|Home | About | Journals | Submit | Contact Us | Français|
Tumor-associated macrophages play critical roles during tumor progression by promoting angiogenesis, cancer cell proliferation, invasion, and metastasis. Cysteine cathepsin proteases, produced by macrophages and cancer cells, modulate these processes, but it remains unclear how these typically lysosomal enzymes are regulated and secreted within the tumor microenvironment. Here we identify a STAT3 and STAT6 synergy that potently upregulates cathepsin secretion by macrophages via engagement of an unfolded protein response (UPR) pathway. Whole genome expression analyses revealed that the TH2 cytokine IL-4 synergizes with IL-6 or IL-10 to activate UPR via STAT6 and STAT3. Pharmacological inhibition of the UPR sensor IRE1α blocks cathepsin secretion and blunts macrophage-mediated cancer cell invasion. Similarly, genetic deletion of STAT3 and STAT6 signaling components impairs tumor development and invasion in vivo. Together, these findings demonstrate that cytokine-activated STAT3 and STAT6 cooperate in macrophages to promote a secretory phenotype that enhances tumor progression in a cathepsin-dependent manner.
Tumor-associated macrophages (TAMs) play critical roles in multiple stages of tumorigenesis and also contribute to chemoresistance (Quail and Joyce 2013; Noy and Pollard 2014; Ruffell and Coussens 2015). Through reciprocal interactions with cancer cells and immune cell infiltrates, TAMs sculpt the tumor microenvironment (TME) to regulate critical aspects of tumor progression, including local inflammation, angiogenesis, cancer cell invasion and intravasation into the circulation (Qian and Pollard 2010). At the molecular level, this is achieved via a wide range of TAM-derived cytokines, growth factors and proteases that mediate autocrine/paracrine signaling as well as modifications to the extracellular matrix (ECM). These factors often enhance the malignant phenotype of cancer cells and thus impart additional advantages to these cells within the TME. Therefore, understanding how TAM-derived factors interact and collectively regulate cancer cell behavior is essential for elucidating TAM biology and developing novel TAM-based therapies.
Amongst TAM-supplied proteases, members of the cysteine cathepsin family have been implicated in multiple aspects of carcinogenesis, including cancer cell proliferation, angiogenesis, invasion and metastasis, in different tumor types (Olson and Joyce 2015). Typically, cathepsins are synthesized as pro-form zymogens that can be activated in the acidic lysosomal compartment, and thus execute intracellular degradation of proteins delivered to the lysosome (Turk et al. 2012). However, numerous studies have shown that several cathepsin family members also perform extra-lysosomal functions, such as antigen presentation (Riese et al. 1996), inflammasome activation (Hornung et al. 2008), and growth factor processing (Wiley et al. 1985). Importantly, when secreted into the extracellular space these enzymes can degrade the ECM and basement membranes, cleave cell-adhesion molecules, and participate in proteolytic cascades resulting in sequential protease activation, which collectively promote tumor invasion and metastasis (Olson and Joyce 2015). Therefore, elucidation of the molecular mechanisms governing the secretion and extracellular activities of cathepsins is critical.
Previously, we demonstrated that IL-4 signaling is at least partially responsible for inducing high cathepsin activity levels in TAMs, in addition to its role in controlling other facets of macrophage biology (Gocheva et al. 2010b; Wang and Joyce 2010). However, as TAMs typically reside in complex TMEs composed of diverse cytokines, it remains to be determined whether other factors operate in concert with IL-4 to mediate coordinated responses. Indeed, aside from IL-4, other cytokines including IL-13, IL-6 and IL-10 readily activate macrophages (Biswas and Mantovani 2010). It has recently been reported that several TH2-associated cytokines coordinately activate STAT6 (the downstream effector of IL-4 and IL-13) and STAT3 (the downstream effector of IL-6 and IL-10) to regulate TH2 cell differentiation (Stritesky et al. 2011), suggesting a crosstalk mechanism responsive to multiple cytokine inputs. However, it is currently unknown whether similar machinery also operates in macrophages to modulate their activation and functions.
To address this question we have used genetic approaches to demonstrate that both STAT3 and STAT6 signaling critically contribute to tumor development, and cancer cell invasion in vitro and in vivo. We determined that STAT3/STAT6-activating cytokines cooperate to regulate cathepsin expression and more potently, secretion, in bone marrow-derived macrophages (BMDMs). Whole genome expression analyses revealed that combined cytokine treatment of BMDMs led to pronounced transcriptional reprogramming highlighted by the upregulation of secretion-associated genes. Furthermore, we found that the synergistic induction of secretion of pro-cathepsins is predominately regulated by activation of the inositol-requiring enzyme 1α (IRE1α) axis of the unfolded protein response (UPR) pathway.
TAMs reside in complex microenvironments that are typically TH2-skewed (Biswas and Mantovani 2010; Shiao et al. 2011). Previously, we showed that the TH2 cytokine IL-4, which is progressively upregulated during pancreatic neuroendocrine tumor (PanNET) development in the RIP1-Tag2 (RT2) mouse model, is an important inducer of high global cysteine cathepsin activity in TAMs which peaks in the most invasive of RT2 tumors (Gocheva et al. 2010b). However, in this earlier study we did not address whether the overall increase in pan-cathepsin activity was the result of transcriptional, translational or localization changes in specific cysteine cathepsin family members, of which there are eleven in humans (Turk et al. 2012). Moreover, given that Il4 deletion only reduced the proportion of cathepsin activity-high TAMs by approximately 40% (Gocheva et al. 2010b), additional cytokines in the TME are likely to regulate this important phenotype of TAMs.
To directly interrogate TH2-associated cytokine signaling programs in vivo, we first quantified the expression levels of the signaling components in individual cell populations within the PanNET TME. We used fluorescence-activated cell sorting (FACS) to isolate TAMs, T cells and cancer cells, based on their distinct patterns of surface marker expression. Separation of pure cell populations was confirmed by exclusive expression of Cd68 in macrophages, Cd3e in T cells, and SV40-Tag in cancer cells (Supplemental Fig. S1A). qRT-PCR analysis showed that Il4, Il10 and Il13 were mainly expressed in T cells or TAMs, while Il6 was uniformly expressed across all cell populations (Fig. 1A). Specifically, T cells expressed the highest levels of Il4 and Il13, whereas Il10 expression was most prominent in macrophages (Fig. 1A). TAMs expressed the corresponding cytokine receptors at both the mRNA and protein level (Fig. 1,C), as well as the downstream transcriptional mediators Stat3 (for IL-6 and IL-10) and Stat6 (for IL-4 and IL-13) (Supplemental Fig. S1B). We found that TAMs expressed high levels of genes associated with an IL-4 driven alternative-activation phenotype including Ccl8 (Hiwatashi et al. 2011), Ccl12, and F13a1 (Ostuni et al. 2013), indicating the capacity of TAMs to receive the TH2 cytokine inputs (Fig. 1D).
We next examined the role of TH2-associated cytokine signaling in tumor development by genetically deleting multiple components of the IL-4 signaling axis in the RT2 model. End-stage analysis at 13.5 weeks revealed that mice with constitutive of the Il4ra receptor led to a significant decrease in tumor burden compared to wild-type (WT) RT2 animals (Fig. 2A). Genetic deletion of Stat6, the critical downstream mediator of IL-4Rα signaling, also resulted in a significant decrease in cumulative tumor volume (Fig. 2B). Because heterozygous Stat6 deletion resulted in a similar reduction in tumor burden compared to Stat6 homozygous knockouts (Fig. 2B), we sought to determine whether Stat6+/− BMDMs still respond to IL-4 stimulation (Supplemental Fig. S1C). Our data suggests that while Stat6+/− BMDMs retained some responsiveness to IL-4, the magnitude of IL-4-mediated activation was significantly attenuated, as indicated by the diminished induction of Arg1, Ccl22, Ccl8 and Ccl12 (Supplemental Fig. S1C). Therefore, the reduction of tumor volume observed in Stat6+/− RT2 mice may be the result of partial Stat6 loss in both TAMs and tumor cells.
To define the contribution of bone marrow (BM)-derived cells to these phenotypes, we performed bone marrow transplantation (BMT) experiments. We have previously shown that the vast majority (88%) of BM-derived cells in RT2 tumors differentiate into TAMs (Gocheva et al. 2010b), and thus BMT provides a strategy to experimentally manipulate the expression of TAM-supplied factors such as STAT6 in vivo. We transferred WT or Stat6−/− donor BM into lethally irradiated WT RT2 recipients at 4 weeks of age, and subsequently assessed end-stage tumor volume at 13.5 weeks. Mice transplanted with Stat6−/− BM showed a markedly lower tumor burden compared to WT BM counterparts (Supplemental Fig. S1D). Together, these findings demonstrate that the IL-4/STAT6 signaling pathway critically contributes to tumor development.
The presence of intratumoral STAT3-activating cytokines (IL-6, IL-10) in the TME prompted us to also delineate the role of STAT3 in PanNET development. Stat3-deficient animals are embryonic lethal (Takeda et al. 1997), and therefore we conditionally deleted Stat3 in myeloid cells of RT2 mice using the LysM:Cre line (hereafter Stat3Δ/Δ) (Takeda et al. 1999). Mice with heterozygous or homozygous Stat3 deletion showed a significantly lower tumor burden compared to WT RT2 (Fig. 2C). To next determine if Stat6 and Stat3 act in an additive or epistatic manner, we generated Stat6−/− RT2 mice with either heterozygous or homozygous Stat3 deletion, and analyzed end-stage tumor burden. Compared to Stat6−/− mice, animals with additional Stat3 deletion had significantly lower tumor volume (Fig. 2D). Collectively, these genetic experiments show that STAT6 and STAT3 synergistically regulate PanNET progression in vivo.
Previously we have shown that IL-4 increases pan-cathepsin activity in TAMs, and TAMs in turn promote cancer cell invasion (Gocheva et al. 2010b). We therefore examined the effect of genetically ablating STAT3/STAT6 signaling components on tumor invasiveness in vivo. Based on histological features, RT2 tumors are classified into encapsulated tumors, microinvasive carcinomas (IC1) and frankly invasive carcinomas (IC2), as previously described (Lopez and Hanahan 2002; Gocheva et al. 2010b). WT RT2 tumors comprise all three grades, with IC1 being the predominant type. To determine the impact of STAT6 and STAT3 on tumor invasion, we compared Stat6−/− tumors, Stat3Δ/Δ tumors, and Stat6−/− Stat3Δ/Δ tumors with their WT counterparts. Stat6 deletion significantly impaired tumor invasion compared to WT. While deletion of Stat3 also trended toward reduced tumor invasion, this was not statistically significant compared to WT tumors (Fig. 2E). Similarly, the combined ablation of both Stat6 and Stat3 did not further decrease tumor invasion compared to Stat6 deletion alone, though there was a trend towards less invasive IC2 tumors. In sum, these results suggest that perturbation of the STAT6 signaling pathway is sufficient to limit the invasive capacity of tumors.
Given that STAT3 and STAT6 cooperate during TH2 cell development (Stritesky et al. 2011), and their synergistic functions in promoting RT2 tumor growth and invasion (Fig. 2), we hypothesized that concurrent activation of STAT6 and STAT3 could potently enhance the pro-tumorigenic functions of macrophages. We specifically focused on macrophage-derived cathepsins, because abundant evidence has established their importance in PanNET progression and invasion (Joyce et al. 2004; Gocheva et al. 2006; Gocheva et al. 2010a; Gocheva et al. 2010b; Akkari et al. 2014; Akkari et al. 2016). We first stimulated BMDMs with either IL-4 alone or a ‘cytokine cocktail’ (comprising IL-4, IL-6, IL-10 and IL-13) for six consecutive days. We specifically analyzed expression of the six of eleven cathepsin family members (CtsB, C, H, L, S, and Z) that we had previously shown to be upregulated during multistage RT2 tumorigenesis (Joyce et al. 2004). Expression analysis demonstrated modest additional upregulation of CtsB, CtsC, CtsS and CtsZ mRNA after combined cytokine stimulation compared to IL-4 stimulation alone (Fig. 3A). The mRNA level of CtsL, however, was dramatically elevated by the cytokine cocktail compared to IL-4 (Fig. 3A). CtsH was the only family member that did not show further enhanced expression by combined cytokines (Fig. 3A). These findings highlight the differential transcriptional regulation of cathepsin family members following TH2-associated cytokine treatment.
Next, we analyzed the levels of secreted cathepsins under individual and combinatorial cytokine conditions using macrophage-conditioned media (CM) labeled with DCG-04 (Greenbaum et al. 2000), a pan-cathepsin probe which has been widely used to assay cathepsin abundance and activity (e.g. (Joyce et al. 2004; Rooney et al. 2005; Vasiljeva et al. 2006; Chandramohanadas et al. 2009)). While IL-4 and the other TH2-associated cytokines increased pan-cathepsin secretion, the ‘cocktail’ treatment resulted in the most robust response (Fig. 3B), with all treatments leading to some increase in overall protein secretion compared to controls (Supplemental Fig. S2A). Interestingly, a substantial proportion of cathepsins were secreted as pro-forms, as indicated by the alterations of their molecular weights following incubation in cathepsin activation buffer (Fig. 3B). Western blots for individual cathepsins confirmed that combined cytokine treatment robustly upregulated the secretion of CtsB, CtsL, CtsS and CtsZ (Supplemental Fig. S2B). Compared with single cytokines, combined cytokines also triggered upregulation of intracellular CtsL protein but not the other cathepsins (Supplemental Fig. S2C), which corresponds to the pronounced transcriptional induction of CtsL (Fig. 3A). By contrast, discordance between transcriptional and secretory regulation is highlighted by the lack of alteration in CtsC secretion following combined cytokine treatment (Supplemental Fig. S2B), despite a modest increase at the mRNA level (Fig. 3A). Cathepsin secretion was more potent with IL-4, as IL-13 did not further promote cathepsin secretion when combined with IL-6 or IL-10 (Supplemental Fig. 3A-D). These results demonstrate that cathepsin abundance and localization are differentially regulated by TH2-associated cytokines, with secretion being the most prominent effect.
To gain further insight into these regulatory complexities, we focused on the downstream mediators of TH2 cytokine signaling, STAT6 and STAT3. We combined IL-4 with either IL-6 or IL-10 to non-redundantly stimulate WT, Stat6−/−, or Stat3Δ/Δ BMDMs. Both cytokine combinations resulted in a synergistic upregulation of cathepsin secretion by WT BMDMs, which was abolished by Stat6 or Stat3 inactivation (Fig. 3C). Taken together, our data suggest that both STAT6 and STAT3 are required for the TH2-associated cytokine-induced synergistic secretion of cathepsins by macrophages.
We determined that while STAT6 and STAT3 were capable of mediating cathepsin transcription, their direct transcriptional regulation was modest and thus not sufficient to explain the greater magnitude of the enhanced secretory phenotype for several cathepsin family members. We reasoned that additional STAT6 and STAT3 targets might mediate the synergistic cathepsin secretion from stimulated BMDMs. Therefore, we performed whole genome expression profiling to gain systematic insights into cytokine-mediated transcriptional changes in macrophages. Principal component analysis (PCA) revealed that IL-4 induced a distinct repertoire of transcriptional changes in BMDMs compared with IL-6 or IL-10, and combined cytokines modified the transcriptional landscape differently from single cytokines (Fig. 4A). When analyzing the interactions between IL-4 and IL-6, we defined six patterns of gene expression: (I) genes upregulated by IL-4, but not IL-6 nor the combination of IL-4 + IL-6, (II) genes upregulated by IL-6, but not IL-4 nor the combination of IL-4 + IL-6, (III) genes upregulated by IL-4 and IL-4 + IL-6 together, but not IL-6 alone, (IV) genes upregulated by IL6 and IL-4 + IL-6, but not IL-4 alone, (V) genes upregulated by IL-4, IL-6, and the combination of IL-4 + IL-6, (VI) genes responsive only to the combination of IL-4 + IL-6 (Fig. 4B; complete gene expression results can be found in Supplemental Table 1). We found 133 genes upregulated by IL-4 treatment, including well-known responsive genes such as Arg1, Ccl17, Ccl22, and Irf4 (El Chartouni et al. 2010; Wang and Joyce 2010) (Fig. 4B,C). In addition we find that both IL-6 and IL-4+IL-6 stimulation induce ll4ra (Fig. 4B,C; Supplemental Fig. S4B), a finding previously proposed to support the role of IL-6 in alternative macrophage activation (Mauer et al. 2014). Moreover, 187 genes (pattern IV, including Cd163, Cd209a/DC-SIGN, and Cd274/PD-L1) were upregulated by IL-4 + IL-6 in combination but showed no response to either cytokine alone (Fig. 4B,C). These data highlight that cytokine interactions result in complex changes in the transcriptome in a non-additive manner.
In accordance with PCA, we found that there was considerable transcriptional similarity between IL-6-stimulated and IL-10-stimulated BMDMs, with only Rsad2 and Ifit3 being specifically upregulated in IL-6-treated macrophages (Fig. 4A, Supplemental Fig. S4A). Despite such a substantial overlap in transcriptional regulation, concurrent IL-4 stimulation dramatically amplified the differences between IL-6 and IL-10, with many more genes being differentially regulated in the combination conditions (Fig. 4A, Supplemental Fig. S4A,B). Specifically we found 33 genes upregulated by combined IL-4 + IL-6 compared to IL-4 + IL-10, including Il4ra, Socs3, and Cd209a (Fig. 4C, Supplemental Fig. S4A,B). Meanwhile, 13 genes were upregulated in response to combined IL-4 + IL-10 compared to IL-4 + IL-6, including Ccl17, Ccl22, and Irf4 (Fig. 4C, Supplemental Fig. S4B). These findings demonstrate the non-redundancy of the two STAT3-activating cytokines IL-6 and IL-10 when combined with IL-4 stimulation.
To comprehensively assess these complex cytokine interactions, we used an interaction-based linear model to statistically quantify the extent to which target genes were regulated by single and combination cytokine treatment (Ritchie et al. 2015). This approach allowed us to more accurately distinguish IL-4 responsive genes (pattern III) from exclusively IL-4 + IL-6 responsive genes (pattern IV) for broader gene ontology analyses. In the most extreme cases ‘Induced’ genes (n=82) showed minimal response to single cytokines but striking increases in gene expression upon combined cytokine stimulation (Fig. 5A). Conversely, ‘Repressed’ genes (n=61) showed the opposite trend where either of the combined cytokine treatments led to a decrease in expression (Fig. 5A). In addition, we observed globally that the magnitude of synergy was most pronounced in the combined IL-4 + IL-6 condition, whereas IL-4 + IL-10 treatment led to a modest response (Supplemental Fig. S4C), confirming the trends observed for individual genes (Fig. 4D). We next sought to summarize these global changes in gene expression using gene ontology (GO) analysis. We focused on the combined IL-4 + IL-6 interaction term to assign a ‘synergy score’ to each gene and assessed GO enrichment using iPAGE (Goodarzi et al. 2009). Interestingly, GO analysis uncovered an enrichment of Golgi-vesicle trafficking for the genes with the highest synergy score (Fig. 5B, far right column), which contains GO terms related to secretion, vesicle transport, and exocytosis (Supplemental Fig. S4D). These findings highlight the phenotypic switch of macrophages to a secretory state upon combined cytokine stimulation.
Given these findings, we next sought to understand the mechanisms regulating protease secretion following combined cytokine treatment. Having already identified a role for STAT6 and STAT3 in mediating cathepsin secretion (Fig. 3C), we utilized Motif Activity Response Analysis (MARA) (Suzuki et al. 2009) to gain additional mechanistic insight into putative downstream pathways regulating this process. As expected, transcription factor (TF) motifs representing STAT6 and STAT3 demonstrated increased activity under both individual and combined cytokine conditions compared to the unstimulated condition (Supplemental Fig. S4E). We next focused our attention on TF motifs that demonstrated synergistic induction in response to combined cytokine treatment, in order to identify pathways regulating secretion and vesicle trafficking-related processes. We identified 8 TF motifs that were induced under the combined cytokine condition: NFE2, ATF6, Rfx-family members, HOXA5/B5, NR1H4, ATF2, ARNT, and XBP1 (Fig. 5C). Only XBP1 and ATF6 targets were enriched for Golgi vesicle-related GO terms (Fig. 5C,D). Interestingly, these factors have been shown to be the major mediators of the unfolded protein response (UPR), and XBP1 in particular has been implicated in antibody secretion from plasma cells (Reimold et al. 2001; Osorio et al. 2014). Collectively, these analyses suggest that combined cytokine stimulation activates UPR-associated TFs in macrophages, potentially resulting in a phenotypic switch to a secretory state.
There are three arms of the classical UPR, mediated via IRE1α/XBP1, PERK, and ATF6 respectively, which are critical for regulating protein aggregates and thereby preventing ER stress (Bettigole and Glimcher 2015). We first analyzed Xbp1 splicing, which serves as a read-out of upstream IRE1α activation. Consistent with the TF MARA analyses, spliced-Xbp1 (sXbp1) was significantly upregulated under the combined cytokine conditions (Fig. 6A, Supplemental Fig. S5A). A time-course analysis of Xbp1 splicing revealed induction as early as 2 hours post-stimulation, which peaked at 24 hours, and remained elevated after 48 hours (Fig. 6B). By contrast, Stat6−/− and Stat3Δ/Δ BMDMs showed either severely attenuated or no sXbp1 induction upon combined cytokine stimulation (Fig. 6C-D, Supplemental Fig. S5A). Similar kinetics of IRE1α phosphorylation and its Stat3/Stat6 dependence were identified, indicating activation of this pathway (Supplemental Fig. S5B).
We next assessed activation of the PERK pathway, another arm of the UPR, which was modestly activated by combined cytokines (Fig. 6E). To determine if there was a functional contribution of the PERK pathway to cathepsin secretion, BMDMs were pre-incubated with a pharmacological inhibitor of PERK, GSK2606414, followed by cytokine stimulation. However, PERK inhibition did not significantly alter the synergistic cathepsin secretion induced by combined cytokines (Fig. 6F). Finally, we examined the third component of the classical UPR, ATF6. Following IL-4 plus IL-6 treatment, we did not detect robust alterations in ATF cleavage nor its nuclear localization (Fig. 6E). When compared with UPR triggered by bona fide activators, including tunicamycin and thapsigargin, combined cytokine-mediated UPR shows modest induction of targets (Supplemental Fig. S5A, S6A). In addition, IL-4 is a more potent inducer of UPR than IL-13, when combined with IL-6 or IL-10 (Supplemental Fig. S6B). Together, these findings suggest that TH2-associated cytokines trigger an attenuated, non-canonical UPR in macrophages, which potentially leads to the secretory phenotype.
Having excluded a role for PERK or ATF6 in controlling the secretory phenotype, we therefore sought to determine whether IRE1α was the key regulator. Transient siRNA-mediated knockdown of IRE1α resulted in significant attenuation in UPR induction by combined cytokines (Supplemental Fig. S7A,B). We also used a small molecule inhibitor of the RNase domain of IRE1α, STF-083010 (Papandreou et al. 2011) to assess the role of IRE1α in UPR induction in macrophages. We found that STF-083010 blocked cytokine-induced Xbp1 splicing (Supplemental Fig. S7C) and sXBP1 protein abundance (Supplemental Fig. S7D), indicating functional inhibition of IRE1α. Similarly, the induction of UPR genes (Bip, Pdi) was also dependent upon IRE1α activity (Supplemental Fig. S7C).
We then investigated whether perturbation of IRE1α would blunt TH2 cytokine-mediated cathepsin secretion. IRE1α knockdown completely abolished cathepsin secretion from BMDMs treated with combined cytokines (Fig. 7A). This finding was further corroborated by using three independent IRE1α pharmacological inhibitors, STF-083010, 4μ8C and KIRA6 (Fig. 7B, Supplemental Fig. S7E,F). Interestingly, this phenomenon appears to extend beyond cathepsins, as secretion of the lysosomal enzyme legumain (LGMN) also depended on IRE1α activity (Fig. 7A,B, Supplemental Fig. S7E,F). By contrast, the secretion of MMP9 and MMP13, two representative MMPs produced in abundance by macrophages, was not synergistically regulated by combined cytokines, and IRE1α inhibition did not block MMP secretion (Fig. 7A,B). Together, these findings indicate that IRE1α-dependent synergistic secretion of proteins from macrophages is specific to a subset of lysosomal proteases.
Our in vivo data suggest that TAMs can enhance tumor invasion. To investigate how cytokine signaling in TAMs promotes this process, we utilized a transwell invasion assay, in which βTC374 cancer cells (a cell line derived from a WT RT2 tumor) were plated on reconstituted extracellular matrix pre-processed with conditioned media from cytokine-primed BMDMs. We found that conditioned media from combined cytokine-treated BMDMs resulted in the highest enhancement of cancer cell invasion, whereas single cytokine priming led to no significant alteration in invasion (Fig. 7C). Considering the marked effect of TH2-associated cytokines on cathepsin secretion, we used a pan-cathepsin inhibitor JPM-OEt to determine whether cathepsins are responsible for macrophage-mediated cancer cell invasion. Indeed, addition of JPM-OEt abolished cancer cell invasion mediated by combined cytokine-treated BMDMs, indicating the crucial role of cathepsins in driving this process (Fig. 7C). Moreover, the IRE1α inhibitor STF-083010 similarly impaired the ability of combined cytokine-stimulated BMDMs to promote tumor cell invasion (Fig. 7C), which was likely caused by diminished cathepsin secretion. Taken together, our data demonstrate that TH2-associated cytokines synergistically promote cathepsin secretion by macrophages in an IRE1α-dependent manner, which in turn facilitates cancer cell invasion.
Macrophages are capable of performing diverse functions in response to various signaling inputs. Previously we showed that, in addition to its role in driving M2-like/alternative polarization, IL-4 upregulates cathepsin activities in TAMs to facilitate tumor development (Gocheva et al. 2010b). Here, using genetic strategies, we find that deletion of TH2 cytokine signaling components in TAMs is sufficient to impair tumor growth and invasion in vivo; a finding which mirrored previous results where we showed that TAM-derived CtsB and CtsS played critical roles in RT2 tumor growth and invasion (Gocheva et al. 2010b). In the current study, we demonstrate that IL-4 cooperates with other cytokines, specifically IL-6 and IL-10, to mediate synergistic induction of cathepsin transcription and secretion in BMDMs. We found that the enhanced secretory phenotype of macrophages is concomitant with engagement of the UPR. Strikingly, transient knockdown and pharmacological inhibition of IRE1α led to a complete blockade in cathepsin secretion and a subsequent reduction in macrophage-mediated cancer cell invasion.
This study presents evidence for the first time that TH2 cytokines and IRE1α regulate protease secretion from macrophages, an effect that was specific to the secretion of pro-form lysosomal proteases, but not MMPs. This pro-form protease secretion suggests a potential re-routing of lysosomal proteases before they reach the lysosome, as we found no change in lysosomal function or acidification (data not shown). In contrast, previous reports demonstrated that IFN-γ, an inducer of M1-like macrophage activation, increases secretion of active-form CtsL (Beers et al. 2003). One explanation for this difference may be that combined TH2 cytokine-treated macrophages upregulate CtsL and concomitantly downregulate the manose-6-phosphate (M6P) receptors Igf2r and M6pr (Supplemental Table 1). This could lead to saturation of the M6P sorting system, whereby excess lysosomal enzymes are secreted into the extracellular space as pro-forms.
Moreover, it has been shown that the downstream effector of IRE1α, XBP1, controls the secretion of a subset of pro-inflammatory cytokines upon TLR2 or TLR4 ligation, which polarizes macrophages to an M1-like state (Martinon et al. 2010). Combined with our findings, these data suggest that the IRE1α/XBP1 pathway may impact secretion across different macrophage polarization states, albeit with distinct outputs. It is also important to note that IRE1α/XBP1 activation may not always be accompanied by a pronounced ER stress. In our studies, stimulation with combined cytokines induced classical ER stress targets (Bip, Chop and Pdi) and activated PERK, but did not engage the ATF6 axis. Moreover, the magnitudes of such induction were significantly lower than those triggered by classical ER stress inducers, such as tunicamycin and thapsigargin.
These findings indicate that combined cytokine stimulation may instigate a low level of ER stress, or perhaps no stress at all. Recently, CD8a+ dendritic cells have been shown to engage the UPR in the absence of ER stress to mediate antigen cross-presentation (Osorio et al. 2014). Similar physiological engagement of the UPR has been identified in differentiating B cells where expansion of the ER is dependent upon ATF6 and IRE1α, and importantly, precedes the production of immunoglobulin (Iwakoshi et al. 2003; van Anken et al. 2003). In addition, thyroid-stimulating hormone-stimulated thyrocytes activate the UPR preemptively to maintain thyroglobulin secretion (Christis et al. 2010). In light of these reports, our data suggests that macrophages stimulated with IL-4 and IL-6/IL-10 may engage the IRE1α/XBP1 pathway to mediate the secretory load of lysosomal proteases.
Our results also emphasize cathepsin protease production as a functional aspect of the tumor-promoting TAM phenotype. We previously found that cathepsin activity increases during the course of disease progression in PanNETs and breast cancer, and that TAM-supplied cathepsins substantially contribute to tumor growth, invasion and angiogenesis in a PanNET mouse model, with the highest level of cathepsin activity present in invasive IC2 tumors (Gocheva et al. 2010b). In concordance with previous reports showing IL-6-mediated transcriptional control of CtsB (Mohamed et al. 2010), our data herein indicate that cathepsin production in TAMs is at least partially regulated by a group of TH2-associated cytokines, which are abundant in the PanNET TME. These results also represent an interesting comparison with the report that LPS-stimulated macrophages robustly upregulate a variety of M1-associated cytokines, including CXCL1 and IL-1β, but not cathepsins (Meissner et al. 2013). Therefore, we propose that a high level of pro-form cathepsin secretion is a characteristic of M2-like macrophages such as TAMs, and extracellular cathepsin-mediated proteolysis coordinates with other TAM-mediated processes, such as immunosuppression and angiogenesis, to promote tumor development.
In conclusion, we have demonstrated that complementary TH2-associated cytokines lead to a distinct macrophage activation state characterized by a robust secretory capacity. These cytokines engage a non-canonical IRE1α axis to promote cathepsin secretion, thus promoting tumor progression and invasion. Given that several cathepsin inhibitors have minimal toxicities (Palermo and Joyce 2008) and deliver therapeutic efficacy in preclinical PanNET models (Joyce et al. 2004; Elie et al. 2010), our findings further emphasize the potential of cathepsin inhibition as a therapeutic strategy in pancreatic neuroendocrine tumors and other cancers.
The generation of RT2 (Hanahan 1985), Il4ra−/− (Noben-Trauth et al. 1997) Stat6−/− (Kaplan et al. 1996), LysM:Cre (Clausen et al. 1999), and Stat3Flox/Flox (Takeda et al. 1998) mice have been reported previously. The Il4ra−/− and Stat6−/− mice were purchased from Jackson Laboratories. LysM:Cre and Stat3Flox/Flox mice were obtained from Dr. Alexander Rudensky. The Il4ra−/− mice, which were originally in the BALB/c background, were backcrossed into the C57BL/6 background for 10 generations. All mouse strains were maintained in the C57BL/6 background. The animal studies were approved by the Institutional Animal Care and Use Committee at Memorial Sloan Kettering Cancer Center (MSKCC). The βTC374 cancer cell line was derived from a WT RT2 tumor in the Joyce lab. The following inhibitors were commercially available: STF-083010 (EMD Millipore), GSK2606414 (Tocris), KIRA6 (Calbiochem), and 4μ8C (Selleckchem). The pancathepsin inhibitor JPM-OEt was synthesized by the Organic Synthesis core at MSKCC.
Tumor burden was determined at 13.5 weeks of age for RT2 mice of all genotypes. For invasion grading, stained pancreatic tissues were graded as previously described (Lopez and Hanahan 2002; Gocheva et al. 2010b). For more details, see the Supplemental Experimental Procedures.
Single cell suspension from RT2 tumors was stained with the antibodies summarized in Supplemental Table 2. An LSR II flow cytometer was used for data acquisition, and data were analyzed using FlowJo software. For cell sorting, samples were sorted on a FACSAria II or MoFlo cell sorter. For more details, see the Supplemental Experimental Procedures.
RNA was isolated using the TRIzol/chloroform method (Invitrogen), and first-strand cDNA was synthesized using the High-Capacity cDNA Reverse Transcription Kit (Invitrogen). For more details, see the Supplemental Experimental Procedures.
Bone marrow (BM) was harvested from WT, Stat6−/−, and LysM:Cre; Stat3Flox/Flox mice. BM-derived cells were cultured for 7 days to generate mature macrophages. For more details on cell culture and cell-based assays, see the Supplemental Experimental Procedures.
Biotinylated DCG-04 (Greenbaum et al. 2002) was synthesized by the Organic Synthesis core facility at MSKCC. Labeling was performed at room temperature. For western blotting, samples were resolved in NuPAGE® SDS-PAGE gels (Invitrogen), transferred to PVDF membranes, and detected using chemiluminescence (Thermo Fisher Scientific). For IRE1α activation analysis, Phos-tag SDS-PAGE (Wako) was performed. For more details, see the Supplemental Experimental Procedures.
BMDMs were transfected with pooled siRNA against Ern1, the gene encoding IRE1α, (Dharmacon) or scrambled siRNA (Invitrogen), using Viromer BLUE (Lipocalyx). For more details, see the Supplemental Experimental Procedures.
BMDMs were generated as described above. On day 7, BMDMs were treated with control media, IL-4, IL-6, IL-10, IL-4+IL-6, or IL-4+IL-10 for 24 hours (all cytokines were used at 10 ng/ml; CSF-1 was supplemented at 10 ng/ml). Cells were harvested and lysed in TRIzol. RNA isolation, library preparation, and preprocessing were completed by the MSKCC Genomics Core Facility using Affymetrix Mouse430A 2.0 microarrays. All downstream bioinformatic analyses were completed in R 3.0.1 using the Bioconductor suite of packages Differentially expressed genes and were identified using the ‘limma’ package (Ritchie et al. 2015) with a fold change cutoff of +/− 2, and a false discovery rate of 10%. These cutoffs were used to identify synergistic and antagonistic genes in the interaction based linear model described below:
Linear model coefficients plotted in Fig. 5A and Supplemental Fig. S4C were normalized to baseline for each gene. Gene ontology analysis was performed with iPAGE and motif activity response analysis (MARA) was performed at http://ismara.unibas.ch/fcgi/mara, further details can be found in the Supplemental Experimental Procedures. Raw microarray data has been deposited to the GEO under accession number GSE70626.
Data are presented throughout as mean ± s.e.m, analyzed by the indicated tests and a significance cutoff of P < 0.05. All statistical analyses were completed in GraphPad Prism 6.0 and R 3.0.1. All tumor volume box and whisker plots are drawn as Tukey boxplots, with upper whiskers extending to the 75% value multiplied by 1.5 × the inter-quartile range and lower whiskers extending to the 25% value multiplied by 1.5 × the inter-quartile range (GraphPad Prism default Tukey options). All plotted data points were included in subsequent statistical analyses as described in the figure legends. All code used to analyze the data can be found at the following website (https://bitbucket.org/bowmanr/joycelab-macrophage). For more details, see the Supplemental Experimental Procedures.
Yan et al. report that the TH2 cytokine IL-4 synergizes with IL-6 and IL-10 in macrophages to promote pancreatic neuroendocrine tumor growth and invasion. The authors show that such synergy depends on STAT3 and STAT6 interaction to activate IRE1α leading to a pronounced secretion of cathepsin proteases and induction of unfolded protein response-related genes.
We thank Xiaoping Chen and Kenishana Simpson for excellent technical support, and Claire Hamilton for assistance with the Stat6 null mouse experiments. We thank members of the Joyce lab for insightful comments and discussion. We are grateful for help from Elyn Reidel, MSKCC Biostatistics Department for assistance with statistical analyses of tumor invasion. We thank Dr. Alexander Rudensky for generously providing the Stat3Flox/Flox and LysM:Cre animals. This research was supported by the following: American Cancer Society (121839-RSG-12-076-01-LIB) (J.A.J.), a National Cancer Institute Cancer Center Support Grant awarded to MSKCC (P30 CA008748), a National Cancer Institute fellowship 5F31CA167863 (R.L.B.), and the graduate training programs of Weill Cornell Medical School (H.W.W.) and the Gerstner Sloan-Kettering Graduate School (R.L.B.).
Author Contributions: D.Y., H.W.W, R.L.B., and J.A.J designed experiments and analyzed data. D.Y., H.W.W., and R.L.B. performed experiments. R.L.B. performed computational analyses. J.A.J. conceived and supervised the study. D.Y., R.L.B., and J.A.J. wrote the manuscript.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.