Nongenetic surface engineering of mesenchymal stromal cells with polyvalent antibodies to

In this study, we used DNA hybridization chain reaction (HCR) technology to prepare polyvalent antibodies33. The system includes a DNA initiator (DI) and two DNA monomers (DM); DM1 and DM2 (Supplementary Fig. 1). The DI sequentially opens the DNA monomers one by one and initiates polymerization to form a DNA polymer (Fig. 1a). The DNA monomer was designed with an amino attached to the end to connect to the antibody. Electrophoresis gel images demonstrated that the DNA polymers could be synthesized by DM1 and DM2 via the HCR in the presence of DI (Fig. 1a). Then, we covalently connected DM1 and DM2 to anti-VCAM1 (Fig. 1b). Electrophoretic gel images showed that the bands of synthesized DNA-anti-VCAM1 were higher than those of anti-VCAM1, indicating the increased molecular weight of anti-VCAM1. The UV‒vis absorption spectra revealed that DNA-anti-VCAM1 had a significantly higher absorbance than anti-VCAM1 at 260 nm and 280 nm. These results demonstrated the ligation of anti-VCAM1 to DNA. We further examined whether binding to DNA affected the function of the anti-VCAM1 antibody. We used two cell lines, the endothelial cell line C166 that overexpresses VCAM1 and the control cell line K562 that does not express VCAM1. Flow cytometry showed that the anti-VCAM1 protein was able to bind to C166 but not K562 cells with or without DNA ligation, suggesting that DNA ligation did not alter the specificity of anti-VCAM1 (Fig. 1c). To polymerize DNA-anti-VCAM1 monomers into polyvalent structures, DNA-anti-VCAM1 monomers and DI were incubated for 3 h in a neutral buffer. Electrophoresis gel images revealed that DNA-anti-VCAM1 monomers could form protein multimers with larger molecular weights in the presence of DI (Fig. 1d). Overall, these data suggest that polyvalent biomolecules can be efficiently and systematically constructed by DNA self-assembly under physiological conditions.

Construction of PAV on the cell surface

We used a lipid-DNA-directed bottom-up self-assembly strategy to construct PAV on the cell surface. As shown in Fig. 2a, the DI with cholesterol at the end was inserted into the cell membrane by hydrophobic intercalation, thus anchoring the oligonucleotide on the cell surface. PAV-engineered cells were then prepared by simultaneously incubating DI-modified cells with DM1-anti-VCAM1 and DM2-anti-VCAM1. As a control group, monovalent anti-VCAM1 (MAV)-engineered cells were also prepared by incubating DI-modified cells with DM1-anti-VCAM1 alone.

We first performed a series of experiments to screen for pairs of DNA sequences that, when formed into multimers on the cell surface, could load sufficient antibodies. In addition to DNA sequence 1 (Fig. 1), we synthesized four different sets of DNA sequences. All five sets of DNA monomers were able to self-assemble into DNA polymers after the addition of DI (Supplementary Fig. 2). The DNA monomers in each set were individually conjugated with FITC-labeled IgG to form DNA-IgG monomers. Then, monovalent or polyvalent IgG was formed on the surface of K562 cells by the bottom-up approach. The number of IgG molecules in the polyvalent structure formed by different sets of DNA sequences could be quantified by determining the fluorescence intensity of FITC. After comparing the fluorescence intensities of the five sets of DNA sequences by flow analysis, it was found that DNA sequence 1 could load as many as 8 IgG proteins per DNA scaffold on the cell surface. Thus, we used the first set of DNA sequences to construct PAV on the cell surface for the following experiments.

To better evaluate the adhesion of the engineered cells to vascular endothelial cells, we used the suspension cell line K562 as the model. Thus, the primary driving force for cell adhesion would be specific recognition but not the natural state of cell adhesion. After engineering K562 cells with PAV using the method described above, the cells were first characterized using confocal fluorescence microscopy, and the fluorescence images showed that the surface of the K562 cells exhibited an intense green fluorescence signal (Fig. 2b). Further quantification of the fluorescence signal on the surface of K562 cells by flow cytometry showed an 8-fold increase in the fluorescence intensity in the PAV group compared to the MAV group (Fig. 2c and Supplementary Fig. 3), indicating an average of 8 anti-VCAM1 molecules assembled on each DNA scaffold. We also quantified the density of PAV on the cell surface by measuring the density of DI, since each DI generates one DNA scaffold. For example, at 0.5 μM DI, the density of PAV was ~1.6 × 107 units/cell (Supplementary Table 1). To better view the structure of the polyvalent antibody on the cell surface, we replaced the protein with similarly sized quantum dots (QDs) and used the same method to assemble DNA-QDs into multimers on the cell surface. Scanning transmission electron microscopy (STEM) images clearly revealed the extension of the multimeric structure of QDs from the cell surface, suggesting that the DNA assembled monomers into multimers through self-assembly (Fig. 2d). We then examined the stability of both monovalent and polyvalent surface modifications. Following a 24 h exposure of engineered cells to serum-containing culture medium, PAV-engineered cells maintained a surface antibody level of nearly 20%, corresponding to 2.4 × 107 antibodies. While MAV-engineered cells exhibited a surface antibody level of less than 10%, corresponding to 0.15 × 107 antibodies on the surface (Supplementary Fig. 4a, b). Notably, the data highlights a staggering 16-fold increase in the remaining surface antibodies on PAV-engineered cells relative to MAV-engineered cells. We also conducted an experiment to investigate the potential for protein exchange between the engineered cells and neighboring native cells. The results demonstrated that no antibodies were detected on the surface of native cells following co-culture with engineered cells, indicating that protein exchange is not occurring between the engineered cells and neighboring cells (Supplementary Fig. 4c). This finding suggests that the engineered cells do not affect other cells in their vicinity.

Enhanced adhesion of PAV-engineered K562 cells to vascular endothelial cells

To determine whether surface engineering with PAV could mediate and enhance the adhesion of K562 cells to vascular endothelial cells, we incubated engineered K562 cells with C166 cells under static conditions. Fluorescence microscopy images showed that MAV and PAV could mediate the adhesion of K562 cells to C166 cells. As expected, the number of adherent K562 cells in the PAV group was much higher than that in the MAV group (Supplementary Fig. 5).

Considering the in vivo shear stress conditions generated by blood flow in the vasculature34, a cell adhesion assay under static conditions does not accurately mimic the adhesion of engineered cells to vascular endothelial layers in vivo35. Therefore, we designed a flow adhesion experiment to study cell adhesion under well-defined shear conditions (Fig. 2e). In this experiment, a precision syringe pump combined with a well-defined flow chamber (µ-Slide I Luer) was used to provide stable shear stress conditions (0−8 dyn/cm2). C166 vascular endothelial cells were seeded in the flow chamber, and native or engineered K562 cells rolled and adhered to C166 cells under defined shear stress. First, K562 cells were modified with MAV or PAV at a density of 1.6 × 107 units/cell, and different shear stress conditions were used in the cell adhesion assay. The results revealed that the number of adherent K562 cells in the PAV group was higher than that in the MAV group under all shear stress conditions (Fig. 2f). As the shear stress increased from 2 dyn/cm2 to 4 dyn/cm2, the ratio of the number of adherent cells in the PAV group to the MAV group increased from 2- to 3-fold, suggesting that polyvalent engineered K562 cells had better adhesion at higher shear stress. Under the highest shear stress condition of 8 dyn/cm2, MAV mediated negligible cell adhesion. In comparison, PAV-engineered cells could adhere to C166 cells. We then fixed the shear stress at 4 dyn/cm2 and gradually increased the density of MAV or PAV on the cell surface (Fig. 2g, h). Consistent with the previous results, the number of adherent K562 cells in the PAV group was higher than that in the MAV group for all density conditions. Collectively, we demonstrated the successful construction of PAV on the cell surface using a DNA-templated protein assembly strategy, and this polyvalent engineering technology significantly enhanced the adhesion of the modified cells to vascular endothelial cells under static and dynamic conditions.

Adhesion and migration of engineered MSCs under shear stress conditions

Encouraged by the compelling results from experiments on K562 cells, we proceeded to produce engineered MSCs via the same strategy. Confocal fluorescence images showed strong FITC signals localized on the cell membrane (Fig. 3c). Flow cytometry indicated an 8-fold increase in the fluorescence intensity in the PAV group compared to the MAV group (Fig. 3b). These results demonstrated the successful preparation of PAV-engineered MSCs.

To study the adhesion and transendothelial migration of engineered MSCs under flow conditions, we used a 3D flow chamber to simulate an in vivo-like blood vessel. The bottom of the chamber was lined with collagen I containing SDF-1α inside, and C166 cells were cultured on the gel matrix to generate an activated endothelial monolayer (Fig. 3a). The microfluidic device mimicked a blood vessel with inflammation. Then, MSCs were modified with MAV or PAV at a density of 1.6 × 107 units/cell and perfused into the flow chamber under defined shear stress conditions. As shown in Fig. 3d, MAV and PAV engineering enhanced the adhesion of MSCs to target C166 cells. We specifically quantified the increase in the efficiency of cell adhesion mediated by engineered MSCs compared to native MSCs (Fig. 3e). Under a shear stress of 2 dyn/cm2, cell engineering with MAV enhanced the adhesion efficiency by 230%, whereas the enhancement efficiency in the PAV group was 560%. When the shear stress was increased from 2 dyn/cm2 to 8 dyn/cm2, the ratio of enhancement efficiency in the PAV group to the MAV group further increased from 2.4- to 3.8-fold. The sharp increase in the enhancement efficiency suggests that polyvalence can lead to a higher degree of cell adhesion than monovalence, especially under higher shear stress conditions. We also varied the modification density of MAV or PAV on the cell surface and compared the difference between monovalent and polyvalent engineering on MSC adhesion. As expected, cell engineering with PAV resulted in a higher adhesion efficiency for all modification density conditions (Supplementary Fig. 6 and Fig. 3f).

After demonstrating the successful adhesion of MSCs to C166 cells under shear flow conditions, we investigated whether these MSCs migrated toward SDF-1α in the chamber. Confocal microscopy images were taken at 0 h and 24 h to track the migration of MSCs in the 3D flow chamber. It was clear from the 3D imaging results that native and engineered MSCs could migrate from the endothelial layer into the gel matrix at 24 h (Fig. 3g). Quantification analysis revealed that the number of migrating cells was higher in the PAV group than in the control and MAV groups (Fig. 3h, i). These results suggest that surface engineering does not affect MSC migration across the vascular endothelium and that polyvalent engineering increases the number of migrating cells by enhancing adhesion efficiency. We also found that most of the antibodies modified on the engineered cells were shed during migration (Supplementary Fig. 7). Taken together, above results demonstrate that surface engineering of MSCs with PAV holds great potential for enhancing cell adhesion to the endothelial layer and transendothelial migration under a physiological shear‒stress range in vascular microcirculation.

Effect of surface engineering on MSC functions and in vivo biosafety

Before proceeding to the in vivo study, we first investigated whether cell surface engineering affected MSC functions. MSCs were modified with MAV or PAV at densities of 1.6 × 107 and 2.4 × 107 units/cell, and we tested cell viability, proliferation, adhesion, and paracrine effects. The results of live/dead staining showed that native and engineered MSCs exhibited a steady growth state with almost no signs of cell death (Supplementary Fig. 8a). The CCK-8 assay indicated that engineered MSCs exhibited comparable proliferation potential relative to native MSCs (Supplementary Fig. 8b). In addition, engineered MSCs could effectively adhere to tissue culture plastic substrate with no abnormalities in morphology (Supplementary Fig. 8c). Increasing evidence indicates that MSCs exert their beneficial effects mainly through the secretion of factors. There are several paracrine cytokines released by MSCs that are involved in tissue repair, such as vascular endothelial growth factor (VEGF), fibroblast growth factor (FGF), interleukin 6 (IL-6), stromal cell-derived factor 1 (SDF-1), monocyte chemoattractant protein 1 (MCP-1), transforming growth factor-beta (TGF-β1), and hepatocyte growth factor (HGF)36. Therefore, we further tested the paracrine effect using ELISA. As shown in Supplementary Fig. 9a, in comparison to that of native MSCs, the expression level of several paracrine cytokines in engineered MSCs showed no discernable variation after 72 h of culture, which indicated that cell surface engineering did not affect the secretory behaviors of MSCs. Taken together, our results revealed that the essential function of MSCs was not considerably changed by cell surface engineering.

Considering that DNA may induce an adaptive immune response37,38, we further tested the immunogenicity of the engineered MSCs in vivo. We collected whole blood and spleens from mice after intravenously administering engineered MSCs and analyzed T-cell activation and cytokine release. As shown in Supplementary Fig. 9b–f, neither T-cell activation nor inflammatory cytokine release was observed, suggesting that these engineered MSCs did not induce adaptive immune responses in vivo. Next, the biosafety of the engineered MSCs was further assessed after 14 days of treatment. The results showed that mouse body weight was not abnormal within 14 days. In addition, histological examination showed that MSC treatment did not induce any toxic effects or tissue damage (Supplementary Fig. 10). The major organs, including the heart, liver, spleen, lungs, and kidneys, did not show visible damage, as indicated by H&E-stained sections. These results demonstrate that the engineered MSCs possess excellent biosafety in mice.

In vivo targeting properties of engineered MSCs

After demonstrating the in vivo biosafety of engineered MSCs, we evaluated whether PAV engineering could promote more efficient delivery of MSCs to sites of inflammation in vivo. We established a mouse model of acute local ear inflammation by intradermally injecting bacterial-derived lipopolysaccharide into the pinna and used saline in the contralateral ear as a control39. This design enabled the quantitative comparison of MSC delivery efficiency between inflamed and noninflamed contralateral tissue within the same mouse (Fig. 4a)40. The successful establishment of the mouse model was verified by immunofluorescence staining, which showed that VCAM1 expression was significantly upregulated in the blood vessels of the inflamed ear compared to the control ear (Supplementary Fig. 11).

MSCs were labeled with Vybrant-DiD and administered intravenously to the mice 6 h after LPS injection, and IVIS imaging was subsequently performed at four time points to track MSC trafficking to inflamed tissue. Inflamed and control ears from the same animal were imaged in pairs to allow direct comparisons. From 12 h to 72 h after administration, PAV-engineered MSCs showed a dramatic increase in homing to the inflamed ear compared to native or MAV-engineered MSCs (Fig. 4b). The highest fluorescence intensity in the PAV group occurred at 48 h after administration, which was 3.5-fold higher than that in the MAV group (Supplementary Fig. 12). We also used intravital confocal microscopy to image the vasculature in the inflamed ear. PAV-engineered MSCs effectively adhered to inflamed blood vessels (Supplementary Fig. 13), suggesting that the improved homing efficiency was mediated by enhanced cell adhesion to the activated endothelium. To obtain more quantitative information, we performed flow cytometric analysis of cells collected from mouse ears. The results showed that 5.4 % of PAV-engineered MSCs eventually reached the target tissue, which was 3.2 times higher than that in the MAV group and 6.6 times higher than that in the native group (Fig. 4c), which was consistent with the IVIS imaging results. In addition, there were no differences in the numbers of native and engineered MSCs localized within the noninflamed ear, suggesting that PAV-mediated MSC homing is specific. We compared the difference between MAV and PAV engineering on MSC homing efficiency. Engineering with MAV and PAV enhanced the homing efficiency of cells by 110% and 560%, respectively (Fig. 4d). These quantitative data demonstrated that cell surface engineering with multivalent anti-VCAM1 enabled MSCs to be specifically and efficiently delivered to inflamed tissue in vivo. Finally, we used immunofluorescence staining to examine whether the engineered MSCs could extravasate across the endothelium into surrounding tissues (Fig. 4e). Fluorescence images revealed that the majority of engineered MSCs were located outside the vessel lumen or colocalized with the endothelial layers, indicating that engineered MSCs retained the ability of transendothelial migration in vivo.

We subsequently evaluated the in vivo biodistribution of engineered MSCs, which demonstrated a predominant accumulation in the lung, liver and spleen, with minimal presence noted in the heart and kidney (Supplementary Fig. 14a, b). One obstacle for intravenous administration of therapeutic cells is that the cells can get trapped in the lung capillaries41. Interestingly, IVIS imaging and flow cytometry showed reduced entrapment of engineered MSCs in the lung compared to native MSCs (Supplementary Fig. 14c). Although the specific mechanism is not clear, the reduction in lung entrapment contributed to the improvement in MSC homing efficiency. In addition, it should be emphasized that while MAV- and PAV-engineered MSCs had similarly low levels of lung entrapment, PAV-engineered MSCs showed 3.2 times more homing efficiency to the inflamed ear than MAV-engineered MSCs, suggesting that polyvalent engineering enhanced MSC homing to inflammation sites in vivo.

Therapeutic efficacy of engineered MSCs against IBD

We further tested the therapeutic efficacy of engineered MSCs in a disease model of IBD. Various studies have shown that transplanted MSCs exert their therapeutic effects on IBD through immunomodulation and angiogenesis and represent a promising alternative treatment42,43. Here, we established a dextran sulfate sodium (DSS)-induced colitis model to test the ability of engineered MSCs to provide enhanced therapeutic benefits in mice with colitis44. As shown in Fig. 5a, normal C57 mice were given 3.5% DSS for 7 consecutive days after one week of acclimatization to create a colitis model. On Day 7, MSCs were intravenously administered. Additionally, we included control groups that received PBS, MAV, or PAV treatment45. We first examined the adhesion of engineered MSCs in the colons of mice at 48 h after intravenous administration. IVIS images and microscopic observations of tissue sections indicated that, compared with native and MAV-engineered MSCs, PAV-engineered MSCs showed the strongest targeting capacity to bowel tissue (Fig. 5b, Supplementary Fig. 15). Quantitative analysis showed that the homing efficiency for PAV-engineered MSCs was 4.7% (Fig. 5c), which was 3.5 times higher than that in the MAV-MSCs group and 6.0 times higher than that in the native group. Biodistribution assays in mice with colitis, consistent with findings in mouse model of acute ear inflammation, indicated that the engineered MSCs were less entrapped in the lung compared to native MSCs (Supplementary Fig. 16).

Then, we evaluated the therapeutic effects of engineered MSCs by monitoring the body weight and disease activity index (DAI, including the consistency index, weight loss index, and fecal bleeding index) of each group of mice. Compared to mice in the normal group, mice in the PBS group showed obvious weight loss and sustained increases in DAI scores, indicating the successful establishment of the colitis model. However, mice that received treatment showed varying degrees of improvements in the disease, indicating the gradual restoration of intestinal function after treatment. Mice in the PAV-MSCs group exhibited superior weight recovery and lower DAI scores compared to other groups, with statistically significant differences observed between the PAV-MSCs and MAV-MSCs groups (Fig. 5d, e). We also examined colon length, which is an essential index to evaluate the treatment efficacy of colitis in mice. As shown in Fig. 5f, g, mice treated with PAV-engineered MSCs had significantly longer colons than those in other treatment groups, suggesting that PAV-engineered MSCs effectively promoted the repair of damaged bowel tissue.

In addition, various inflammation-related mediators in bowel tissue have been measured to better understand the biological mechanism of engineered MSCs. Myeloperoxidase (MPO) activity, which correlates with the level of neutrophil infiltration in the colon46, was significantly elevated in the PBS group and was suppressed to varying degrees after treatment. In particular, MPO activity in the PAV-MSCs group was decreased close to the level of the normal group (Fig. 5h). Moreover, PAV-MSCs group exhibited a statistically significant decrease in the levels of proinflammatory cytokines such as IL-6 and TNF-α, and a statistically significant increase in the level of the anti-inflammatory cytokine IL-10 compared to MAV-MSCs group (Fig. 5i–k). We then examined histological sections. H&E staining revealed inflammatory infiltration, mucosal destruction, and structural disruption of crypt foci in colitis mice in the PBS group. However, mice treated with PAV-engineered MSCs showed virtually normal pathological structures, suggesting that PAV-engineered MSCs ameliorated histological damage caused by DSS (Fig. 5l).

Notably, we found that treatment with antibodies alone led to a modest reduction in colitis severity, with no discernible variance in therapeutic outcomes between PAV and MAV (Fig. 5d–l). By contrast, PAV-MSCs could promote tissue repair more effectively. These results suggest that the primary therapeutic impact within the PAV-MSCs construct can be attributed to the MSC component, and the strategic incorporation of PAV modification on the MSC surface enhance the delivery of MSCs to damaged tissues, consequently yielding a more potent therapeutic effect. The above assessment further highlights the pivotal role of surface modification in bolstering the overall therapeutic efficacy of PAV-MSCs. Furthermore, the enhanced targeting efficiency and therapeutic efficacy of PAV-engineered MSCs were demonstrated through comparison with several engineered MSCs previously reported in the literature for intravenous injection into mice with colitis (Supplementary Table 3). Taken together, the in vivo results demonstrate that cell surface engineering with PAV can enhance the therapeutic effect of MSCs by enhancing MSC adhesion to damaged tissue and has the potential for significant applications in a wider range of diseases.