Ndrg2 expression is reduced in tolerogenic DCs
To identify potential targets for gene editing in DCs, we compared transcriptomic profiles of tolerogenic DCs, which confer a variety of clinical benefits,11-17 with untreated DCs. Bone marrow-derived DCs were cultivated from wild-type (WT) mice (C57/BL6) according to standard protocols. Vitamin D3 (calcitriol, VD3) was used to induce tolerogenic DCs, and scRNA-seq of VD3 stimulated and untreated DCs was performed (Fig 1a, b). Nine transcriptionally distinct clusters resembling pre-mature (characterized by high expression of Cd34, Csf1r (encoding CD115), Ccr2, and Cx3cr125) and mature BM-DCs (characterized by high expression of Flt3 (encoding CD135)26 and Ccr727,28) were identified (Fig 1b, Extended Data Fig 1a). We found that tolerogenic DCs induced by VD3 showed a downregulation of Ndrg2, which has previously been reported to inhibit growth factor expression, angiogenesis, and cell proliferation.29-33 In accordance, tolerogenic DCs showed an overexpression of vascular endothelial growth factor (Vegfa), indicating pro-angiogenic properties potentially related to Ndrg2 inhibition. Thus, we identified Ndrg2 as an interesting target for further investigation (Fig 1c, Extended Data Fig 1b).
Ndrg2 marks DC progenitors
Among the untreated DCs, the strongest Ndrg2 expression was found in cluster 8 (C8, Fig 1d, Extended Data Fig 1d), which co-expressed the hematopoietic stem cell marker Cd34, and Csf1r (encoding CD115), as well as Flt3 (encoding CD135) and Clec9a (encoding DNGR-1). This expression profile is consistent with previously defined myeloid progenitor populations, namely MDP (macrophage and dendritic cell progenitor) and CDP (common dendritic cell progenitor), for which phenotypic overlap has been described.34-36 Interestingly, Ndrg2+ progenitors also expressed Itage (CD103) while being negative for Cd8 (Fig 1e, Extended Data Fig 1d).
Compared to untreated cells, VD3 treatment almost completely abrogated the expression of Ndrg2 and its co-expressed genes Cd34, Flt3, Itgae, and Clec9a, while inducing Kit expression (Fig 1e, Extended Data Fig 1d). This was further confirmed with flow cytometry showing that 3% of all CD34+ DC progenitors in untreated cultures co-expressed Flt3 (CD135) and CD103, while this population was reduced to 0.8% after VD3 treatment (Fig 1f).
Ndrg2 inhibition in DCs promotes EC tube formation in co-cultures
To investigate whether inhibition of Ndrg2 could promote a pro-angiogenic behavior in DCs, we co-cultured BM-DCs with endothelial cells (ECs) and analyzed their angiogenic capacity in EC tube formation assays. Ndrg2 was inhibited according to established protocols using VD3 and lipopolysaccharide (LPS) or dexamethasone (Fig. 1g).29 We found that Ndrg2-inhibited DCs stimulated EC tube formation in vitro, as indicated by a higher EC branch and junction number, longer branch length, and higher total mesh area compared to EC co-cultures with untreated DCs and ECs cultured in basal media (Fig. 1i, Extended Data Fig. 1e, f). The strongest angiogenic response was found in co-cultures with VD3-treated DCs (Fig. 1i, h). This was confirmed in a Luminex multiplex ELISA for 43 proteins performed on culture media of DC monocultures, which showed that DCs treated with VD3 had the highest VEGFA secretion compared to untreated DCs, DCs treated with LPS, or even DCs treated with VD3 and LPS (Fig. 1j).
Given the strong expression of Ndrg2 in DC progenitors and the pro-angiogenic response created by pharmacologic Ndrg2 inhibition, we hypothesized that KO of Ndrg2 would allow us to alter the transcriptional fate and development of BM-DCs, thus allowing us to generate cells that can favorably impact the wound healing process. Furthermore, precise KO of Ndrg2 would avoid pleiotropic effects of VD3, in which unpredictable off target or systemic effects could occur and would also precisely interrogate whether the observed pro-angiogenic features of tolerogenic DCs were in fact due to loss of Ndrg2 expression.
Development of a CRISPR/Cas9 platform for KO of Ndrg2 in primary DCs
To develop an approach for CRISPR/Cas9 KO of Ndrg2 in primary murine BM-DCs, we used nucleofection with ribonucleoprotein complexes assembled from Cas9 protein and sgRNA (Cas9-RNP; Fig. 2a,b). This technique for CRISPR/Cas9 gene editing has recently been demonstrated to be very effective for editing primary murine and human DCs37,38 and has several advantages compared to previously employed lentiviral vector based approaches.39 To achieve the highest editing efficiency, we compared indels and KO rates after Cas9-RNP nucleofection of DCs using two sets of three sgRNAs from different manufacturers targeting the Ndrg2 genomic locus as well as different numbers of cells per reaction and different RNP concentrations (Extended Data Fig. 2a, Table S1). After optimization, using GFP-fused Cas9 protein (dCas9-GFP) and flow cytometry, we confirmed a >90% transfection efficiency of DCs with our approach for nucleofection (Fig. 2c). Sanger Sequencing and the Interference of CRISPR Editing (ICE) tool (Synthego) were used to compute indel and KO rates, which were 91% resp. 88% using a triple-guide RNP approach (Fig. 2d, Extended Data Fig. 2a, Methods). We also performed immunocytochemical staining of CRISPR-edited and untreated DCs in vitro and observed a significant reduction in Ndrg2 protein expression (Fig. 2e).
Gene editing with Cas9-RNP showed no observable off-target effects
To analyze specific mutations introduced by CRISPR gene editing in DCs at the target site as well as any potential off-target effects, we performed whole genome sequencing (WGS) with an ultrahigh depth (~100X coverage) of CRISPR-edited DCs using our triple-guide RNP approach (Fig. 2f). Wild-type DCs were analyzed as controls. A variety of different on-target effects were identified, the most common of which were upstream gene variants, followed by 5’ UTR variants (Fig. 2g). Cas-OFFinder was used to identify potential off-target sites, which allows for unbiased extensive mutation searching. Importantly, indels were not detected within 15 base pairs (bp) up- and downstream of the sites predicted by Cas-OFFinder40. When each site was broadened to 200bp up- and down-stream, indels in 14 sites were identified (Fig. 2h, Table S2). By manual inspection of these loci, we found that only 7 loci showed indels that occurred only in the Ndrg2-KO group and not in the control group. After identifying the most likely off-target sites, we further aligned the sequence between the predicted Cas-OFFinder cut sites and the three sgRNA sequences targeting Ndrg2. Among the three sites that showed the highest alignment, we only identified a protospacer adjacent motive (PAM) in two of the sites. All other predicted off-target sites did not show similarity to the sgRNA sequences and therefore are unlikely true off-target sites (Extended Data Fig. 2b, c). This set of analysis together confirmed that the designed guide RNAs and the use of RNP in DCs cells had little or no observable off-target effects.
Ndrg2-KO prevents DC maturation and induces regenerative gene expression profiles
To assess the impact of Ndrg2-KO on the transcriptional fate of DCs, we performed scRNA-seq of CRISPR-edited DCs and compared their transcriptomic profiles with those of untreated DCs and DCs treated with VD3, which pharmacologically inhibits Ndrg2 (Fig. 1a). A total of 23,871 cells were analyzed (Fig. 3a). We employed RNA velocity analysis using dynamical modeling with the scVelo package to determine changes in DC differentiation in response to Ndrg2-KO41. A differentiation stream was identified originating from the Ndrg2+ progenitors (cluster 9) that progressed through clusters 1 and 3 and ended in the mature DCs in cluster 0 (Fig. 3a, b). Most Ndrg2-KO cells were found in cluster 1, whereas control and VD3 treated cells mostly localized to cluster 0 (Fig. 3c). Using CytoTRACE, we compared the relative differentiation states of individual cells based on the distribution of unique mRNA transcripts and found that cluster 1, which mostly contained Ndrg2-KO cells, was among the least differentiated clusters, whereas cluster 0, which contained mostly VD3 treated and untreated cells, showed more advanced differentiation states (Fig. 3d)42. In accordance, KO of Ndrg2 led to a downregulation of the DC maturation markers Cd80 and Cd83 and an upregulation of Csf1r, which marks premature DCs (Fig. 3e). Hence, our findings indicate that Ndrg2 is highly expressed in DC progenitor cells and promotes DC differentiation toward mature phenotypes. In addition to a reduction of global Ndrg2 expression (Extended Data Fig. 3a), Ndrg2-KO using our Cas9-RNP approach almost eliminated cluster 9, which contained the Ndrg2+ progenitors and thereby halted DC maturation, preserving a pre-mature DC state (Fig. 3f). These engineered pre-mature DCs exhibited an even stronger Vegfa expression than DCs treated with VD3, and in addition exhibited an up-regulation of extracellular matrix associated genes that promote wound healing, such as Fn1 (encoding fibronectin-1) and Mmp12 (matrix metalloproteinase-12), as well as anti-oxidative genes Prdx4 and Mgst1 (Fig. 3g, Extended Data Fig. 3b). By contrast, control DCs highly expressed pro-inflammatory markers, such as S100a6 and S100a4 (Fig. 3h).33
To investigate differential regulation of cellular signaling pathways between our conditions, we then used GeneTrail 3, a computational pipeline for over-representation analysis (ORA) of specific gene sets on a single-cell level, and found a significant upregulation of pathways related to angiogenesis, epithelial cell migration, and degradation of the ECM in Ndrg2-KO cells, indicating a beneficial impact on wound healing (Fig. 3i).43
When analyzing the expression levels of genes located at the aforementioned potential off-target sites, Amy1, Ftsj1, Sema5a, and Minar1 (Table S2), no detectable expression among all groups was found for Sema5a and Minar1. Ftsj1 and Amy1 were expressed at very low levels in less than 6% resp. 0.06% of all cells without significant differences, confirming that our approach for gene editing is highly specific and does not cause significant off-target effects (Extended Data Fig. 3c).
Hydrogel delivery of Ndrg2-KO DCs promotes healing of non-diabetic wounds
Having developed and characterized our approach for gene editing to specifically induce regenerative gene expression profiles in DCs, we next aimed to test the suitability of these cells as a cell-based therapy for wound healing. The effectiveness and biocompatibility of hydrogels for cell delivery has been shown in murine and porcine wound models in combination with multiple cell types.5,44-48 We seeded engineered DCs into a previously developed pullulan-collagen hydrogel to facilitate effective cell delivery onto the wound bed.44,48 High viability and uniform distribution within the hydrogel were confirmed using calcein AM staining of live cells cultured on these hydrogels for 7 days (Fig. 3j).
To determine the therapeutic potential of Ndrg2-KO-DCs, we treated splinted full-thickness excisional wounds in wild-type mice (C57BL6) with hydrogels seeded with either control or Ndrg2-KO DCs. For controls, we subjected DCs to the same Cas9-RNP protocol, using 3 non-targeting sgRNAs to account for any potential effect of the Cas9-RNP or the nucleofection itself on wound healing. We also used a second control group of wounds treated with blank, unseeded hydrogels (Fig. 4a). Wounds treated with our engineered Ndrg2-KO DCs healed significantly faster and were completely re-epithelialized by day 11, 5 days faster than in either control group (Fig. 4b-d). Histologically, healed wounds treated with Ndrg2-KO-DCs showed a fully regenerated epithelium and a thicker dermis compared to wounds treated with control DCs or blank hydrogels (Fig. 4c-e). Immunofluorescent staining for CD31 revealed a significantly stronger vascularization of healed wound tissue that had been treated with engineered DCs compared to healed wound tissue from the control groups, confirming that the engineered DCs promoted angiogenesis and accelerated wound healing (Fig. 4f).
Hydrogel delivery of Ndrg2-KO DCs promotes healing of diabetic wounds
We then investigated whether our novel DC therapy would also be beneficial for chronic wounds with impaired healing potential, such as in diabetes mellitus. Therefore, we treated excisional wounds in db/db mice with either Ndrg2-KO-DCs, control DCs (nucleofection with 3 non-targeting sgRNAs), or blank hydrogels (n = 5 per group, Fig. 4g). The db/db mice have a homozygous mutation in the leptin receptor gene (Lepr), resulting in morbid obesity, chronic hyperglycemia, and pancreatic beta cell atrophy. It has been shown that excisional wounds in db/db mice show a significant delay in wound closure, decreased granulation tissue formation, severely impaired vascularization, and reduced cell proliferation compared to wounds in WT and other diabetic mouse strains.49 In accordance with our findings in WT mice, Ndrg2-KO DCs significantly accelerated wound healing in db/db mice, leading to complete re-epithelialization by day 16, five days earlier than in the control groups (Fig. 4h, i, Extended Data Fig. 4a, b) and similar to what we had observed in WT mice. The engineered DCs also promoted angiogenesis, similar to the effects seen in the WT mouse experiments, significantly improving vascularization of diabetic wound tissue compared to control DCs and unseeded hydrogels (Fig. 4j).
Ndrg2-KO DCs target wound fibroblasts via growth factor signaling
We next aimed to further elucidate the molecular mechanisms underlying the accelerated wound healing and increased neovascularization in response to engineered DC therapy. To interrogate how the transplanted DCs interact with different cell types of the wound healing environment and affect their individual transcriptional signatures, we performed scRNA-seq of wound tissue treated with Ndrg2-KO-DCs, control DCs, or blank hydrogels (Fig. 5a). Tissue was explanted from a second batch of mice (n = 5 per group) on day 10 after wounding, when the wound healing curves of the different groups began to diverge (Fig. 4d), indicating critical differences in wound biology. We performed scRNA-seq on 10,612 cells across the three conditions, which were identified as fibroblasts, myeloid cells, neutrophils, lymphoid cells, and erythrocytes (Fig. 5b, c). Fibroblasts had the highest number of differentially expressed genes (DEGs, FC > 0.5, p< 0.05) between the groups among all cell types, indicating that our engineered DC therapy had the strongest impact on fibroblast gene expression within the wound bed (Fig. 5d, e). Among the most differentially expressed genes expressed by fibroblasts, we found several genes that have been shown to be critically involved in wound healing. Fibroblasts from wounds treated with Ndrg2-KO DCs almost exclusively expressed Ngfr, the receptor for nerve growth factor, which has been shown to promote cell proliferation, angiogenesis, and wound healing (Fig. 5f, g).50 Moreover, treatment with Ndrg2-KO DCs strongly upregulated the expression of Lgals1 (encoding galectin-1) in fibroblasts, which promotes wound healing in diabetic and non-diabetic wounds.51,52 Interestingly, we also found an upregulation of Plod2, encoding lysyl hydroxylase 2, which catalyzes the first step of collagen cross linking, and Ppib, encoding peptidyl prolyl cis/trans isomerase, which catalyzes the folding of ECM proteins such as collagen. These findings indicate that wound fibroblasts may be activated by Ndrg2-KO DCs to increase ECM protein synthesis and collagen cross linking in the proliferative phase of wound healing, leading to accelerated wound closure.
To study the intercellular communication networks between the transplanted DCs and the local cells of the wound, we integrated scRNA-seq data from Ndrg2-KO and control DCs (Fig. 3a) with transcriptomic data from the respective local wound environments (Fig. 5b) and used CellChat to analyze ligand-receptor interactions.53 We found that Ndrg2-KO DCs showed a stronger outgoing signaling activity within their wound environment compared to control DCs (Fig. 5h) and also demonstrated a stronger total interaction strength within the aggregated cell-cell communication network (Fig. 5i). CellChat analysis demonstrated significant VEGF signaling from Ndrg2-KO DCs to fibroblasts, which was not present in control DCs, confirming our previous observations (Fig. 5j). In contrast to control DCs, Ndrg2-KO DCs also communicated with wound fibroblasts via other growth factor pathways, such as the insulin-like growth factor (IGF) and platelet-derived growth factor (PDGF) pathways (Fig. 5l, m), which stimulate cell proliferation and angiogenesis, and showed an upregulation of Spp1 (osteopontin) signaling, which has been shown to be a critical driver of early wound repair in diabetic wounds (Fig. 5m).54-56