Inositol Requiring Enzyme 1 (IRE1) is an endonuclease embedded in the ER membrane that together with PKR-like endoplasmic reticulum kinase (PERK) and Activating Transcription Factor 6 (ATF6), coordinates the unfolded protein response (UPR)1–3. The UPR is an adaptive response that is typically activated upon accumulation of unfolded proteins in the ER. IRE1 represents the most conserved branch of the UPR and splices X-box binding protein 1 (Xbp1) mRNA to generate the transcription factor XBP1s. XBP1s in turn helps to restore protein homeostasis by inducing the expression of chaperones, redox enzymes and ER-associated degradation (ERAD) components. In addition, IRE1 endonuclease activity targets mRNAs and miRNAs for degradation through a poorly understood process termed regulated IRE1 dependent decay (RIDD)4–7. In vivo, IRE1 is typically activated in secretory cells such as goblet cells8 or plasma cells9, or in immune cells that are actively proliferating during a viral infection10, underscoring IRE1’s canonical role in protein folding. On the contrary, in type 1 conventional dendritic cells (cDC1s), IRE1 is active in absence of prototypical ER stress and without the induction of typical XBP1s target genes11–15. Why and how IRE1 is specifically activated in cDC1s but not in the highly related cDC2 subset remains enigmatic.
Loss of XBP1/IRE1 affects gene expression in cDC1s, not cDC2s
To gain insights into the function of IRE1 in cDC1s, we performed bulk RNA sequencing (RNA-seq) in different genotypes of the XBP1/IRE1 axis. Loss of XBP1 induces IRE1 hyperactivation and strong activation of RIDD-associated mRNA degradation in cDC1s, not in the closely related cDC2 subset12,13. This complicates the interpretation of RNA-seq data, since any gene downregulated in Xbp1fl/fl Itgax-cre (called XBP1∆DC) cDC1s could either be transcriptionally regulated by XBP1s or targeted for degradation through RIDD12. To dissect the functions of XBP1 and IRE1 separately, we compared cDC1s sorted from the spleens of XBP1(/IRE1)fl/fl (WT), XBP1∆DC and a XBP1/IRE1∆DC mouse strains (Fig. 1a). IRE1 can be found as two different paralogues, IRE1a and IRE1b. IRE1a is ubiquitously expressed, while IRE1b expression is limited to epithelial cells lining the mucosal tracts16. When referring to IRE1 in this study, we refer to IRE1a, encoded by the gene Ern1.XBP1∆DC cDC1s have lost Xbp1 mRNA expression and show high RIDD activity, as reflected by downregulation of the prototypic RIDD target gene Bloc1s17(Fig. 1b). XBP1/IRE1∆DC cDC1s do not express Xbp1 nor Ern1, hence lost RIDD activity, as confirmed by restoration of Bloc1s1 expression (Fig. 1b). The two major splenic conventional DC subsets, cDC1s and cDC2s, were sorted (gating strategy Extended Data Fig. 1) and processed for bulk RNA-seq analysis. In cDC2s, very few differentially expressed (DE) genes could be identified, revealing that in steady state the XBP1/IRE1 pathway does not play a major role in cDC2s (Supplementary Table 1). On the contrary, in cDC1s many DE genes were uncovered in all 3 pairwise comparisons (Fig. 1c and Supplementary Table 2), which was validated by RT-qPCR (Extended Data Fig. 2). Overall, three major directionalities of DE genes could be distinguished, with representative genes indicated on Fig. 1c and validated by RT-qPCR in Fig. 1b. A first group of genes (indicated in yellow on the rose plot, Fig. 1c) represents genes that were specifically downregulated in XBP1∆DC cDC1s and comprises classical RIDD targets such as Bloc1s1, Tapbp or Stim1 in addition to established XBP1 target genes (Fig. 1d-e, Extended Data Fig. 2, Extended Data Fig. 3a-b). Loss of XBP1 in cDC1s does not lead to a major loss in XBP1 target gene expression as only a few XBP1 target genes were retrieved as differentially expressed genes (DEGs), all with low fold inductions (e.g. Sec61a1, Txndc11, Sec24d). Of note, several genes previously annotated as XBP1 transcriptional target genes18 appeared regulated by RIDD rather than by XBP1 based on their restored expression in XBP1/IRE1∆DC cDC1s (Fig. 1d, Extended Data Fig. 3b). A second group of genes (indicated in green and blue on the rose plot, Fig. 1c) is highly upregulated in XBP1∆cDC1s and XBP1/IRE1∆cDC1s and comprises genes that belong to the integrated stress response (ISR) (Fig. 1d-e, Extended Data Fig. 2, 3d, 4), confirming earlier data13. Finally,the pink group (Fig. 1c) represents genes that were downregulated in absence of XBP1/IRE1. Based on their directionality, they appeared more affected by the loss of IRE1 than by the loss of XBP1s transcriptional activity (Fig. 1b, c). To probe the function of the DE genes in this group, we assessed whether particular gene ontology (GO) terms were enriched in this direction. This revealed GO immunological terms such as “negative regulation of inflammation” and “tolerance induction” associated with the pink group of genes (angle 5-6 Extended Data Fig. 5a, Supplementary Table 3), comprising well-established DC maturation genes such as Tmem176a, Fsnc1 or Ccr719,20 (Extended Data Fig. 3e). None of these genes were known as IRE1 or XBP1s target genes before. In addition, ingenuity pathway analysis (IPA) highlighted ER associated (GO) terms such as “unfolded protein response”, “superpathway of cholesterol biosynthesis” or “tRNA charging” in the comparison XBP1∆DC vs WT while the comparisons XBP1/IRE1∆DC vs WT, and XBP1∆ vs XBP1/IRE1∆DC additionally retrieved DC specific categories such as “Th1 and Th2 activation pathway”, “dendritic cell maturation”or “graft-versus-host-disease signaling” (Extended Data Fig. 5b).
Overall, the bulk RNA-seq analysis revealed that the absence of the IRE1/XBP1 signaling branch does not affect gene expression in cDC2s, in contrast to cDC1s. In line with our earlier observations12,13, the high basal activity of IRE1 in cDC1s was not reflected by strong expression of canonical XBP1 target genes. On the contrary, a large group of DC specific genes appeared downregulated in absence of IRE1 rather than XBP1 and were associated with DC maturation. While we could not rule out the possibility that these genes might be indirectly regulated rather than being direct targets of IRE1 endonuclease activity, it suggested that in cDC1s, IRE1 might hold functions beyond its traditional role in protein folding.
IRE1 is essential for homeostatic cDC1 maturation
DEG analysis revealed that genes belonging to the homeostatic and common DC maturation program19 showed decreased expression levels in XBP1/IRE1 deficient cDC1s (Fig. 2a, Extended Data Fig. 3e, 4), which was validated by RT-qPCR (Extended Data Fig. 2).
To assess whether the differences in expression of DC maturation genes were reflected by differences in DC maturation, we immunophenotyped the DC compartment in the different genotypes. In line with our previous findings12,13, the total number of splenic cDC1s and cDC2s was not altered in XBP1∆DC or XBP1/IRE1∆DC mice (Fig. 2b). However, the percentage of mature CCR7+ cDC1s was strongly affected, particularly in XBP1/IRE1∆DC mice (Fig. 2b), while CCR7+ cDC2s were not affected. To exclude the possibility that the difference in mature cDC1s was due to downregulation of the marker gene Ccr7, we assessed the percentage of CD86+ cDC1s, as Cd86 was not affected in the RNA-seq analysis (Fig. 2a) which confirmed the specific decrease in the mature cDC1 population (Fig. 2c). Furthermore, we noticed that the surface expression of CCR7 appeared unaffected in XBP1∆DC and XBP1/IRE1∆DC cDC1s, even though Ccr7 came out as a DE gene (Extended Data Fig. 6a). This indicated that Ccr7 might not be directly regulated by IRE1, but that the loss of IRE1 in cDC1s is associated with a block in maturation, which is reflected by a general decrease in maturation genes in a bulk RNA-seq dataset. Recently, our lab extensively mapped homeostatic DC maturation pathways in the spleen through CITE-Seq analysis and lineage tracing experiments20. This led to the identification of markers that could be used by flow cytometry to distinguish different maturation stages of cDC1s20 (Extended Data Fig. 6b). We used this gating strategy to assess at which point loss of the XBP1/IRE1 signaling branch affected the DC maturation program and found significant loss of the mature CCR7+ cDC1s and the late immature subset (identified as CD62L-CD103+ESAM+CCR7-). On the other hand, the early immature cDC1 subset (CD62L+CD103-ESAM-CCR7-)was increased, suggesting that loss of IRE1 led to a reduction of mature cDC1s and an accumulation of the early immature cells (Fig. 2d).
Signals driving homeostatic DC maturation in steady state conditions have long remained enigmatic21. Recently, several labs uncovered an essential role for apoptotic cell (AC) engulfment and cholesterol metabolic pathways at the heart of cDC1 maturation20,22,23. We previously noted that injection of exogenous apoptotic thymocytes boosted engulfment in splenic cDC1s and triggered their homeostatic maturation with a peak observed at 12h post- intravenous injection20 (Fig. 2e). In line with the observed defects in the mature cDC1 compartment, injection of ACs in XBP1/IRE1∆DC mice did not lead to a similar increase in cDC1 maturation (Fig. 2e). Also, injection of empty non-adjuvanted lipid nanoparticles (eLNPs)20, comprising 40% cholesterol (Fig. 2f left panel), only slightly increased cDC1 maturation in absence of XBP1/IRE1. On the contrary, injection with LNPs coupled to poly(I:C), a potent TLR3 ligand, did induce full cDC1 maturation in XBP1/IRE1∆DC mice, showing that the immunogenic maturation program did not depend on the presence of the IRE1 signaling branch (Fig. 2f right panel).
In summary, these data highlight a function for the ER stress sensor IRE1 in the homeostatic maturation process of cDC1s, not cDC2s. Kinetics experiments further revealed that the process of cDC1 maturation induced by injection of ACs or cholesterol rich eLNPs was strongly impaired in absence of IRE1, while TLR ligand-induced cDC1 maturation appeared largely unaffected.
IRE1 is triggered by apoptotic cell engulfment
Several labs meanwhile noted that IRE1 shows a higher basal activity in cDC1s compared to cDC2s11–15, still the mechanism explaining this subset specific activation of IRE1 has remained enigmatic. A recent study proposed that antigen-derived peptides can engage IRE1 in a TAP1-dependent manner by binding the IRE1 lumenal domain upon their import in the ER14. We tested this hypothesis in vivo by monitoring the IRE1 activity in TAP1-deficient cDC1s, by crossing the well-established IRE1 reporter line ERAI12,24 on TAP1-deficient mice. Unexpectedly, absence of TAP1 did not result in any difference in ERAI activity, suggesting that in vivo TAP1-dependent import of peptides into the ER does not contribute to the cDC1 specific steady state activity of IRE1 (Extended Data Fig. 6c).
We previously noted that AC engulfment in cDC1s is reflected by changes in endogenous cholesterol levels: cholesterol levels first rise due to the uptake of apoptotic cargo followed by a steep decline once cDC1s start to mature and migrate to the white pulp of the spleen20. In steady state conditions, cDC2s do not engulf ACs20,25–34 and therefore do not show these drastic changes in cholesterol levels20. Intrigued by recent data showing IRE1 activation by accumulation of aberrant lipids such as cholesterol at the ER membrane35–37, we tested the premise that the selective activation of IRE1 in cDC1s could be explained by their unique capacity to engulf ACs in vivo. A strict correlation could be observed between IRE1 activity, as monitored by ERAI reporter activity12,13,24, and intracellular cholesterol content as measured by BODIPY 493/503, which stains all neutral lipids including cholesterol esters (Extended Data Fig. 6d). IRE1 activity increases from the early immature to the late immature state and then declines as soon as cells gain CCR7 expression (Extended Data Fig. 6d). Injection of exogenous CTV-labeled ACs led to an increase in IRE1 activity specifically in CTV+ cDC1s 2h post injection (p.i.), but not in cDC2s (Fig. 3a). More detailed examination of the different maturation stages revealed an increase in ERAI signal in all CTV+ subsets, which was least prominent in the late immature cDC1s, that already have a high basal IRE1 activity due to engulfment of endogenous (CTV-) ACs20 (Fig. 3a). We recently generated a Rac1∆DC Rac2-/- mouse line in which cDC1s are deficient in the engulfment of ACs and therefore have lower levels of neutral lipids and cholesterol esters20. We crossed the ERAI reporter to this line and observed that blocking engulfment causes a strong decrease in IRE1 activity in cDC1s, whereas IRE1 activity in cDC2s remains low (Fig. 3b). To assess whether the increase in IRE1 activity was due to the engulfment process itself or due to the influx of lipids, we injected inert beads and eLNPs, respectively. Engulfment of inert beads did not induce IRE1 activity while eLNPs did, following a similar pattern as observed upon AC engulfment (Fig. 3c-d). Of note, the difference in IRE1 activity between CTV+ or LNP+ and CTV- or LNP- cDC1s was most prominent in the CCR7+ stage (Fig. 3a, d). This can be explained by the fact that CTV+ CCR7+ cDC1s are still in the early mature state, as reflected by their intermediate expression level of CCR7 (blue dots in Extended Data Fig. 6e), compared to CTV- CCR7+ cDC1s which consist mainly of late mature cDC1s. From early to late mature cDC1s, IRE1 activity progressively declines (Extended Data Fig. 6c), to reach basal levels 12h post-injection of ACs (Fig. 3e). Altogether, these data establish that uptake of ACs and more specifically influx of lipids triggers IRE1 endonuclease activity in cDC1s.
Furthermore, the engulfment capacity of XBP1/IRE1 deficient cDC1s appeared reduced, especially at 12 hours post-injection (Fig. 3f). Since the engulfment defect appeared less pronounced at early timepoints, the reduction in the CTV+ cDC1 population at later timepoints could also be explained by a specific survival deficit of CTV+ mature cDC1s over time in XBP1/IRE1∆DC mice (Fig. 3f). In contrast, engulfment of inert latex beads in cDC1s was not hampered by deletion of XBP1/IRE1 neither at the 2 or 12 hours time point (Fig. 3g).
In summary, these data reveal that lipids derived from ACs are the major trigger for IRE1 activity in steady state cDC1s rather than the import of peptides into the ER. This explains the selective activation of IRE1 in cDC1s11–13 and is in line with earlier studies showing IRE1 activation by aberrant lipids such as accumulation of cholesterol at the ER membrane35–41. Deficiency of XBP1/IRE1 in cDC1s leads to a specific loss of cDC1s that engulfed ACs, potentially explaining why homeostatically matured cDC1s are reduced in XBP1/IRE1∆DC mice.
IRE1 controls cholesterol efflux in homeostatic mature cDC1s
Earlier studies showed a role for IRE1-dependent RIDD in regulating lipid metabolism42–44. Our bulk RNA-seq data in cDC1s confirmed this and showed a specific deficit in cholesterol biosynthesis genes like the transcription factor Srebf2, or key enzymes in the mevalonate pathway, such as Sqle, Cyp51 and Hsd17b17 upon XBP1-deficiency (Fig. 4a, Extended Data Fig. 2, Extended Data Fig. 3c). In IRE1/XBP1 deficient cDC1s, the cholesterol efflux gene Abcg1 and the apolipoproteins Apol7c and Apol10b, genes that we previously noted to be induced specifically in homeostatic mature cDC1s20, were downregulated. On the other hand, genes related to cholesterol esterification like Soat2 or genes associated with lipotoxicity such as Chop and Trib3 were upregulated in IRE1/XBP1 deficient cDC1s (Fig. 4a, Extended Data Fig. 2, Extended Data Fig. 3c). Since the loss of IRE1 and XBP1 leads to a strong reduction in the number of homeostatic mature cDC1s, we decided to verify whether these genes were affected at single cell level in absence of IRE1 and performed cellular indexing of transcriptomes and epitopes by sequencing (CITE-seq). CITE-Seq was performed on sorted CD64- CD11c+ MHCII+ XCR1+ CD172a- cDC1s (65% of total), CD64- CD11c+ MHCII+ XCR1- CD172a+ cDC2s (10% of total), CD64- CD11c+ MHCII-/Lo CD135+ CD172adim pre-DCs (15% of total) and live cells (10% of total) from the spleen of WT (XBP1/IRE1fl/fl) and XBP1/IRE1∆DC mice. Unsupervised clustering and UMAP dimensionality reduction yielded several clusters for cDC1s that were annotated based on expression of data-driven marker genes and led to the identification of pre-cDC1s, proliferating cDC1s, early immature cDC1s, late immature cDC1s, early mature cDC1s, Cxcl9+ cDC1s and late mature cDC1s (Extended Data Fig. 7a). We recently identified these clusters as consecutive steps in cDC1 maturation, except for the Cxcl9+ cluster which was previously included in the early mature cDC1 cluster20. Comparing the relative abundances of each subcluster of cDC1s confirmed the reduction of mature cDC1s in XBP1/IRE1∆DC mice (Extended Data Fig. 7b). DE genes (both up and down) could be identified in each cDC1 subcluster, with only a few in the pre-cDC1s and most in the late immature and mature subsets (Supplementary Table 4). Previously identified DE genes with a high logFC (>2) linked to DC maturation (like Ccr7, Fscn1, Tmem176a, Tmem176b, Slco5a1, Nudt17) were not differentially expressed anymore in the late mature cDC1s in the CITE-seq analysis, indicating that they were differentially expressed due to a loss of the CCR7+ cDC1 subset in the bulk RNA-seq rather than being directly regulated by IRE1. However, we could still identify DE genes related to cholesterol metabolism and apolipoproteins (Apol7c, Apoe, Apol10b and Abcg1) (Supplementary Table 4, Extended Data Fig. 7c), a few remaining DC maturation genes (Il4i1, Mical3, H2-M2, etc.) and Dnase1l3, which is related to the degradation of self-DNA. The upregulated genes mainly contained genes related to the ISR pathway (Atf4, Ddit3, Cars, Yars, …) confirming the findings of the bulk RNA-seq (Supplementary Table 4). One of the top upregulated genes was Cox6a2, a subunit of the mitochondrial complex IV of the oxidative phophorylation pathway, which was also one of the few upregulated genes in XBP1/IRE1 deficient cDC2s (Supplementary Table 1). Together, the CITEseq data confirmed the specific loss of the CCR7+ mature cDC1 subset and revealed genes important for cholesterol efflux, such as Apoe and Abcg1 and apolipoproteins Apol10b and Apol7c as potential IRE1 target genes.
We previously showed that during cDC1 homeostatic maturation, cholesterol efflux is activated to restore cholesterol levels after AC engulfment in an LXR-dependent manner20. Based on the gene expression signature obtained in the XBP1/IRE1 CITE-seq data and the activation of IRE1 by ACs and LNPs, we decided to assess the role of the IRE1 branch in regulating cholesterol efflux during cDC1 maturation. RT-qPCR analysis confirmed the decrease in expression of Abcg1, Apoe, Apol10b and Apol7c most prominently in XBP1/IRE1∆DC cDC1s (Fig. 4b). This was reflected by a small increase in BODIPY 493/503 levels (Fig. 4c), which can be used to monitor the amount of neutral lipids amongst others cholesterol esters. To assess specifically the cholesterol content in the cell, we used an enzymatic assay that detects both free and esterified cholesterol and confirmed the increase in cholesterol levels in mature XBP1/IRE1∆DC cDC1s, while we did not detect any difference in immature cDC1s or in cDC2s (Fig. 4d). We previously noted that late during homeostatic cDC1 maturation, likely as a compensatory response to ABCG1-mediated shuttling of cholesterol from the ER45, SREBP2-dependent gene transcription becomes reactivated20. We observed that SREBP2-target genes Ldlr and Sqle were less induced in XBP1/IRE1∆DC cDC1s, potentially due to the loss of ABCG1 expression and therefore a potential defect in cholesterol export from the ER (Fig. 4b).
Abcg1 and Apoe are known target genes of the LXR signaling pathway and we previously established a role for LXRß in mediating cholesterol efflux upon AC engulfment in cDC1s20. Furthermore, LXRs have been shown to mitigate ER stress by upregulating Lpcat3, an enzyme that favors the incorporation of polyunsaturated fatty acids into phospholipid46. Hence, we were keen to understand if the IRE1 and LXR signaling pathways operated independently from each other or impacted each other (Extended Data Fig. 8a). We reasoned that if activation of the LXR signaling pathway was upstream of IRE1 (scenario 1), this would be reflected by a loss of IRE1 reporter activity in LXRa/LXRß∆DC mice, hence we crossed ERAI mice onto LXRa/LXRß∆DC mice to assess this. Neither in total cDC1s or cDC2s, neither in any of the previously identified cDC1s states, IRE1 activity was affected by loss of LXRa/LXRß, indicating that activation of IRE1 occurs independently from LXR signaling (Extended Data Fig. 8b). Due to the lack of specific LXRß antibodies or reporter assays in vivo, we addressed scenario 2, whether IRE1 would be needed for triggering LXRa/LXRß activation, in an indirect way. We measured the expression of interferon stimulated genes (ISGs) in XBP1/IRE1∆DC cDC1s as we had previously shown that they were induced in LXRa/LXRß deficient cDC1s20. Loss of XBP1/IRE1 in cDC1s did not result in increased expression of ISGs, suggesting that the LXRa/LXRß pathway was still active in IRE1 deficient cDC1s (Extended data Fig. 8c).
In conclusion, in the absence of IRE1 and XBP1, the efflux of cholesterol levels after AC engulfment in cDC1s is not properly controlled, leading to enhanced cholesterol levels particularly at the mature state. This indicates that IRE1 plays a crucial role in regulating cholesterol metabolism in maturing cDC1s, likely in an LXR-independent manner (Extended Data Fig. 8a scenario 3).
IRE1 controls cholesterol efflux genes in a RIDD/miRNA-dependent manner
Cholesterol efflux genes were downregulated, not stabilized in absence of IRE1 endonuclease activity. This suggested that they were not directly targeted by IRE1-dependent RIDD activity, but that they might be targets of miRNAs that would be regulated by IRE1 (Fig. 5a, WT). IRE1 dependent RIDD activity has been previously linked to the degradation of miRNAs, especially in conditions of prolonged ER stress44,47–49. To test the hypothesis that IRE1-dependent regulation of cholesterol efflux genes, and hence DC maturation, was controlled by the degradation of miRNAs, we generated IRE1/Dicer deficient conditional knockout mice (IRE1/Dicer∆DC). We postulated that in absence of IRE1, miRNA cleavage is prevented, leading to degradation of cholesterol efflux mRNAs (Fig. 5a, IRE1∆DC). In DCs deficient for both IRE1 and Dicer, pre-miRNAs might no longer mature, leading to a restoration of downstream cholesterol efflux genes and cDC1 maturation (Fig. 5a, IRE1/Dicer∆DC). In line with this hypothesis, the absence of Dicer on top of IRE1 led to a restoration of homeostatic mature cDC1s (Fig. 5b). Similarly, we noted a complete restoration of the expression levels of Abcg1, and a partial restoration of Apoe and Apol10b, while the levels of Apol7c remained unaffected in the absence of Dicer expression (Fig. 5c). To address whether IRE1 regulates the expression of miRNAs in steady state conditions in cDC1s, we performed small RNA-seq on sorted splenic cDC1s derived from WT, XBP1∆DC and XBP1/IRE1∆DC mice and highlighted in red all significantly DE miRNAs between XBP1/IRE1∆DC and WT cDC1s on a Triwise plot (Fig. 5d, Supplementary Table 5). Four DE miRNAs were identified that were specifically upregulated in absence of IRE1, hence in absence of RIDD activity (Fig. 5d), which were all validated by RT-qPCR (Fig. 5e). Interestingly, miR-92a has been previously linked to regulating cholesterol efflux in macrophages by targeting Abca150. Inspection of the sequence of miR-92a revealed a potential IRE1 cleavage site UUGCAC that was present in the stem, rather than the stemloop (Extended Fig. 9). In none of the other DE miRNAs a potential IRE1 cleavage site could be found (data not shown). To address whether IRE1 could cleave miR-92a, an in vitro cleavage assay was set up with human recombinant IRE1 cytosolic endonuclease domain incubated with RNA oligonucleotides encoding Xbp1 stemloop, miR-92-a-1 or miR-30b, the latter being one of the most abundantly expressed miRNAs in cDC1s (data not shown). Both the Xbp1 stemloop and miR-92a-1 were cleaved by IRE1, while miR-30b remained unaffected (Fig. 5f).
Altogether, these data indicate that in homeostatic conditions IRE1 can cleave miR-92a-1 through a RIDD-dependent mechanism in cDC1s. Loss of miRNAs by removal of Dicer on top of IRE1 deficiency leads to restoration of Abcg1 expression levels and restores the mature homeostatic cDC1 population.
Homeostatic mature cDC1s can be restored by enforcing cholesterol efflux
When analyzing the cholesterol metabolic gene signatures in absence of XBP1/IRE1 signaling, we noted that the upregulation of Soat2, encoding the cholesterol esterifying enzyme acetyl-CoA acetyltransferase51 and the induction of Trib3 and Ddit3, potentially reflected signs of lipotoxicity (Fig. 4a, Extended Data Fig. 2c). To test the hypothesis that the loss of homeostatic mature cDC1s in absence of IRE1 was caused by lipotoxicity, we aimed to restore cholesterol levels by injecting reconstituted high-density lipoprotein (rHDL) into IRE1/XBP1 deficient mice. However, it did not manage to reach the limited amounts of cDC1s in the spleen. Therefore, we established an in vitro system for cDC1s by cultivating bone marrow (BM) cells in presence of FLT3L and on a feeder layer of OP9 fibroblasts expressing the Notch ligand DLL1 (Flt3 Notch cDC1s)52,53. This protocol yielded cDC1s with IRE1 activity levels that were comparable to what we noticed in the spleen (Fig. 6a). Similar to what we had observed in vivo, loss of XBP1 leads to a reduction in CD11c expression, due to activation of RIDD, which is restored by concomitant loss of IRE113 (Fig. 6b). When analyzing the cDC1s by microscopy, we observed the same aberrant ER structures, as we described before in sorted XBP1∆DC and XBP1/IRE1∆DC cDC1s (Fig. 6c)12. Also in vitro, the absence of XBP1/IRE1 led to a defect in cholesterol efflux gene expression (Fig. 6d), which was reflected by an increase in total cholesterol levels (Fig. 6e). The effect of XBP1/IRE1 deficiency on cDC1 survival in Flt3 Notch DC cultures becomes prominent from day11 of seeding the BM cells on the OP9 feeder layer onwards, while at day9 the survival of WT versus XBP1/IRE1∆DC cDC1s is still similar (Fig. 6f). Treatment of XBP1/IRE1∆DC Flt3 Notch cDC1s on day9 for 2 days with rHDL to enforce cholesterol efflux led to a complete restoration of their survival at day11 (Fig. 6f), supporting the hypothesis that homeostatic mature IRE1-deficient cDC1s die due to accumulation of cholesterol in absence of IRE1.
Loss of IRE1 in cDC1s leads to hampered cross-priming
We speculated that the loss of homeostatic mature cDC1s in XBP1/IRE1∆DC mice will impact the cross-priming of dead-cell derived antigens. Previous studies from our lab showed that loss of XBP1, but not loss of IRE1 in DCs, leads to decreased cross-presentation in an ex vivo co-culture assay12. At the time, we assumed that this was due to activation of canonical RIDD in XBP1-deficient cDC1s which led to degradation of Tapbp and Ergic3. In XBP1/IRE1∆cDC1s Tapbp and Ergic3 mRNA expression was restored, leading to the restoration in cross-presentation12. However, in this type of ex vivo co-culture assays, DCs and T cells are brought in close proximity bypassing any potential deficits in processes like cDC1 migration to the T cell area in the dLN. To take this into account, we assessed the presentation of dead-cell derived antigens by XBP1/IRE1∆DC cDC1s in vivo. We adoptively transferred CTV-labeled CD45.1.2 OT-I cells in CD45.2 acceptor WT and XBP1/IRE1∆DC mice. One day later, we injected apoptotic thymocytes from Act-mOVA mice or wild-type mice as a control. Three days later, we sacrificed the mice and analyzed the proliferation of OT-I cells in the spleen (Extended Data Fig. 10a). As expected, injection of mice with WT apoptotic thymocytes, not containing any OVA, did not result in proliferation of OT-I cells, and the percentage and number of OT-I cells remained low. Injection of OVA-containing apoptotic thymocytes led to increased percentage, number and CTV-dilution of OT-I cells. XBP1/IRE1∆DC mice showed a reduced percentage and number of OT-I cells after injection with OVA-ACs compared to WT littermates (Extended Data Fig. 10b). Furthermore, the proliferation index, a measure for CTV-dilution, was lower in the XBP1/IRE1∆DC mice compared to WT littermates (Extended Data Fig. 10c). These data therefore indicate that in vivo the loss of mature cDC1s in XBP1/IRE1∆DC does affect their capability to prime naïve antigen-specific T cells.
In summary, our data establish IRE1 as a sensor of cholesterol influx in cDC1s during AC engulfment. IRE1 regulates cholesterol efflux from cDC1s by regulating the stability of the cholesterol efflux transporter Abcg1 through RIDD-mediated miRNA-92a degradation, while LXRß drives Abcg1 expression. In absence of IRE1, cholesterol might accumulate at the ER, potentially explaining the aberrant ER aggregates. This causes lipotoxicity, leading to a loss of cDC1s that have recently engulfed and matured in XBP1/IRE1∆DC mice and results in defective cross-presentation of dead-cell derived antigens (Extended Fig. 10d). On the contrary, neither pIC-triggered immunogenic cDC1 maturation nor homeostatic cDC2 maturation depend on IRE1 signaling. These data therefore confirm the previously observed link between AC engulfment, cholesterol metabolism and homeostatic cDC1 maturation and establish a central role for IRE1 in this process.