IRE1 endoribonuclease signaling promotes myeloid cell infiltration in glioblastoma

Abstract

Perturbation of endoplasmic reticulum (ER) protein homeostasis is 1 of the hallmarks of highly proliferative and/or secretory cells. Moreover, many intrinsic and environmental conditions, such as low oxygen levels, acidification, or nutrient shortage also increase the risk of misfolded proteins accumulation in the ER lumen, leading to ER stress. To cope with the latter, cells use the unfolded protein response (UPR) which is transduced by the ER transmembrane proteins PERK, ATF6, and the most conserved UPR sensor IRE1 alpha (referred to as IRE1 hereafter). The UPR restores ER homeostasis or induces apoptosis when stress cannot be resolved.¹ Once activated, IRE1, which harbors cytosolic kinase and ribonuclease (RNase) activities, dimerizes/oligomerizes, thus inducing the activation of JNK1, XBP1 mRNA splicing (XBP1s) and regulated IRE1-dependent decay (RIDD) of RNA.² XBP1s activates the transcription of genes involved in protein glycosylation, ER-associated degradation, protein folding, and lipid synthesis. RIDD also controls cell fate under ER stress.³ The UPR has emerged as an adaptive mechanism supporting tumor progression and resistance to treatment by impacting almost all cancer hallmarks.⁴ Mounting evidence also suggests that the UPR shapes the tumor microenvironment by regulating angiogenesis, inflammation, and immune response.^5,6

The consequences of UPR signaling have been studied in various cancers such as breast, liver, lung, prostate, pancreas cancers, and in GB, the most lethal and aggressive primary brain tumors. IRE1 contributes to GB development by regulating tumor growth, invasion, and vascularization.^7,8 Loss of IRE1 signaling also results in decreased expression of proangiogenic and proinflammatory VEGFA, IL1β, IL6, and IL8 in GB cells.⁸ Importantly, we have shown a pivotal role of IRE1 in the immune remodeling of GB stroma, which mostly depends on XBP1s.^9,10 However, the precise IRE1-dependent mechanisms by which pro-tumoral inflammation/immune responses are regulated in GB remain elusive. IRE1 signaling induces the expression of proinflammatory chemokines in various models through XBP1s and JNK-dependent pathways or through GSK3β induction. IRE1 also interacts with TRAF2, recruiting IκB kinases (IKK), to promote the degradation of IκB thereby enabling NFκB nuclear translocation.¹¹ Since brain tumors are infiltrated by a large number of immune cells, which modulate GB aggressiveness and response to treatment, we aimed here to investigate the molecular mechanisms by which IRE1 controls GB tumor infiltration by immune/inflammatory cells. We showed that IRE1 governs myeloid cell recruitment to GB. This involves a novel IRE1/UBE2D3/NFκB signaling axis, thereby leading to proinflammatory chemokine production and the subsequent recruitment of immune/inflammatory cells to the tumor site.

Materials and Methods

Antibodies and Other Reagents

Primary antibodies are listed in Supplementary Table S1. Secondary antibodies used were horseradish peroxidase-conjugated polyclonal goat anti-rabbit IgG, goat anti-mouse IgG and rabbit anti-goat IgG (all from Dako). Unless specified, all other reagents were from Sigma Aldrich.

Cell Culture and Treatments

The U87, U87 IRE1.NCK DN cells (referred to as U87 DN), U251, and GL261 lines (all from ATCC) were grown in DMEM medium (Invitrogen) supplemented with 10% fetal bovine serum (FBS) (Lonza) in a 5% CO₂ humidified atmosphere at 37°C. Primary GB lines were previously described.^10,12

Generation of Stable Cell Lines

For stable overexpression of UBE2D3, RADH87, and GL261 cells were transfected with 3 µg of pCMV3-UBE2D3-Flag plasmid using lipofectamine LTX and 2000, respectively (ThermoFisher), following the manufacturer’s protocol. For stable ube2d3-downexpression, GL261 cells were transfected with 1 µg of pRS shRN-ube2d3. Transfections of GB lines with IRE1 WT and Q780* constructs¹⁰ were performed using Lipofectamine LTX (ThermoFisher), according to the manufacturer’s instructions. IRE1 WT and Q780* overexpressing cells were selected using puromycin (ThermoFisher).

Quantitative Real-Time PCR

Total RNA was extracted using the TRIzol reagent (Invitrogen). RNAs were reverse-transcribed with Maxima Reverse Transcriptase (ThermoFisher), according to the manufacturer’s protocol. qPCR was performed with a QuantStudio™ 5 Real-Time PCR System and the PowerUp™ SYBR Green Master Mix (ThermoFisher). Experiments were performed at least in triplicate for each data point. Each sample was normalized based on the expression of the GAPDH or ACTB gene using the 2^ΔΔCT-method. Primer pairs are listed in Supplementary Table 2.

Gel Shift Assay

Gel shift assay was performed using the LightShift Chemiluminescent EMSA Kit (ThermoFisher) according to the manufacturer’s instructions. Biotin end-labeled probes were prepared by PCR using 5ʹ biotin-labeled primers (Supplementary Table 3).

Flow Cytometry Analyses

Cells were washed in PBS 2% FBS and incubated with saturating concentrations of human immunoglobulins and fluorescent-labeled primary antibodies as indicated for 30 min at 4° C. Cells were then washed with PBS 2% FBS and analyzed using a FACSCanto and Novocyte flow cytometers (BD Biosciences and Acea Biosciences).

Myeloid Chemoattraction Assay

Myeloid cells were washed in DMEM and placed in 3 μm Boyden chambers (Merck Millipore), at 0.5 × 10⁵ cells/chamber in 200 µl DMEM. Chambers were simultaneously placed in 500 µl of either control DMEM or tumor cell-conditioned media, and incubated for 2 h (for PMN) or 16 h (for Mo) at 37°C. Migrated myeloid cells (under the Boyden chambers) were analyzed by flow cytometry.

Detailed methods are in the Supplemental Materials and Methods. This study does not include any data deposited in external repositories.

Results

Myeloid Cells are Recruited to GB in an IRE1-Dependent Manner

We previously demonstrated that IRE1 activity in GB cells controls the recruitment of inflammatory cells such as macrophages and microglial cells.¹⁰ Herein, we tested whether this also applied to other immune cells from the myeloid or lymphoid lineages. To this end, we first investigated whether immune gene signatures were associated with IRE1 activity¹⁰ using the TCGA GB datasets. These immune gene signatures characterize microglia/macrophages (MM) (ie, blood monocytes-derived macrophages (MDM), microglia cells (MG)^13–15); polymorphonuclear neutrophils (PMN) and T cells (T). Myeloid MM (indifferently MDM and MG cells) and PMN cells were strongly linked to the high IRE1 activity gene signature, whereas T cells were not (Figure 1A; Supplementary Figure 1A and B). Using flow cytometry, we confirmed the presence of both myeloid populations in freshly dissociated samples from an in-house GB cohort (n = 82) by quantifying expression of specific markers for each cell type (Figure 1B, C; Supplementary Figure 1C–F). Immune cell infiltration was higher in GB compared to grade II and III gliomas (Figure 1B), in particular CD45 + cells expressing MM and PMN markers (Figure 1C; Supplementary Figure 1F). MM constituted a majority of GB infiltrates,¹⁶ PMN were found in approximately 20% of the specimens, and T-cell infiltration was rare (below 10%) (Figure 1C). Remarkably, infiltration of immune cells was higher in tumors with high IRE1 activity (Figure 1D), including CD11b positive cells expressing MM markers CD14, CD64 (specific for MDM/MMG) and CD163 (specific for MDM); and PMN markers CD15 and CD66b (Figure 1E). Again, T cell recruitment was not dependent on IRE1 activity (Figure 1E). This suggests that IRE1 signaling could contribute to myeloid attraction, which in turn may promote tumor aggressiveness. To test this hypothesis, monocytes and neutrophils attraction was assessed in vitro. Myeloid cells from healthy donors’ blood, including monocytes (Mo) and PMN, were characterized based on the expression of CD14, CD15, CD66b, and CD16 (Supplementary Figure 1G). In addition, we generated the RADH87 line stably overexpressing wild-type (WT) IRE1 or the truncated IRE1 variant, Q780* which is devoid of its kinase and RNase domains (Supplementary Figure 1H and I). The latter mutation attenuated IRE1 signaling (Supplementary Figure 1J), therefore resembling the characteristics of U87 DN cells which are deficient for IRE1 activation.¹⁰ As shown in Figure 1F and G, conditioned media derived from U87 DN and RADH87 Q780* did not recapitulate the migratory abilities of Mo and PMN in the Boyden chamber assay, when compared to conditioned media from parental U87 and RADH87 cells or RADH87 IRE1 WT cells. Thus, IRE1 activation in GB cells promotes PMN and Mo chemotaxis. Lastly, we evaluated the role of IRE1 activity in myeloid recruitment in vivo using 2 syngeneic GB mouse models. As such, mice were orthotopically injected with GL261 cells. Fourteen days later, mice underwent surgical tumor removal, followed by insertion of either an empty gel implant (CTR) or an implant containing the IRE1 RNase inhibitor, MKC8866 (MKC).^5,9 Tumor size and overall survival obtained in both CTR and MKC groups were similar.⁹ We showed that pharmacological abrogation of IRE1 activity by MKC8866 significantly reduced PMN infiltration in the recurring tumors mainly at the tumor core (Figure 1H and I) and led to increased MM recruitment at the tumor periphery (Supplementary Figure 1K). Additionally, mice were orthotopically injected with CT2A cells and treated with a novel IRE1 inhibitor B2-1, derived from the IRE1 inhibitor Z4P,¹⁷ that crosses the brain blood barrier.¹⁸ Again, groups of CTR and B2-1 treated mice displayed similar tumor sizes and overall survival. However, the massive MM recruitment found in the CTR group was reduced upon IRE1 inhibition, similar to what was observed with PMN (Figure 1H and I). Overall, our results demonstrate that GB are infiltrated by myeloid cells, a phenomenon partly mediated by IRE1 signaling in tumor cells.

Figure 1.

Impact of IRE1 on myeloid recruitment to GB in vitro and in vivo. (A) Hierarchical clustering of TCGA GB patients based on high/low IRE1 activity was confronted to immune markers for MM, MDM, MG, PMN, and T cells. UP (n) and P values denote the proportion of signature genes that were found upregulated between the groups (n = number of genes) and the estimated 2-sided directional P-value of test, respectively. (B) Total immune infiltrate of GB (n = 82) and grade II/III (n = 8/14) glioma analyzed by flow cytometry using anti-CD45 antibodies. ^***P < .001 according to unpaired t-test compared to GB. (C) Percentage of specific GB infiltrated leukocytes populations analyzed by flow cytometry using anti-CD45/CD11b antibodies. Tumors were classified according to IRE1 activity (high/blue in red/blue) (n = 31). (D) Total immune infiltrate of GB classified according to IRE1 activity (n = 14/7) and analyzed by flow cytometry using anti-CD45 antibodies. Tumors were classified using transcriptome analysis according to IRE1 signature. ^***P < .001 according to unpaired t-test compared to tumors with high IRE1 activity. (E) Deeper characterization of immune subtypes was performed combining specific markers that is MM markers CD14, CD64 (for MDM/MMG), CD163 (for MDM), PMN markers CD15 and CD66B, and T marker CD3. Tumors were classified according to IRE1 activity (high/low, n = 10/5). ns: not significant, ^*P < .05 and ^**P < .01 according to unpaired t-test compared to tumors with high IRE1 activity. (F) and (G) Freshly isolated Mo and PMN were placed in Boyden chambers toward fresh medium (–), conditioned media from U87 (par.), U87 DN (DN), RADH87 (par.), RADH87 cells overexpressing IRE1 (WT) or Q780* (Q*) (n = 3, mean ± SD). ns: not significant, ^*P < .05, ^**P < .01, ^****P < .0001 according to unpaired t-test compared to media from parental. (H) Representative immunohistological analysis of myeloid infiltration in GB resected from GL261 and CT2A-implanted mice treated with plug with IRE1 inhibitor MKC8866 (for GL261) and B2-1 (for CT2A), respectively. MM and PMN were detected with anti-IBA1 and anti-Ly6G antibodies, respectively. Scale bar 100 µm. (I) Semi-quantitative analysis of IBA1 and Ly6G staining from (H). P values from unpaired t-test compared to CTR; ns: not significant.

Open in new tabDownload slide

IRE1 Activity Regulates Expression of Myeloid-Attracting Chemokines

We next hypothesized that IRE1 activity in tumor cells controls expression of myeloid-attracting chemokines. The latter have been described in different cancer models and include CXCL1, CXCL2, CXCL5, CXCL7, IL6, and IL8 for PMN¹⁹ and CCL2, CCL3, CCL5, CCL8, IL6, and IL8 for MM.²⁰ Using the TCGA dataset, we demonstrated that tumors with high IRE1 activity expressed high mRNA levels of the aforementioned cytokines/chemokines (Figure 2A). IRE1-dependent modulation of CXCL2, CXCL5, IL6, and IL8 was expected as these are part of the IRE1 activity signature.¹⁰ Increased mRNA expression of these cytokines/chemokines in tumors correlated with higher expression of CD14 and CD15/CD16, markers for MM and PMN, respectively (Figure 2B; Supplementary Figure 2A). To test whether those soluble factors promoted myeloid recruitment to GB, freshly isolated Mo and PMN were exposed to media conditioned by various GB lines, which resulted in significant induction of myeloid migration (Figure 2C). Using ELISA, we found that myeloid chemoattraction was correlated with elevated levels of CXCL2, IL6 and IL8 in these conditioned media (Figure 2C). PMN (but not Mo) attraction was partially blocked by SB225002, an antagonist of the IL8 receptor CXCR2 (Figure 2D). In line with our previous work,⁸ we showed that expression of myeloid-attracting chemokines depended on IRE1 activity. Indeed, expression of CXCL2, IL6 and IL8 mRNA but not CCL2 was dramatically reduced in U87 DN and RADH87 Q780* cells when compared to control or IRE1 WT-overexpressing cells (Figure 2E and F). Intriguingly, none of these chemokines/cytokines are known direct targets of XBP1s,²¹ suggesting an indirect contribution of XBP1s to this process. These findings indicate that IRE1 signaling, possibly through an indirect role of XBP1s, promotes the expression of chemokines involved in myeloid infiltration of tumors.

Figure 2.

IRE1-mediated synthesis of myeloid-attracting chemokines. (A) Hierarchical clustering of TCGA GB patients based on high/low IRE1 activity (n = 264/275) confronted to chemokines expression involved in myeloid chemoattraction. P values obtained using unpaired t-test comparing IRE1 high vs low tumors. (B) Chemokines mRNA expression in TCGA GB tumors with high/low MM and PMN infiltration determined according to CD14 or CD15 levels, respectively. CD14/CD15 high/low groups were determined using the median of the marker mRNA expression as cut-off (high/low n = 263/263). P values according to unpaired t-test comparing chemokines mRNA expression between CD14/CD15 high vs low tumors. (C) Correlation between chemokines secretion and Mo/PMN migration toward tumor conditioned media from different GB lines (n = 7 and 20 for Mo and PMN, respectively). R square and P values of the slopes were calculated using Pearson correlation coefficient analysis between chemokines secretion and myeloid attraction; ns: not significant and ^*P < .05. (D) Myeloid migration assay was performed as described in Figure 1, in the presence of SB225002, a CXCR2 antagonist (n = 3 and 4 for Mo and PMN, respectively, mean ± SD). ns: not significant and ^**P < .01 according to unpaired t-test compared to DMSO. (E) Quantification of myeloid-attracting chemokines mRNA abundance using RT-qPCR in parental U87 and RADH87 cells (–) transiently overexpressing TRAP-Nck (TRAP), IRE1-Nck (DN), IRE1-Q780* (Q) (n = 4). P values according to an ANOVA test comparing the 4 conditions. (F) Quantification of chemokines mRNA expression using RT-qPCR in parental U87 (par.), U87 DN (stably overexpressing IRE1-Nck), parental RADH87 or RADH87 cells stably overexpressing wild-type (WT) or Q780* (Q*) IRE1 (n = at least 6, mean ± SD). ns: not significant, ^**P < .01, ^***P < .001, ^****P < .0001 according to unpaired t-test compared to parental.

Open in new tabDownload slide

IRE1/XBP1s/RIDD Control Myeloid Infiltration to GB Via NFκB and UBE2D3

Since the expression of the aforementioned proinflammatory chemokines is controlled by NFκB,²² we tested whether a NFκB signaling gene signature was associated with IRE1, XBP1s and RIDD signatures previously established¹⁰ using the TCGA data. This revealed that both XBP1s and RIDD signaling were oppositely associated with elevated NFκB activation (Figure 3A). We also analyzed NFκB activation in IRE1 signaling deficient U87 DN cells compared to parental cells. The expression and phosphorylation of NFκB was lower in U87 DN cells compared to control (Supplementary Figure 2B), demonstrating a functional relationship between IRE1 signaling and NFκB. Next, we postulated that IRE1 through both XBP1s and RIDD could oppositely regulate molecular actors that control NFκB activation. We thus compared lists comprising XBP1s direct targets identified using ChIPseq²¹ with RIDD targets described in¹⁰; and identified 28 genes (Figure 3B). Among them, expression of 7 correlated (either positively or negatively) with the IRE1 signature and only 3 with XBP1s and RIDD signatures in GB samples (Figure 3B). Additionally, expression of only 1 of them, namely the E2 ubiquitin-conjugating enzyme UBE2D3, showed a marked positive correlation with NFκB activation (Figure 3C). Using TCGA transcriptome and RNAseq datasets, we showed that UBE2D3 mRNA expression was significantly higher in tumors with high IRE1 and XBP1s activities (Figure 3D) and in tumors expressing XBP1s mRNA (Figure 3E); but remained unaffected in tumors with high RIDD activity (Figure 3D). We then focused our attention on UBE2D3 as it also contributes to IκB degradation, and thus NFκB activation.^23,24 Intriguingly, we did not observe any effect on UBE2D3 mRNA or protein expression upon transient IRE1 inhibition using MKC8866 (Supplementary Figure 2C and D) or siRNA targeting either IRE1 (Supplementary Figure 2E) or XBP1 (Figure 3F and G; Supplementary Figure 2F–H). In addition, no effect was observed when the IRE1 activator IXA4 was used (Supplementary Figure 2C and D). However, UBE2D3 expression was reduced when IRE1 activity was stably attenuated in dominant negative U87 and RADH87 cells (Figure 3H). The effect was reversed by overexpressing XBP1s ectopically in these IRE1-defective cells (Figure 3I). This indicates that XBP1s could act as a transcriptional activator of UBE2D3 expression. Using the MatInspector and TFBIND tools, we found multiple potential XBP1s binding sites within the human and mouse UBE2D3 promoter region (Figure 3J; Supplementary Figure 2I). Gel shift assays revealed 2 sites for XBP1s binding on the human (sites h1 and h3, Figure 3K and L) and 1 on the mouse (site m1, Supplementary Figure 2J) UBE2D3 promoter regions. Remarkably, multiple bands presented slower mobilities corresponding to higher molecular weight entities, suggesting that XBP1s could be part of larger complexes. Both human and mouse sites on the UBE2D3 promoter are flanked by other transcription factors binding sites, some of which are known to associate with XBP1s (data not shown). These results suggest that UBE2D3 expression could be regulated by XBP1s. At this stage however, an indirect control of XBP1s on UBE2D3 transcription cannot be excluded and further investigations are needed.

Figure 3.

IRE1/XBP1-dependent regulation of UBE2D3. (A) IRE1, XBP1s, and RIDD signatures were confronted to NFκB signaling gene signature using the TCGA GB dataset (IRE1 high/low n = 264/275; XBP1s high/low n = 261/210; and RIDD high/low n = 285/249). P values obtained with unpaired t-test comparing IRE1, XBP1s, and RIDD high vs low tumors. (B) Venn diagram of the intersection of XBP1s target genes identified by ChIPseq²¹ with RIDD targets.¹⁰ (C) Association of NFκB signature with NCSTN, UBE2D3, and UFM1 low/high GB from TCGA. NCSTN, UBE2D3, and UFM1 high/low groups were determined using the median of the mRNA expression as cut-off (high/low n = 263/263). P values obtained from unpaired t-test comparing NCSTN, UBE2D3, and UFM1 high vs low tumors. (D) UBE2D3 mRNA expression in TCGA GB categorized according to their IRE1, XBP1s, and RIDD signatures (IRE1 high/low n = 258/265; XBP1s high/low n = 261/210; and RIDD high/low n = 252/258). P values obtained with unpaired t-test comparing IRE1, XBP1s, and RIDD high vs low tumors. (E) UBE2D3 mRNA expression in XBP1s low/high TCGA GB (RNAseq dataset; high/low n = 80/86). P value obtained from unpaired t-test comparing XBP1s high vs low tumors. (F) Quantitation of UBE2D3 mRNA expression using RT-qPCR in U87 and RADH87 cells silenced for XBP1 (n = 7 and 4 for U87 and RADH87, respectively). ns: not significant according to unpaired t-test compared to control. (G) Western blot analysis of UBE2D3 in parental (NT), control (siCTR) and XBP1-silenced (siXBP1) U87 and RADH87 cells. Actin (ACT) was used as loading control. Data are representative of 3 biological replicates (see Supplementary Figure S4F). (H) Quantification of UBE2D3 mRNA expression by RT-qPCR in cells with active (parental U87 and RADH87 (par)) and inactive IRE1 signaling (U87 DN and RADH87 Q780*) (n = 4, mean ± SD). ^*P < .05 according to unpaired t-test compared to parental. (I) Quantitation of UBE2D3 expression with RT-qPCR in U87 DN and RADH87 Q* cells and transfected with XBP1s (n = 3). ^*P < .05, ^***P < .001 according to unpaired t-test comparing CTR to XBP1s conditions. (J) Putative XBP1s binding sites on human UBE2D3 promoter regions analyzed with MatInspector and TFBIND. (K) and (L) Gel shift assays performed on 4 putative XBP1s binding sites using U87 nuclear extracts after XBP1s overexpression (K). Validation of putative binding sites h1, h2, and h3 using gradual amounts of unlabeled probes used in competition assay (L).

Open in new tabDownload slide

Conversely, UBE2D3 mRNA expression was decreased upon ER stress and this was significantly reduced upon treatment with MKC8866 (Figure 4A), suggesting that UBE2D3 is also targeted by RIDD upon ER stress. UBE2D3 functions together with E3 ligases including SYVN1.²⁵ Since SYVN1 was previously demonstrated to contribute to IRE1 stability,²⁶ we asked whether UBE2D3 also contributes to IRE1 expression regulation. As such modulation of UBE2D3 expression using either ectopic overexpression or siRNA silencing, respectively, led to reduced or increased IRE1 expression, independent of ER stress (Figure 4B). Furthermore, SYVN1 contributed to UBE2D3-dependent IRE1 degradation as shown using SYVN1 siRNA silencing in UBE2D3-overexpressing cells (Figure 4C). Notably, in contrast to UBE2D3, SYVN1 expression was neither modulated by IRE1 activity (Supplementary Figure 2K) nor associated with NFκB signaling pathway (Supplementary Figure 2L). Taken together, these results suggest that (i) in a basal/chronic ER stress situation, a prolonged inhibition of IRE1 activity (which is achieved by stable IRE1 DN overexpression, but not by a transient knockdown of IRE1/XBP1 or a short-term MKC treatment in vitro) is required to counteract IRE1-XBP1s-driven expression of UBE2D3 and that (ii) upon acute ER stress induction, transient IRE1 inhibition by MKC is sufficient to repress IRE1-mediated (RIDD-dependent) decrease of UBE2D3 expression. Thus, IRE1 via XBP1s and RIDD tightly controls UBE2D3 expression which, in turn, regulates IRE1 stability through SYVN1 as a feed-back mechanism (Figure 4D).

Figure 4.

IRE1/RIDD-dependent regulation of UBE2D3. (A) Western blot analysis of UBE2D3 in U87 cells treated with MKC8866 (MKC) under basal or ER stress condition using thapsigargin (Tg, 50 nM). P value obtained using an ANOVA test. (B) Western blot analysis of IRE1 and UBE2D3 in U87 cells silenced for UBE2D3 and UBE2D3 overexpressing RADH87 cells under basal or ER stress condition using tunicamycin (Tm, 1 µg/mL). (C) Western blot analysis of IRE1 and UBE2D3 in UBE2D3 overexpressing U87 cells silenced for SYVN1 under basal or ER stress condition using Tm. (D) Schematic representation of IRE1 regulation of UBE2D3 expression with a retro-control loop involving SYVN1.

Open in new tabDownload slide

Next, both XBP1s and RIDD signatures previously established¹⁰ were applied to the GB TCGA datasets to test whether these IRE1 signaling branches were associated with immune cell signatures. Importantly, the XBP1s signature was strongly linked to MM (including MG and MDM), PMN but not T immune signatures (Supplementary Figure 3A). In contrast, only the PMN signature was associated with the RIDD branch (Supplementary Figure 3A). Furthermore, amongst the IRE1 high/XBP1s high group, RIDD activity had no effect on myeloid recruitment, whereas PMN were positively regulated by XBP1s activity in the IRE1 high/RIDD high group (Supplementary Figure 3B). Taken together, these data suggest that XBP1s is the main responsible of IRE1-controlled GB infiltration by myeloid cells.

UBE2D3 Cooperates With MIB1 to Trigger NFκB Proinflammatory Response

To investigate how UBE2D3 modulates NFκB activation, we evaluated the impact of UBE2D3 overexpression in GB lines on the NFκB regulator IκB and on the subsequent activation of the NFκB pathway. UBE2D3 overexpression led to IκB degradation and concomitant increased phosphorylation of NFκB (Figure 5A; Supplementary Figure 4A–D). To identify the putative E3 ligase(s) involved in IκB degradation and/or NFκB activation as well as to investigate the global effect of UBE2D3 on protein ubiquitination in GB, we carried out a label-free quantitative MS/MS analysis using cells stably overexpressing UBE2D3 (Supplementary Figure 5). Details of this analysis are in the Supplemental material (Results). Among proteins modulated by UBE2D3 expression, we identified the E3 ligase MIB1, a known regulator of NFκB activation²⁷ and interactor with UBE2D3.²⁸ Therefore, we investigated whether MIB1 cooperates with UBE2D3 to trigger the degradation of IκB. We found that MIB1 silencing partially prevented the UBE2D3-mediated degradation of IκB protein (Figure 5B; Supplementary Figure 6A and B). However, MIB1 also controlled UBE2D3 directly as MIB1 silencing led to increased expression of both UBE2D3 mRNA (Supplementary Figure 6C) and protein (Figure 5B; Supplementary Figure 6A), suggesting a complex interaction between these molecules. Overall, these results suggest the existence of a signaling axis involving UBE2D3 and MIB1 that is sufficient to trigger the activation of NFκB-dependent proinflammatory response in GB (Figure 5G), and which needs to be further analyzed.

Figure 5.

Impact of UBE2D3 on NFκB activation and chemokines synthesis. (A) and (B) Western blot analysis of NFκB, phospho-NFκB, IκB, and phospho-IκB in control (EV) and transiently (U87) or stably (RADH87) UBE2D3 overexpressing cells (A); and after MIB1 downregulation (siMIB1) (B). UBE2D3 overexpression was checked. (C) Chemokines mRNA expression in TCGA GB specimens categorized according to UBE2D3 expression. UBE2D3 high/low (red/blue, n = 263/263) groups were determined using the median of the mRNA expression as cut-off. P values obtained from unpaired t-test comparing UBE2D3 high vs low tumors. (D) Quantification of chemokines expression using RT-qPCR in control (CTR) U87, parental (par.) RADH87, transient U87 and stable RADH87 cells overexpressing UBE2D3 (n = 3, mean ± SD). ^*P < .05, ^**P < .01, ^***P < .001, ^****P<.0001 according to unpaired t-test compared to control. (E) Quantification of chemokines expression using RT-qPCR in U87 (EV) or UBE2D3 overexpressing cells treated with 5 µM JSH-23 (n = 3, mean ± SD). ^*P < .05, ^**P < .01, ^****P < .0001 according to unpaired t-test compared to control. (F) Myeloid migration (Mo and PMN) was performed as described in Figure 1, toward media conditioned by U87 control (CTR) and UBE2D3 overexpressing cells (n = 3, mean ± SD). ^*P < .05 and ^***P < .001 according to unpaired t-test compared to control. (G) Schematic representation of UBE2D3 impact on inflammatory response in GB.

Open in new tabDownload slide

IRE1/UBE2D3 Control Myeloid Recruitment to GB Through NFκB Proinflammatory Response

To further investigate the role of UBE2D3 in myeloid mediated immunity, we stratified the TCGA GB cohort according to UBE2D3 mRNA levels and tested mRNA expression of the main myeloid-attracting cytokines/chemokines. We demonstrated that tumors with high UBE2D3 expression levels also expressed significantly higher levels of CCL2, CXCL2, IL6, and IL8 mRNA (Figure 5C). Next, we evaluated the expression of myeloid-attracting chemokines by RT-qPCR in UBE2D3-overexpressing cells. This showed that CCL2, CXCL2, IL6, and IL8 expression was markedly induced upon UBE2D3 overexpression (Figure 5D). In addition, treatment of U87 cells with the NFκB inhibitor JSH-23, which prevents NFκB activity, blunted the observed UBE2D3-dependent increase in CCL2 CXCL2, IL6, and IL8 mRNA expression (Figure 5E). We next found that MIB1 silencing partially prevented the UBE2D3-mediated upregulation of CCL2, CXCL2, IL6, and IL8 in UBE2D3-overexpressing cells compared to control (Supplementary Figure 6D). Finally, we analyzed Mo and PMN attraction properties of conditioned media from GB cells modified for UBE2D3 expression. We confirmed that media conditioned by UBE2D3-overexpressing cells increased Mo and PMN migration compared with media conditioned by control cells (Figure 5F), which was partially reversed after blocking IL6 and IL8 with antibodies (Supplementary Figure 6E). In contrast, UBE2D3 downregulation decreased Mo and PMN migration compared to control siRNA (siGL2) (Supplementary Figure 6F), and this effect was abolished by adding exogenously IL6 and IL8 (Supplementary Figure 6G). As expected, both cytokines were involved in Mo attraction, whereas only IL8 impacted PMN attraction (Supplementary Figure 6E–G), a result consistent with those presented in Figure 2D. In addition, Mo and PMN migration was reduced with media conditioned by UBE2D3 expressing RADH87 cells after MIB1 silencing, and this was reversed with exogenous IL6 and IL8 (Supplementary Figure 6H). Hence, we demonstrate that the signaling circuit involving IRE1/UBE2D3/MIB1 controls NFκB-mediated chemokines synthesis and inflammatory response in GB (Figure 5G).

UBE2D3 Promotes Pro-Tumoral Inflammation In Vivo and Increases GB Aggressiveness

To assess the importance of IRE1/UBE2D3 signaling axis in GB, we used a syngeneic mouse model.⁹ GL261 control and UBE2D3 overexpressing cells (Figure 6A) were orthotopically injected in immunocompetent C57BL/6 mice. Twenty-four days postinjection, tumors were resected, subjected to immunohistochemical analyses and the immune infiltrate was quantified. Interestingly, UBE2D3 GL261 cells produced larger tumors than their control counterparts (Figure 6B), although did not impact on mouse survival. This difference was not attributed to difference in proliferation rates in vitro (Supplementary Figure 7A), which suggests that the growth advantage of UBE2D3-overexpressing tumors might emerge from interaction with stroma and/or tumor microenvironment. Accordingly, UBE2D3 tumors showed elevated NFκB expression (Figure 6C) and recruited significantly higher numbers of MM and PMN (Figure 6D and E), consistent with our in vitro findings. Furthermore, in UBE2D3-overexpressing cells, ccl2, and cxcl2 mRNA levels were increased, as well as other NFκB-dependent factors including g-csf, il1b, and lif mRNA (Supplementary Figure 7B). Conditioned media from these cells also triggered Mo and PMN increased migration in vitro in a xenogeneic setting (Supplementary Figure 7C). Importantly, when ube2d3 was silenced in GL261 cells (Figure 6F), resulting tumors were smaller (Figure 6G), and mouse survival was extended compared to control cells (Figure 6H). Again, ube2d3 modulation did not influence GL261 proliferation in vitro (Supplementary Figure 7D), but affected mRNA expression of ccl2, cxcl2 as well as the factors under the control of NFκB activity, that is, g-csf, il1b, and lif mRNA (Supplementary Figure 7E). Furthermore, conditioned media from GL261 shube2d3 cells were less efficient to promote Mo and PMN attraction in vitro (Supplementary Figure 7F). Since mice lack both il8 and il8 receptor (cxcr1) genes, the PMN recruitment observed in our models could be due to cxcl1 and cxcl2, as previously reported in other mouse-bearing tumors infiltrated by PMN.^29,30 Using transcriptome data from 2 independent GB cohorts, GBMmark and TCGA-GBLGG, we observed a strong correlation between UBE2D3 expression and that of a large number of proinflammatory cytokines/chemokines (Supplementary Figure 7G and H). Intriguingly, high UBE2D3 expression was also associated with increased infiltration of monocytes, T cells and M2-polarized macrophages (Supplementary Figure 7I and J). Finally, to evaluate the clinical and prognostic relevance of UBE2D3 expression in GB, we investigated UBE2D3 expression in brain malignancies and showed that UBE2D3 was markedly increased in GB specimens compared to low-grade gliomas (Figure 6I). Patients whose tumors displayed high UBE2D3 expression were found associated with poorer prognosis, consistent with the fact that high IRE1 activity status was also of poor prognosis (Figure 6I). Our findings unveil a novel IRE1-dependent mechanism promoting pro-tumoral inflammation that integrates UPR signaling and ubiquitin system. Here, we demonstrated that IRE1/UBE2D3 axis controls GB secretome composition through NFκB signaling activation (Figure 6J).

Figure 6.

Impact of UBE2D3 overexpression on inflammation in vivo. (A) UBE2D3 protein overexpression in 5 UBE2D3 transfected GL261 stable lines. UBE2D3 protein level was measured with anti-Flag antibodies. (B) Left panel: brain sections from mice injected with GL261 control (CTR) or GL261_UBE2D3 cells analyzed for vimentin expression. Scale bar 1 mm. Right panel: tumor volume in control and UBE2D3 overexpressing (oe) group (n = 3 and 10, mean ± SD). ns: not significant according to unpaired t-test compared to control. (C)–(E) Left panel: Representative immunohistological NFκB expression (n = 7/15, mean ± SD) (C), macrophages/microglia infiltration (n = 18, mean ± SD) (D) and neutrophils infiltration (n = 24, mean ± SD) (E) in GL261 control or GL261_UBE2D3 tumors detected by anti-NFκB, anti-IBA1 and anti-Ly6G antibodies, respectively. Scale bar 100 µm. Right panel: semi-quantitative analyses of NFκB (C), IBA1 (D) and Ly6G staining (E) in control and GL261_UBE2D3 tumors. ^*P < .05, ^**P < .01 according to unpaired t-test compared to control. (F) UBE2D3 protein silencing in 2 GL261 stable lines transfected with shube2d3 construct. UBE2D3 protein level was measured using anti-ube2d3 antibodies. (G) Left panel: brain sections from mice injected with GL261 shCTR or shube2d3 GL261 cells analyzed for vimentin expression. Scale bar 1 mm. Right panel: tumor volume in shCTR and shube2d3 group (n = 4, mean ± SD). ^*P < .05 according to unpaired t-test comparing to control. (H) Mouse survival of mouse-bearing parental, shCTR and shube2d3 GL261 cells. ^***P < .001 according to unpaired t-test comparing to parental. (I) UBE2D3 expression in LGG and GB (mean; P value according to unpaired t-test compared to GB); and its impact on patients’ survival. (J) Schematic representation of IRE1/UBE2D3 axis in the regulation of pro-tumoral inflammation.

Open in new tabDownload slide

Discussion

In recent years, we have characterized the relevance of UPR signaling, in particular of the IRE1 arm, in GB^7,8 and found that characteristics of IRE1 signaling represent a predictive factor for GB aggressiveness.¹⁰ As such, modulating ER stress signaling pathways represents an attractive therapeutic avenue for GB treatment aimed at either increasing ER stress to levels that trigger apoptosis or decreasing the adaptive UPR, leading to loss of cellular selective advantages or increased sensitivity to treatments, and subsequent death.³¹ In addition to the intrinsic aggressiveness of GB cells, the brain tumor microenvironment, which contains among others endothelial and immune cells, is emerging as a crucial regulator of cancer progression. The most abundant immune cells in GB microenvironment are tumor-associated macrophages and microglial cells that might reach up to 30% of the tumor mass and are often linked to disease aggressiveness.^32,33 However, brain tumors are also infiltrated by other immune cells such as T, myeloid and plasmacytoid DCs, and neutrophils.³⁴

In the present study, we demonstrated that IRE1 signaling in tumor cells plays a key role in regulating the GB microenvironment, by promoting myeloid recruitment to the tumors. We previously found that IRE1 signaling was involved in macrophages and microglial cells recruitment to the tumors¹⁰ and that IRE1 controls proinflammatory chemokines expression.^5,35 Herein, we showed that pharmacological inhibition of IRE1 signaling decreased the extent of PMN infiltration into GB in vivo (Figure 1), which might be of clinical importance because elevated PMN recruitment correlated with poor outcome in GB patients. We also found that IRE1 activation in tumor cells was correlated with higher expression of myeloid-attracting chemokines (Figure 2). Importantly, this occurred in the absence of ER stress induction, underlining the important function of constitutive IRE1 activity in GB cells, that is modulating tumor secretome composition.

Mechanistically, we showed that IRE1 tightly and oppositely controlled the expression of UBE2D3 in GB cells by engaging the XBP1s and RIDD branches (Figures 3 and 4). This result might be indicative of a stress-dependent (nature, time, and intensity) regulation of UBE2D3 expression which could in turn find some spatial relevance in the whole tumor. We found that activation of IRE1/UBE2D3 signaling axis was in part responsible for myeloid chemoattraction through NFκB activation (Figure 5). UBE2D3 is an E2 ubiquitin-conjugating enzyme that, together with E1 ubiquitin-activating enzyme and E3 ligase, mediates attachment of ubiquitin moieties to target proteins. This posttranslational modification affects a broad range of biological processes, including protein quality control, trafficking, differentiation, cell division, signal transduction as well as inflammation.^36,37 UBE2D3 has been shown to control proteasomal degradation of, among others, p53,³⁸ cyclin D1,³⁹ p12 subunit of DNA polymerase δ,⁴⁰ and IκBα.^23,41 It was also reported to mediate RIG-I ubiquitination and thereby promote its activation upon viral infection to initiate a type I interferon-dependent innate immune response.⁴² In this study, we further demonstrated the crucial role of UBE2D3 in immunity/inflammation regulation in pathological conditions, such as cancer. We found that UBE2D3 is overexpressed in GB compared to low-grade gliomas, and that its elevated expression correlates with a high abundance of proinflammatory molecules. We delineated a novel IRE1-dependent mechanism for NFκB activation, which involves upregulation of UBE2D3 leading to IκB degradation through, at least partially, the E3 ligase MIB1 activity (Figure 5), the subsequent nuclear translocation of NFκB and its downstream signaling activation. Hence, IRE1 controls proinflammatory chemokines synthesis, including CXCL2, IL6, and IL8 as demonstrated herein. Once secreted, they not only sustain the pro-tumoral inflammatory microenvironment but can also mobilize immune cells recruitment to tumor sites further promoting cancer progression. As such, we showed in vivo that UBE2D3-overexpressing tumors were bigger in size and were infiltrated by significantly higher numbers of immune cells, such as MM and PMN (Figure 6). Furthermore, reduction of UBE2D3 led to decreased tumor growth in vivo and to the subsequent increased mouse survival. However, our findings indicate that the aforementioned mechanism might be applicable to the infiltration by a large number of lymphocytes, highlighting the importance of understanding the IRE1/UBE2D3 axis in other cancer models, particularly in “immune hot” tumors. Recent studies using single-cell technologies reveal a large range of transcriptome phenotypes from myeloid cells including both MDM and MG cells in GB.^13–15 The multidimensional heterogeneity of these myeloid populations (ie, resting, repressed MG, primed MDM) renders complex their identification and studies.^43,44 Notably, IRE1 is involved in macrophage polarization in melanoma.⁴⁵ Whether and how IRE1 could be involved in controlling MDM/MG cell states and/or functions has to be further studied. Furthermore, the crosstalk between myeloid cells might determine GB cell fate and nature that is, pro-neural to mesenchymal transition.⁴⁶

In the context of those findings, immunotherapy which has clearly proven its efficacy as an anti-cancer treatment,^47,48 has so far led to disappointing results in GB, likely because of the powerful immunosuppressive features of those tumors.⁴⁸ Neutrophil-activating therapy is emerging as a powerful anti-cancer immunotherapeutic approach in mouse breast, colon, and melanoma models.⁴⁹ Specific PMN activators including TNF, agonistic anti-CD40 and anti-tumor antibodies allow the PMN recruitment at metastatic sites and activate PMN-mediated antibody-dependent cytotoxicity against tumor cells. The current work implies that targeting IRE1 signaling might impede GB aggressiveness by reducing tumor cell invasion and angiogenesis,^8,10 but also by attenuating pro-tumoral inflammation and immunity. This study opens a new avenue for therapeutic approaches to improve the efficacy of current and future immunotherapies.

Conflict of interest statement

E.C. is the founder of Thabor Therapeutics (www.thabor-tx.com). The other authors declare no conflicting interests.

Funding

This work was funded by grants from la Ligue Contre le Cancer (Comités de la region Grand-Ouest: Côtes d’Armor, Ille-et-Vilaine, Indre, Morbihan et Vendée), from INSERM, from Région Bretagne and from l’Institut des Neurosciences Cliniques de Rennes to A.T.; from INSERM (IRP 2020 Tupric), Institut National du Cancer (INCa; PLBio 2017, 2019, 2020), Région Bretagne, Rennes Métropole, Fondation pour la recherche Médicale (F.R.M.; équipe labellisée 2018 and DEQ20180339169), EU H2020 MSCA ITN-675448 (TRAINERS), la Ligue Contre le Cancer and MSCA RISE-734749 (INSPIRED) to E.C.; ELIXIR-GR “The Hellenic infrastructure for biological data management and analysis” (MIS: 5002780), “Reinforcement of the Research and Innovation Infrastructure”, and the Operational Programme “Competitiveness, Entrepreneurship and Innovation” (NSRF 2014-2020) co-financed by Greece and the European Union (European Regional Development Fund) to A.C.

Data Availability

Materials described in the manuscript, including all relevant raw data, are freely available to any researcher wishing to use them for noncommercial purposes, without breaching participant confidentiality.

References