Selective reduction of splenic cDC1s in DC-specific Zeb1-deficient mice
The Zeb protein family of transcription factors consist of two members Zeb1 and Zeb2, which are best known to drive epithelial to mesenchymal transition (EMT) by repressing epithelial genes30. Zeb1 and Zeb2 are widely expressed in murine myeloid immune cells, but the expression of Zeb1 in cDC1s is higher than that of Zeb2 31, 32. It has been demonstrated that Zeb2 switches the DC fate specification from cDC1 to pDC or cDC2 by antagonizing Id2 expression33, 34, 35. High association of Zeb1 binding motif with increased chromatin accessibility in cDC1s suggested a potential role for Zeb1 in regulating some aspects of cDC1 function28. However, further investigations are needed to examine the precise role of Zeb1 in dendritic cells. To address this, we generated mice with floxed Zeb1 alleles using gene targeting strategy, in which exon 4 was flanked by LoxP sites, and then crossed them with mice harboring a transgene encoding Cre recombinase under the control of the Itgax promoter (CD11c-cre) to generate DC-specific Zeb1 conditional knockout mice (Zeb1fl/fl CD11c-cre, called Zeb1-dcKO hereinafter) (fig. S1a). We confirmed specific and efficient deletion of Zeb1 in CD11c+ dendritic cells, but not in T cells and B cells by immunoblot analysis of Zeb1 (fig. S1b). Flow cytometric analysis demonstrated that the frequencies of cDC1 populations were markedly decreased in the spleen of Zeb1-dcKO mice compared with those in wild-type (WT) littermates (Zeb1fl/fl or Zeb1+/+ CD11c-cre). The absolute numbers of both cDC1s and cDC2s were substantially reduced, due to lower abundance of myeloid cells in the spleen of Zeb1-dcKO mice (Fig. 1a-b). In contrast, the proportions and absolute numbers of cDC1s and cDC2s in other lymphoid tissues (peripheral lymph node (pLN) and thymus) and non-lymphoid tissues (liver and lung), were not affected by the absence of Zeb1 (Fig. 1a-d). In addition, the frequencies of pDCs were similar in the lymphoid tissues of WT and Zeb1-dcKO mice, nevertheless, the numbers of pDCs were dramatically decreased in the spleen of Zeb1-dcKO mice (Fig. 1e-f). All other myeloid cells were normally present in the spleen of Zeb1-dcKO mice, with the exception of a small decrease of neutrophils and inflammatory monocytes due to lower abundance of total splenic myeloid cells (fig. S1, c-d). Zeb1-dcKO mice had an increased frequency of CD8 single positive thymocytes, which was also observed in Zeb1 mutant Cellophane mice36, but possessed normal frequencies and numbers of mature T cells in the second lymphoid organs (fig. S1, e-f). This selective reduction of splenic cDC1s was further confirmed by single-cell RNA sequencing (scRNA-seq) analysis of splenic cDC populations (Fig. 1g). Taken together, these data indicated that Zeb1 deficiency in DCs selectively impeded the generation of cDC1s in the spleen but not in the other lymphoid and non-lymphoid tissues.
cDCs develop from pre-DCs, which are derived from CDPs in the BM. To determine at which stage Zeb1 regulates cDC1 development, we next analyzed pre-DC compartment in the BM and spleen. It has been reported that commitment to the cDC1 and cDC2 lineage is already apparent as Siglec-H−Ly6C− pre-cDCs develop exclusively into cDC1s whereas Siglec-H−Ly6C+ pre-cDCs give rise preferentially to cDC2s37. We did not observe significant difference in CD135+ pre-DCs and their sub-populations distinguished by Siglec-H and Ly6C expression in the BM of WT and Zeb1-dcKO mice (fig. S1, g-h). Although the frequencies of Siglec-H−Ly6C− pre-cDC1s was only mildly decreased in the spleen of Zeb1-dcKO mice, the absolute numbers of splenic pre-DCs and pre-cDC1s were notably reduced due to lower total numbers of splenic myeloid cells (fig. S1, i-j). To further investigate whether absence of Zeb1 affected the development of DC progenitors into cDC1s, we then examined in vitro development of BM cells from WT and Zeb1-dcKO mice upon stimulation with Flt3L, which generates bona fide counterparts of the splenic DC subsets38. In this setting, WT and Zeb1-dcKO BM precursors exhibited similar capacity to develop into pDCs, cDC1s and cDC2s (fig. S1, k-l), suggesting that Zeb1 might control the homeostasis rather than the development of splenic cDC1s. To explore how Zeb1 regulated the homeostasis of splenic cDC1s, we performed R package Seurat (v3.0) analysis of single-cell sequencing data. Feature plots revealed that residual Zeb1-deficient cDC1s maintained cDC1 identity characterized by the expression of cDC1 and cDC2 signature genes including Xcr1, Sirp-α, Clec9a, Irf8, etc. (fig. S2a). Nevertheless, there were 106 genes up-regulated and 49 genes down-regulated in Zeb1-deficient cDC1s (fig. S2b). Gene ontology (GO) analysis revealed that in addition to pathways of tight junction and cell adhesion, pathways of TNF signaling, apoptosis and necroptosis were also enriched in these differentially expressed genes (DEGs) (fig. S2c). We next asked whether the depletion of splenic cDC1s in Zeb1-dcKO mice might be caused by impaired survival or proliferation. By staining fresh isolated splenocytes with AnnexinV and 7-aminoactinomycin D (7-AAD), we detected a dramatically increased percentage of dying cells (AnnexinV+ 7-AAD+) in splenic cDC1s but not splenic cDC2s from Zeb1-dcKO mice (Fig. 1h-i). In contrast, the percentage of either apoptotic (active caspase-3+) or proliferating (Ki67+) splenic cDC1s was not affected by Zeb1 deficiency (fig. S2, d-g). Taken together, these results suggested that Zeb1 was required for the homeostasis of splenic cDC1s by maintaining their survival, and that loss of Zeb1 led to excessive non-apoptotic cell death in splenic cDC1s.
We next assessed whether the reduction of splenic cDC1s in Zeb1-dcKO mice was cell-intrinsic or caused by external environmental factors. To this end, we generated mixed bone marrow chimeric mice, whereby lethally irradiated B6.SJL (CD45.1+) mice were reconstituted with a mixture of C57BL/6J (B6) (CD45.2+) BM cells plus either WT (CD45.1+CD45.2+) or Zeb1-dcKO (CD45.1+CD45.2+) BM cells at a ratio of 1:1. Eight weeks after reconstitution, Zeb1-dcKO BM cells repopulated cDC populations almost equally as WT BM cells in the same recipient. However, in this competitive environment, Zeb1-dcKO (CD45.1+CD45.2+) BM cells reconstituted lower percentage of cDC1s in the spleen, similar to that in non-chimeric Zeb1-dcKO mice (Fig. 1j-k). Unexpectedly, B6 (CD45.2+) BM cells in the same recipient with Zeb1-dcKO BM cells also generated less splenic cDC1s, whereas B6 (CD45.2+) BM cells together with WT (CD45.1+CD45.2+) BM cells in the same host developed into normal frequencies of splenic cDC1s (Fig. 1j-k). Together, these results suggested that the effect of Zeb1 deletion on splenic cDC1s was cell-intrinsic and even dominant as it also affected Zeb1-sufficient cells in the same host.
Resistance of DC-specific Zeb1-deficient mice to Listeria infection
Previous studies have established that CD8α+ dendritic cells (cDC1s) are the obligate cellular entry points for productive infection by intracellular bacterium L. monocytogenes, and lack of cDC1s enhances host resistance to the infection due to the loss of access into periarterial lymphocyte sheath (PALS)39, 40. To examine the host defense of Zeb1-dcKO mice against listeria infection, we infected WT and Zeb1-dcKO mice intravenously with L. monocytogenes. All WT mice succumbed to infection with high dose (1×106 CFU) of L. monocytogenes, whereas almost all Zeb1-dcKO mice survived (Fig. 2a). There were 100 ~ 1000 folds fewer CFU of bacteria in the spleen and liver of Zeb1-dcKO mice than in WT littermates at day 2 after infection with low dose (2.5×104 CFU) of L. monocytogenes (Fig. 2b). The decreased listeria burden observed in Zeb1-dcKO mice was in complete accord with the dramatically reduced mortality. Furthermore, histological analysis showed that Zeb1-dcKO mice displayed substantially reduced lesions in the spleen and liver compared to WT littermates at day 2 after infection. Micro-abscesses with depletion of apoptotic lymphocytes or necroptotic hepatocytes and infiltration by inflammatory cells were observed in white pulp of spleen or near vessels of liver from WT mice but not from Zeb1-dcKO mice (Fig. 2c). Collectively, these data suggested loss of Zeb1 in DCs enhanced host defense against Listeria infection.
To examine the innate immune responses of Zeb1-dcKO mice to Listeria infection, we measured the secretion of inflammatory cytokines into serum of mice within 24 hours post infection by ELISA. As expected, Zeb1-dcKO mice exhibited a severe reduction in secretion of all tested serum cytokines including IL-6, IL-12, TNF-α and IFN-γ (Fig. 2d). We next asked whether the impaired production of inflammatory cytokines was caused by the decreased Listeria burden or defective innate signaling in dendritic cells. To address this issue, we challenged purified cDCs or subpopulation cDC1s and cDC2s from Flt3L-cultured BM cells with different Toll-like receptor ligands including the TLR4 ligand LPS, TLR9 ligand CpG-B DNA, TLR1/2 ligand Pam3CSK4, TLR3 ligand Poly(I:C), heat-killed L. monocytogenes, and measured cytokine production by intracellular staining or by enzyme-linked immunosorbent assay (ELISA). In both experiments, Zeb1-deficient Flt3L-cDCs or sub-population cDC1s and cDC2s produced similar amount of all tested cytokines including IL-12, IL-6 and TNF-α compare to WT counterparts (fig. S3, a-e). Taken together, these results strongly suggested that the increased resistance to Listeria infection of Zeb1-dcKO mice was likely due to the impaired transmission of Listeria into PALS, resulted from selective reduction of splenic cDC1s, as the innate immune response was intact in Zeb1-deficient cDCs.
Ablation of Zeb1 in DCs dampens antitumor immunity
A previous study suggested that L. monocytogenes may co-opt the cross-presentation pathways of cDC1s for the productive infection40. Although Zeb1-dcKO mice had a reduction of cDC1s only in the spleen but not in the other lymphoid and non-lymphoid tissues, they showed a reduced Listeria burden not only in the spleen but also in the liver. These results indicated that Zeb1-deficient cDC1s from the liver might have an impaired capability of cross presentation. Accumulating evidence has established that cDC1s are particularly adept at uptake of dead tumor cells and at cross-priming tumor-specific CD8+ T cells within tumor microenvironment or after migration to tumor draining lymph nodes (tumor dLNs)41, 42. To examine the role of Zeb1 in the function of cDC1s, we challenged WT and Zeb1-dcKO mice subcutaneously with 2×105 B16F10 melanoma cells. B16F10 tumors grew much faster and caused shorter host survival in Zeb1-dcKO mice than in WT littermates (Fig. 3a-b). Similar to unchallenged Zeb1-dcKO mice, tumor-bearing Zeb1-dcKO mice also had decreased frequencies and numbers of splenic cDC1s, as compared with tumor-bearing WT mice (fig. S4, a-b). By contrast, tumors from WT and Zeb1-dcKO mice displayed comparable composition of cDC1s and cDC2s (Fig. 3c-d). Likewise, tumor dLNs from WT and Zeb1-dcKO mice contained similar numbers of resident and migratory cDC1s (Fig. 3e-f). Collectively, these data suggested that Zeb1 in DCs was required for control of tumor growth, although loss of Zeb1 in DCs did not affect tumor infiltration of cDC1s despite selective reduction of splenic cDC1s.
In line with the increased tumor growth, Zeb1-dcKO mice showed a strong decrease in the recruitment and activation of CD8+ T cells in B16F10 tumors despite the normal infiltration of cDC1s in tumors (Fig. 3g-h). Correspondingly, the frequencies and numbers of IFN-γ-, Granzyme B- and perforin-producing CD8+ T cells were dramatically reduced in B16F10 tumors from Zeb1-dcKO mice, whereas the frequencies and numbers of IFN-γ-producing CD4+ T cells were unaffected (Fig. 3i-j). As such, deletion of Zeb1 in DCs resulted in decrease of both number and effector function of CD8+ T cells in B16F10 melanoma tumors (Fig. 3g-j). The numbers of splenic CD4+ and CD8+ T cells were slightly but significantly declined in tumor-bearing Zeb1-dcKO mice, consistent with the decreased numbers of splenic cDC1s (fig. S4, c-d). In contrast, the CD4+ and CD8+ T cells in tumor dLN remained low activation and their numbers were comparable between WT and Zeb1-dcKO tumor-bearing mice (fig. S4, e-f). These results suggested that either cDC1-mediated priming of tumor infiltrating CD8+ T cells did not occur at tumor dLN, or most of activated CD8+ T cells in tumor dLN migrated into tumor site. Taken together, these results demonstrated that Zeb1 expression in DCs was required for the recruitment and activated of tumor infiltrating CD8+ T cells, and further indicated that loss of Zeb1 in DCs might attenuate cross-presentation of tumor antigens to CD8+ T cells by cDC1s.
Deletion of Zeb1 abrogates the presentation of exogenous antigen by cDC1s
Given that Zeb1 expression in DCs was required for activation of tumor infiltrating CD8+ T cells, we next examined the role of Zeb1 in cross-presentation of cell-associated antigens. To address this issue, we transferred carboxyfluorescein diacetate succinimidyl ester (CFSE)-labeled OT-ⅠT cells to B6 recipient mice followed by immunization with irradiated ovalbumin (OVA)-loaded β2m−/− splenocytes. At day 3 after immunization, OT-Ⅰ T cells showed much stronger proliferation in both spleen and pLNs when transferred into WT mice than those transferred into Zeb1-dcKO littermates (Fig. 4a-b). As loss of Zeb1 in DCs caused reduction of cDC1s only in the spleen but not in the pLNs, these data suggested that in vivo cross-presentation of cell-associated antigens by cDC1s from pLNs was attenuated by the absence of Zeb1. To directly evaluate the effect of Zeb1 deficiency on cross-presentation in cDC1s from LNs, we purified cDC1s from pLNs and mesenteric LNs (mLNs) of WT and Zeb1-dcKO mice and test their ability to cross-present bacteria-associated antigen in the form of heat-killed L. monocytogenes expressing ovalbumin (HKLM-OVA) to OT-ⅠT cells in vitro. cDC1s from pLNs and mLNs of WT mice induced strong T cell proliferation by cross presentation of HKLM-OVA in a dose-dependent manner. However, cDC1s from pLNs and mLNs of Zeb1-dcKO mice induced significantly less T cell proliferation especially at low dose of bacteria (Fig. 4c-d). These results confirmed that Zeb1 promoted cross-presentation of cell-associated and bacteria-associated antigens in cDC1s.
To gain comprehensive insight into the role of Zeb1 in antigen presentation, we sorted cDC1s and cDC2s from BM cells cultured with Flt3L and performed in vitro antigen presentation assays. Compared to WT Flt3L-cDC1s, Zeb1-deficient Flt3L-cDC1s exhibited a striking defect in in vitro cross-presentation of both cell-associated and bacteria-associated antigens that were presented by neither WT nor Zeb1-deficient Flt3L-cDC2s (Fig. 4e-h). Furthermore, Zeb1-deficient cDC1s showed partially reduced efficiency for cross presentation of soluble OVA compared to WT cDC1s, whereas WT and Zeb1-deficient cDC2s equally cross-presented soluble OVA (Fig. 4i-j). In addition, the direct presentation of processed antigen SIINFEKL peptide to OT-ⅠT cells was equally efficient in WT and Zeb1-deficient cDC1s, also in both genotypes of cDC2s, suggesting that MHC-Ⅰlevels were normal in Zeb1-deficient cDCs (Fig. 4k-l). The defect of antigen presentation in the absence of Zeb1 was not limited to cross-presentation to OT-ⅠT cells by cDC1s. The presentation of soluble and bacteria-associated antigen through MHC-Ⅱwas also dramatically decreased in Zeb1-deficient cDC1s (fig. S5, a-b), while the presentation of these antigens as well as cell-associated antigen through MHC-Ⅱwas intact in Zeb1-deficient cDC2s (fig. S5, c-e). Collectively, these results suggested that Zeb1 expression in DCs was required not only for cross-presentation of exogenous antigens to CD8+ T cells but also for presentation of these antigens to CD4+ T cells by cDC1s.
Zeb1 orchestrates transcriptional program that favors cross-presentation
To explore the mechanism by which Zeb1 regulated cross-presentation, we performed high throughput RNA sequencing (RNA-seq) of WT and Zeb1-deficient Flt3L-cDC1s at steady state and after stimulation with HKLM-OVA. Principal component analysis showed that HKLM-OVA stimulation enlarged the transcriptomic diversification from the steady state in WT and Zeb1-deficient cDC1s (fig. S6a). A comparison of WT and Zeb1-deficient cDC1 transcriptomes identified 246 genes activated and 364 genes repressed by Zeb1 at steady state (fig. S6b and Table S1), whereas identified 1,167 genes activated and 1,084 genes repressed by Zeb1 after stimulation (Fig. 5a and Table S2). Functional enrichment analysis of these DEGs revealed that the most enriched Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways including cell adhesion molecules, tight junction and focal adhesion, were overactivated in steady Zeb1-deficient cDC1s relative to WT cDC1s (fig. S6c), consistent with the well-established function of Zeb1 in repressing epithelial genes30. In addition to the pathway of cell adhesion molecules, the pathways of antigen processing and presentation, lysosome and phagosome were enriched or became more significantly enriched after stimulation with HKLM-OVA (Fig. 5b). Gene Set Enrichment Analysis (GSEA) further uncovered that most of the genes in the pathways of antigen processing and presentation (Fig. 5c), phagosome (Fig. 5d) and to a less extent in the lysosome pathway (Fig. 5e) were down-regulated in Zeb1-deficient cDC1s after stimulation. Among the down-regulated genes in the pathways of antigen processing and presentation, and phagosome, genes including H2-T22, H2-M2, H2-Q4, etc. that encode MHC-Ⅰb molecules might not be the causative factors for defective cross-presentation, as they have functions other than antigen presentation to conventional CD8+ T cells43 (Fig. 5f). Two notable genes, Cybb and Ncf2, which encode subunits of Nox2, in the phagosome pathway, were down-regulated in Zeb1-deficient cDC1s after stimulation. Strong reduction of Cybb and slight reduction of Ncf2 were further confirmed by immunoblotting (Fig. 5g), suggesting a potential requirement of Zeb1 for phagosomal ROS production. Given that Nox2 activity is absolutely required for cross-presentation of particulate antigens26, 27, the reduction of Cybb and Ncf2 might contribute to the defective cross-presentation in Zeb1-deficient cDC1s. Although most of the genes in the lysosome pathway were down-regulated in Zeb1-deficient cDC1s, the genes (Ap1b1, Ap1m2 and Ap1s3) encoding subunits of the clathrin adaptor AP-1 that controls not only the distribution of apical proteins, but also the trafficking of endocytic vesicles44, were up-regulated in Zeb1-deficient cDC1s (Fig. 5f). Taken together, these results suggested that Zeb1 promoted cross-presentation probably by regulating the expression of a series of genes involved in the phagosome and lysosome pathways.
In most instances, Zeb1 directly represses the transcription of its target genes by recruiting co-repressor Ctbp. In other cases, Zeb1 also activates transcription by recruiting co-activators such as p300 30. To determine whether the DEGs identified above were directly regulated by Zeb1, we assessed genome-wide Zeb1 binding in WT cDC1s after stimulation with HKLM-OVA by Cleavage Under Targets and Tagmentation (CUT&Tag)45. We identified 23,896 consensus Zeb1 binding sites in cDC1s after stimulation. Zeb1 binding in the gene promoter regions and sequences surrounding a transcription starting site accounted for 46% of the regions bound by Zeb1 in cDC1s after stimulation (fig. S6, d-e). Intragenic binding represented 25%, whereas distal intergenic binding represented another 25% of the binding sites in cDC1s (fig. S6, d-e). As expected, known motif discovery analysis retrieved the canonical Zeb1-binding motif as one of the most enriched motifs in the Zeb1 binding peaks, indicating that the result of CUT&Tag was as reliable as that of Chromatin Immunoprecipitation with high-throughput sequencing (ChIP-seq) (fig. S6f)46, 47. Comparison of Zeb1-bound with Zeb1-regulated genes revealed that 1,888 genes were transcriptionally regulated by Zeb1 (fig. S6g). KEGG analysis of the DEGs in the intersection revealed that Zeb1 target genes were significantly enriched in the pathways of cell adhesion molecules and lysosome (Fig. 5h). This suggested that most of DEGs in these pathways were directly bound by Zeb1 and were its direct target genes. For instance, we detected strong Zeb1 binding peaks in the promoter regions of Ap1b1 and Ap1s3, as well as in the promoter region and intragenic region of Ppt1, which suppresses cross-presentation (Fig. 5i)48. In contrast, the significance of the association of Zeb1 target genes with the pathways of phagosome, antigen processing and presentation, was notably reduced, compared to the association of Zeb1-regulated genes (Fig. 5h and 5b). This meant that significant Zeb1 binding signal was not detected in some of DEGs in the two pathways, e.g. Cybb (Fig. 5i). Altogether, these results demonstrated that Zeb1 might directly regulate the transcription of most DEGs in the lysosome pathway, but indirectly regulate the transcription of some DEGs including Cybb in the pathways of phagosome, antigen processing and presentation.
It has been reported that Zeb1 links EMT-activation and stemness-maintenance by suppressing stemness-inhibiting microRNAs (miRNAs)49. Given that Cybb was not directly regulated by Zeb1, we next explored the possibility that Zeb1 elevated the expression of Cybb by repressing a miRNA that targeted Cybb mRNA. To this end, we performed microRNA deep sequencing (miRNA-seq) analysis of WT and Zeb1-deficient cDC1s after stimulation with HKLM-OVA. We identified 31 miRNA DEGs by comparing the miRNA transcriptomes of WT and Zeb1-deficient cDC1s (fig. S6h and Table S3). The intersection of Zeb1-bound miRNAs and Zeb1-regulated miRNAs revealed that 19 miRNAs were direct targets of Zeb1 (fig. S6h). Among these miRNAs, miR-96 was predicted to bind to the 3’ end of coding sequence (CDS) of Cybb mRNA50. Two Zeb1 binding peaks were observed near the promoter of miR-183-96-182 cluster where miR-96 is located, consistent with previous reports (Fig. 5i)49, 51. As expected, miR-96, miR-182 and miR-183 were all up-regulated in Zeb1-deficient cDC1s after stimulation (Fig. 5j and fig. S6i). Thus, our data supported the finding that Zeb1 directly represses the transcription of the miR-183-96-182 cluster49, 51. To examine the direct regulation of Cybb by miR-96, we performed a dual-luciferase reporter assay. The gene fragment containing miRNA-binding site of each target was cloned into the expression vector that encodes Renilla luciferase controlled by the cloned fragment. These constructs were transfected into HEK293 cells along with an expression vector encoding miRNA. The regulation of protein expression by miRNA was measured using luciferase activity. miR-96 significantly reduced protein expression under the control of gene fragment of the Cybb, while mutation of the miR-96 binding site abolished this reduction (Fig. 5k). By contrast, miR-467d was not able to reduce protein expression through either WT or mutated 3’ untranslated region (UTR) of Cybb, although miR-467d was predicted to weakly bind to 3’UTR of Cybb (Fig. 5k). Therefore, miR-96 was able to suppress the expression of Cybb through its binding site at the 3’end of CDS of the transcript. Collectively, our data demonstrated that Zeb1 activated the expression of Cybb by repressing the transcription of Cybb-targeting miR-96 and further indicated that Zeb1 might be required for phagosomal ROS production.
Phagosomal ROS production and membrane rupture require Zeb1
DCs undergo a program of maturation following recognition of microbes, which transiently enhances the capacity to phagocytose antigens and also increases the expression of costimulatory molecules and inflammatory cytokines as the second and third signal for effective T cell response52. First, the defective cross-presentation was not due to the maturation of Zeb1-deficient Flt3L-cDC1s, because they expressed normal levels of MHC classⅠmolecules and costimulatory molecules including CD80, CD86, CD40, as well as inflammatory cytokines after stimulation with HKLM-OVA or TLR ligands (fig. S7a and fig. S3). Second, Zeb1-deficient cDC1s had similar capability in uptake of soluble OVA, HKLM-OVA and OVA-loaded β2m−/− splenocytes (fig. S7, b-d), despite the defective cross-presentation, relative to WT cDC1s. Together, these data indicated that Zeb1 was dispensable for DC maturation and phagocytosis.
As we observed that Zeb1 deletion curtailed the expression of Cybb via upregulation of miR-96 in cDC1s, we first asked whether the defective cross-presentation in Zeb1-deficient cDC1s was caused by miR-96-mediated Cybb downregulation. To address this, we transduced BM Lin−c-Kithi stem cells from Zeb1-dcKO mice with lentiviral vector expressing Cybb during Flt3L-induced in vitro DC development. The infected Flt3L-cDC1s were sorted and assayed for cross-presentation. Strikingly, lentiviral expression of Cybb in Zeb1-deficient cDC1s completely restored the capability of cross-presentation of bacteria-associated antigens (Fig. 6a). Then we infected BM Lin−c-Kithi stem cells from B6 mice with control retrovirus or retrovirus encoding miR-96 during Flt3L-induced in vitro DC development. As described above, the infected Flt3L-cDC1s were sorted and tested. The cross-presentation of bacteria-associated antigens by infected Flt3L-cDC1s was partially and significantly inhibited by miR-96 overexpression (Fig. 6b), which to some extent mimicked the defective cross-presentation in Zeb1-deficient cDC1s. Collectively, these results demonstrated that the Zeb1-miR-96-Cybb pathway plays a critical role in controlling cross-presentation in cDC1s.
We next examined whether phagosomal ROS was affected by Zeb1-deficient Flt3L- cDC1s. We first challenged Flt3L-cDC1s with HKLM-OVA and measured ROS production in the cytoplasm and mitochondria after phagocytosis, using CellROX and MitoSOX respectively. We observed similar production of mitochondrial ROS and mild reduction of cytosolic ROS in Zeb1-deficient Flt3L-cDC1s compared to WT cDC1s (Fig. 6c). Then we fed Flt3L-cDC1s HKLM-OVA labelled with OxyBURST and Alexa Fluor 647 and measured the oxidative burst as indicator of phagosomal ROS production. The ratio of OxyBURST+ to Alexa 647+ cDC1 population increased progressively in WT cDC1s following HKLM-OVA ingestion. However, this ratio remained at a very low level in Zeb1-deficient cDC1s (Fig. 6d-e). More importantly, Zeb1-deficient cDC1s with phagocytosed HKLM-OVA generated much lower intensity of Oxidative Burst than WT counterparts did (Fig. 6e). Consistent with a role for phagosomal ROS in cross-presentation26, 27, inhibition of the NADPH oxidase by diphenyleneiodonium chloride (DPI) completely blocked cross-presentation of bacteria-associated antigens in both WT and Zeb1-deficient cDC1s (Fig. 6f). Altogether, these results demonstrated that Zeb1 was required for production of phagosomal ROS but not of mitochondrial ROS in cDC1s upon phagocytosis.
Phagosomal ROS has been shown to control phagosomal acidification, proteolysis and membrane rupture26, 27, 53. To determine whether the defective phagosomal ROS production affect phagosomal acidification, we monitored phagosomal pH by exposing Flt3L-cDC1s to FITC-coupled (pH-sensitive), Alexa Fluor 647-coupled (pH-insensitive) or pHrodo Red-coupled (pH indicator) HKLM-OVA. We did not detect any difference in all the three kinds of fluorescence in cDC1s of both genotypes at all tested time points after phagocytosis (fig. S7e), suggesting that phagosomal acidification was not affected by Zeb1 deletion in cDC1s, despite the defective phagosomal ROS production.
After internalization, antigens are processed and loaded onto MHC-Ⅰmolecules via the vacuolar pathway or the P2C pathway21, 22. The P2C pathway is thought to be most dominant mechanism in cross-presentation and immune surveillance. The antigen export into the cytosol is mediated by ERAD machinery or through phagosomal rupture21, 22, 54. A recent study has reported that phagosomal membrane damage induced by phagosomal ROS is necessary for P2C transfer of dead cell-associated antigens and thereby for cross-presentation26. We next assessed whether the defective phagosomal ROS production reduced phagosomal membrane damage in Zeb1-deficient cDC1s. To address this, we measured recruitment of cytosolic galectin-3, which binds to sugar moieties attached to membrane proteins on the luminal side of phagosomes, to the HKLM-OVA-containing phagosomes, as a marker for phagosomal membrane damage26, 55. By confocal microscopy, we observed more galectin-3 puncta in WT cDC1s than in Zeb1-deficient cDC1s following uptake of Alexa Fluor 647-labelled HKLM-OVA, and the galectin-3 puncta were highly colocalized with HKLM-OVA+ phagosomes in WT cDC1s but not in Zeb1-deficient cDC1s (Fig. 6g-h), indicating that phagosomal rupture was diminished in Zeb1-deficient cDC1s. Taken together, these results suggested that impaired phagosomal ROS production caused by Zeb1 ablation in DCs resulted in diminished phagosomal rupture, which probably impeded antigen export to the cytosol and impaired cross-presentation.
Ablation of Zeb1 prevents antigen export to the cytosol in cDC1s
A critical step in P2C pathway is the export of internalized antigens from phagosome to the cytosol54. As the phagosomal ROS production and phagosomal rupture were diminished in Zeb1-deficient cDC1s, we next examined whether antigen export to cytosol was affected by Zeb1 deficiency. To this end, we performed a cytofluorimetry assay based on the enzymatic activity of β-lactamase56. Purified Flt3L-cDC1s were loaded with CCF4, a cytosolic fluorescence resonance energy transfer (FRET) substrate of β-lactamase. Following uptake of soluble β-lactamase or β-lactamase-loaded β2m−/− splenocytes, the enzyme undergoes export to the cytosol and then cleaved the CCF4 dye, resulting in the switch from 525nm emission fluorescence to 450nm fluorescence. Under both experimental setting, cleavage of CCF4 was increased overtime in WT cDC1s, by contrast, CCF4 was cleaved much less in Zeb1-deficient cDC1s, reflecting reduction in translocation of both types of β-lactamase (Fig. 7a-d). These results demonstrated that Zeb1 was required for efficient export of both soluble and cell-associated antigens into the cytosol. Once exported into the cytosol, antigens are degraded into oligopeptides by proteasome. We incubated Flt3L-cDC1s with β2m−/− splenocytes loaded with DQ-OVA, a protease probe consisting of OVA proteins heavily labelled with self-quenched BODIPY dye, which becomes brightly fluorescent upon OVA hydrolysis. Consistently, Zeb1 ablation significantly decreased DQ-OVA fluorescence that was blocked by proteasome inhibitor MG132, suggesting that proteasome-mediated cytosolic antigen processing was crippled in Zeb1-deficient cDC1s (Fig. 7e).
It has been reported that delay of phago-lysosome fusion prevents excessive degradation of internalized antigens and enhances cross-presentation57. The expressions of AP-1 adaptor subunit encoding genes (Ap1b1, Ap1m2 and Ap1s3) were all increased in Zeb1-deficient cDC1s after stimulation (Fig. 5f), implying more trafficking to lysosome. Given the impaired export of internalized antigens into cytosol in Zeb1-deficient cDC1s, we next explored whether these antigens gained more opportunity to be transported from endosome/phagosome to lysosome. We first analyzed various cellular compartments of WT and Zeb1-deficient Flt3L-cDC1s at steady state by confocal microscopy and found no obvious difference in distribution of early endosomes (Rab5), late endosomes (Rab7), recycling endosomes (Rab11) and lysosomes (Lamp1) (fig. S7f). Furthermore, we observed low colocalization of HKLM-OVA and early endosomes/phagosomes, late endosomes/phagosomes, or recycling endosomes in both WT and Zeb1-deficient cDC1s at 4 hours after phagocytosis (Fig. 7f), but found much higher colocalization of HKLM-OVA and Lamp1+ lysosome in Zeb1-deficient cDC1s than in WT cDC1s (Fig. 7g). These data suggested that the fusion of phagosomes containing HKLM-OVA with lysosomes was dramatically increased in Zeb1-deficient cDC1s after phagocytosis. Additionally, we employed a FRET-based assay to measure the relative interaction between a donor Alexa Fluor 488 on the phagocytosed HKLM-OVA and an acceptor Alexa Fluor 594 that was endocytosed and chased into lysosomes. The FRET signal increased much faster early after phagocytosis in Zeb1-deficient cDC1s than in WT cDC1s, confirming that phago-lysosome fusion was enhanced in Zeb1-deficient cDC1s (Fig. 7h). Together these results suggested that internalized antigens in Zeb1-deficient cDC1s lost access to the cytosol, instead gained more access to lysosomes, which was associated with enhanced phago-lysosome fusion.