wound healing52,53. The short half-life of HGF (<3–5 min) in plasma, however, may require repeated administrations to maintain therapeutic windows in clinical applications. Therefore, the sustained production of HGF at the wound sites by ASCs may establish a clinically feasible treatment strategy. To confirm the construction of HGF-encoded saRNAs, the HGF secretion level of DS-ASCs in the conditioned medium was quantified using ELISA assay. The DIM1T-SEC-engineered ASCs secreted a significantly higher level of mature HGF proteins than the mRNA- or saRNA- treated groups from Day 1, the level of which was still detectable 9 days post-treatment (Fig. 4a). For saRNA alone groups, the secretion level of proteins was negligible on Day 2, corresponding to a similar dynamic expression pattern in the FLuc-saRNA assay. This result suggested that the translation of HGF proteins by saRNA was completely hampered in ASCs and the incorporation of E3 mRNA played an essential role in maintaining protein expression.
To evaluate the therapeutic functions of DS-ASCs, we established an excisional wound model in db/db mice, which mimics the delayed wound healing process observed in DFU patients41,54,55. Instead of subcutaneously administrating ASCs in the wound sites, we embedded ASCs with an in situ crosslinked HyStem®-HP hydrogel above the wound bed for sustained cell retention and HGF protein generation41,56. This hydrogel system is composed of crosslinked hyaluronan, heparin, and denatured collagen, and has been extensively utilized in cell culture and delivery41,57. The ASCs were treated with DIM1T-HGF mRNA (DM) or DIM1T-HGF SEC for 12 h before being harvested and encapsulated in the hydrogels. Shortly after wound formation, 3 ×105 DIM1T-engineered ASCs were embedded above the wound bed (n = 10 wounds per group). The wounds were imaged in 3-day intervals until closure (Fig. 4b). Starting from Day 6, we observed pronounced differences in wound healing rates between different groups. Compared with the vehicle group (hydrogel only), all ASC groups exhibited accelerated wound size reduction. Among them, the HGF DS-ASCs yielded the most efficient wound-healing kinetics as the average wound size of this group on Day 18 was 0.00 ± 0.00% (Fig. 4c; Supplementary Fig. 10a). Such rates in other groups were 25.60% ± 6.28% (Vehicles), 18.87% ± 4.70% (WT ASCs), and 7.03% ± 6.29% (HGF DM-ASCs). The area-under-curve (AUC) of each wound was graphically computed and normalized as average values against the vehicle groups. The normalized AUCs of wounds in the DS-ASCs group were significantly lower than those of wounds treated by WT ASCs and HGF DM-ASCs, respectively (Fig. 4d). Moreover, the wounds treated with HGF DS-ASCs showed complete closure starting from Day 15 and were all closed after Day 18 post-wounding, the time-to-closure function of which was significantly more efficient than other groups (Fig. 4e). The wounds treated by HGF DS-ASCs demonstrated more efficient re-epithelialization and displayed well-formed hyperproliferative epidermis compared with other groups, which is an encouraging indication of advanced healing (Fig. 4f, g). Furthermore, enhanced vascularization was observed in the wounds treated with HGF DS-ASCs as evidenced by the higher density of CD31+ vessels than other groups (Fig. 4h, i). Normally, the conversion of fibroblasts to myofibroblasts was limited in diabetic wounds, resulting in deficient wound contraction capability. However, the population of αSMA+ myofibroblasts in the wounds was substantially increased in the HGF DS-ASCs group, demonstrating a more active and efficient healing process (Fig. 4j, k). These data suggested that the delivery of HGF saRNA/E3 mRNA complex using DIM1T LNPs could elevate and elongate the secretion of HGF proteins by ASCs, thereby enhancing their therapeutic efficacy in healing diabetic wounds.
CXCL12 DS-ASCs reprogram local immune microenvironment in diabetic wounds
In addition to growth factors, chemokines have been explored to promote diabetic wound healing58,59. For example, CXCL12 (also known as stromal cell-derived factor 1α) is an immune-regulating chemokine that can bind to C-X-C chemokine receptor type 4 (CXCR4) expressed on immune cells and keratinocytes at wound sites, thereby mitigating local inflammation, enhancing angiogenesis, and promoting cell proliferation60,61. Similar to HGF proteins, CXCL12 has a short half-life in plasma, demanding repetitive administration to reach therapeutically efficacious levels. To compare the therapeutic efficacy between HGF and CXCL12 in healing diabetic wounds, we constructed saRNA encoding CXCL12 for SEC-DIM1T-mediated ASC engineering and reproduced the excisional wound healing experiments in db/db mice using the same procedure described before. In an ex vivo protein secretion assay, the CXCL12 DS-ASCs were able to constantly generate CXCL12 in culture medium, the level of which remained detectable for 9 days post-treatment (Fig. 5a). After being embedded on the wound sites, the CXCL12 DS-ASCs reduced wound size at significantly accelerated healing kinetics (0.00% ± 0.00% in relatively wound size on Day 15) and showed substantially smaller AUCs of wounds compared with HGF-generating groups (10.04% ± 9.14% in relatively wound size on Day 15) (Fig. 5b–d; Supplementary Fig. 10b). Moreover, all the wounds treated with the CXCL12 DS-ASCs were entirely closed on Day 15 while, simultaneously, only 50% of the wounds in the HGF DS-ASCs group demonstrated complete closure (Fig. 5e). Moreover, the wounds treated by the CXCL12 DS-ASCs displayed a significantly thicker layer of epidermis at wound closure sites and a substantially higher density of CD31+ blood vessels and αSMA+ myofibroblast (Fig. 5f–k). Notably, immunofluorescence staining for proinflammatory cytokine, interleukin-6 (IL-6), and anti-inflammatory cytokine, interleukin-10 (IL-10), indicated that the CXCL12 SEC-engineered ASCs significantly mitigated the dysregulated local inflammation at the wound sites compared with HGF DS-ASCs and WT controls (Fig. 5l–o). The sustained generation of CXCL12 chemokines at wound sites by the CXCL12 DS-ASCs facilitated the effective shift from the early inflammation stage to the anti-inflammatory stage, resulting in enhanced efficacy for wound healing.
To investigate the potential synergistic effects of CXCL12 and HGF in wound healing, we conducted the same excisional wound healing experiments in db/db mice and embedded the engineered DS-ACSs generating both CXCL12 and HGF on the wound sites. In comparison with vehicle controls and FLuc DS-ASCs group, the wounds treated with the CXCL12/HGF DS-ASCs or CXCL12 DS-ASCs demonstrated pronounced acceleration in the healing process, which was highlighted by the decrease in wound size and smaller wound healing AUC (Supplementary Figs. 11 and 12a, b). Particularly, the average relative wound size of these groups on Day 15 was 28.87% ± 6.67% (Vehicles), 20.00% ± 7.21% (FLuc DS-ASCs), 0.00% ± 0.00% (CXCL12 DS-ASCs) and 1.3% ± 1.9% (CXCL12/HGF DS-ASCs). Remarkably, 80% of the wounds in CXCL12/HGF SC-ACSs group and a full 100% in CXCL12 DS-ASCs group exhibited complete closure by Day 15 (Supplementary Fig. 12c). Moreover, both CXCL12 DS-ASCs and CXCL12/HGF DS-ASCs substantially increased the thickness of the hyperproliferative epidermis and density of CD31+ blood vessels and αSMA+ myofibroblast relative to vehicles and FLuc-DS ASCs (Supplementary Fig. 12d–i). Both groups also showed increased accumulation of IL10 as well as a decreased level of IL6 in the wound microenvironment (Supplementary Fig. 12j–m). However, no statistically significant differences were observed in the aforementioned parameters between these two groups, suggesting the absence of synergistic therapeutic efficacy when combining CXCL12 and HGF in expediting wound healing.
Furthermore, to assess the potential effects of gender differences on diabetic wound healing rate, we conducted parallel investigations employing identical experimental protocols across treatment groups: vehicle controls, FLuc DS-ASCs, CXCL12 DS-ASCs, and CXCL12/HGF DS-ASCs. Each of these groups was balanced with an equal representation of male and female db/db mice (four each). Throughout the duration of the study, the relative wound size, measured on average or individually, did not reveal any significant variations in the wound healing kinetics attributable to sex differences (Supplementary Fig. 13a–d).