VHL Inactivation in Precancerous Kidney Cells Induces an Inflammatory Response via ER Stress–Activated IRE1α Signaling Free

Genomic mutations and epigenetic inactivation of the VHL tumor suppressor gene underlie the development of a majority of clear-cell renal cell carcinoma (ccRCC; refs. 1, 2). VHL encodes the substrate-recognition component of an E3 ubiquitin ligase. Its most prominent target is the alpha subunit of the hypoxia-inducible factor (HIF-α). HIF is a master transcription factor stabilized in low oxygen tension, which induces a metabolic switch from oxidative phosphorylation to glycolysis (3, 4). HIF also promotes angiogenesis through expression of cytokines such as VEGF (5). In VHL-mutant ccRCC cells, HIF function is activated even in normal oxygen conditions (6, 7), exhibiting heightened pseudohypoxic responses. As such, ccRCC is an excellent model for studying tumor metabolism and the interaction between tumor cells and the microenvironment (8).

The association of tumor with inflammation has been proposed in mid-19th century by Rudolf Virchow, and the concept was reintroduced by H.F. Dvorak more than a hundred years later (9). Since then, accumulated evidence has indeed implicated inflammation in the development of various cancers (10, 11). Recently, the development of ccRCC has also been linked to tissue inflammation (12, 13), and kidney disease has been identified as an important risk factor for renal cancer (14, 15). Inflammation can promote tumor initiation and progression in various ways. It can stimulate cell proliferation (10, 16) and importantly, induce genome instability (11, 17, 18). Consistent with these observations, our recent study shows that inactivation of the mouse VHL tumor suppressor gene (designated Vhlh) in a subpopulation of kidney distal tubule cells, using the HOXB7-Cre-GFP driver, resulted in Hif-1α–dependent hyperplastic clear-cell lesions and severe inflammation and fibrosis (19). This model thus provides a good system for studying the relationship between inflammation and tumorigenesis.

In this study, we ask the critical question of how VHL inactivation in a subpopulation of renal epithelial cells can elicit systemic responses such as inflammation. It has been shown that in VHL-mutant cells, the load of translation is significantly increased (20), probably as a result of PI3K signal transduction and mTOR pathway activation (21). HIF activation in VHL-defective cells also generates excessive reactive oxygen species (ROS; ref. 22). This is partly because HIF indirectly inhibits conversion of pyruvate to acetyl-CoA that is the carbon source feeding into the tricarboxylic acid (TCA) cycle (4, 23). Therefore, the TCA cycle is slowed while the oxygen supply is normal in these VHL-mutant cells. Slowed TCA cycle causes electron relay to stall; the electrons are then prone to be captured by the oxygen, resulting in the generation of reactive oxygen species (ROS; refs. 24, 25).

We hypothesize that such metabolic abnormalities can disrupt normal proteostasis in at least two ways: protein overload in the endoplasmic reticulum (ER) compounded by protein disulfide bond malformation resulting from excessive ROS. These conditions have been shown to cause ER stress (26, 27). ER stress activates the unfolded protein response (UPR) that enables the cell to either reduce the load of unfolded proteins or, if unresolved, initiate apoptosis (28). The cellular response to ER stress is complex, but mainly involves three pathways that lead to production of three active transcription factors: ATF6, ATF4 via PERK activation, and XBP1 via IRE1 activation.

Among these pathways, IRE1 is the most ancient and has been linked to inflammatory response (29, 30). IRE1 is an ER transmembrane protein (31). Two isoforms of IRE1 exist: IRE1α is ubiquitously expressed while the β isoform is mainly expressed in the gastrointestinal epithelia (32). It has recently been shown that one of the sensors that detects mitochondrial ROS and translates the signal to induce ER stress is IRE1α, which is frequently found at the junction between mitochondria and the ER lumen (27). IRE1α contains the dual activity of kinase and endonuclease (RNase; refs. 33, 34). Canonically, ER stress activates IRE1α via autophosphorylation. Activated IRE1α then induces specific alternative splicing (via its RNase activity) that generates a functional “spliced” XBP1 transcription factor (XBP1s; for clarity, we will use XBP1 to denote activated XBP1 in this article). The IRE1α kinase activity is required for the oligomerization that precedes the activation of the RNase activity (34–36). In a separate function, autophosphorylation of IRE1α also results in binding to the adaptor protein TNFα receptor-associated factor 2 (TRAF2) and activation of c-Jun N-terminal kinase (JNK) via ASK1 (37). The IRE1α−TRAF2 complex can also recruit and activate IκB kinase (IKK), which phosphorylates IκB, leading to the degradation of IκB and nuclear translocation of NFκB (38). NFκB and the JNK-activated transcription factor AP1 are two major inducers of the expression of inflammatory genes (39, 40). Therefore, it is possible that the metabolic abnormalities in VHL loss-of-function cells trigger cellular inflammatory pathways via ER stress and IRE1α signaling. In this study, we set out to test this hypothesis.

All procedures were conducted in accordance with the US Public Health Service Policy on Humane Care and Use of Laboratory Animals. Mice were maintained in National Central University, Taiwan, animal housing facility according to protocols approved by the Institutional Animal Care and Use Committee. Mouse strains used were in C57BL/6 background as described previously (19). The Vhlh conditional knockouts were generated by intercross of Hoxb7-Cre-GFP/+; Vhlh^fl/+. Littermates with the genotype Hoxb7-Cre-GFP/+ served as wild-type control. Hoxb7-Cre-GFP is a Hoxb7 promoter driven Cre line with a GFP reporter. All triplicate experiments were performed using three different animals.

HK-2, a human renal proximal tubular epithelial cell line, and HEK293, a human cell line of kidney origin, were purchased from Bioresource Collection and Research Center, Taiwan, in 2014 with identity certification. Each batch of cells were cultured for <10 passages in DMEM/Ham F12 (Gibco-Thermo Fisher) supplemented with 10% FBS, and penicillin–streptomycin (50 U/mL; Sigma-Aldrich) at 37°C in 5% CO₂/95% air. Culture medium was replaced on alternate days. Once the cells reached 60%–70% confluency, they were collected or passaged (via trypsinization) for the subsequent experiments. When appropriate, cells were treated with solvent [DMSO:PBS = 1:3, PBS, or DMSO], tunicamycin (at 1 μg/mL in DMSO/PBS; Sigma-Aldrich), APY29 (at 0.2 μmol/L in DMSO; Tocris Bioscience), or 4-phenylbutyric acid (4-PBA at 1 mmol/L in PBS; Sigma-Aldrich) for 24 hours.

The murine monocyte/macrophage cell line RAW264.7 was kindly provided by Dr. N.H. Ma (Institute of Systems Biology and Bioinformatics, National Central University, Jhongli, Taiwan) and was originally obtained from ATCC in 2012. The cells were cultured in DMEM (Gibco-Thermo Fisher) supplemented with 10% FBS, 4 mmol/L l-glutamine, 4,500 mg/L glucose, 1 mmol/L sodium pyruvate, 1,500 mg/L sodium bicarbonate, and penicillin–streptomycin (50 U/mL; Sigma-Aldrich) at 37°C in 5% CO₂/95% air, for no more than 5 passages. Culture medium was replaced on alternate days. Once the cells reached 50%–60% confluence, they were collected or passaged (via trypsinization) for the subsequent experiments.

Potential contamination by mycoplasma was monitored by PCR for conserved 16S rRNA coding sequence in the newly plated cells, and follow-up routine examination was done by DAPI staining.

Antibodies for Western blot analyses and the final working dilutions were as follows: mouse mAbs against VHL (1:1,000; BD Biosciences), BiP (1:1,000; BioLegend), p-JNK (1:1,000; Santa Cruz Biotechnology), FLAG (1:1,000; Cell Signaling Technology), and β-actin (1:500, Santa Cruz Biotechnology); rabbit polyclonal antibodies against IRE1α (1:1,000; Cell Signaling Technology), p-IRE1α (1:1,000; Abcam), PERK (1:500; Santa Cruz Biotechnology), p-PERK (1:500; Santa Cruz Biotechnology), eIF2α (1:1,000; Cell Signaling Technology), p-eIF2α (1:1,000; Cell Signaling Technology), ATF4 (1:500; Santa Cruz Biotechnology), p65 (1:1,000; GeneTex), fibrillarin (1:1,000, Santa Cruz Biotechnology), TNF-α (1:1,000, GeneTex), and IL1β (1:1,000, GeneTex); and goat polyclonal antibody against albumin (1:250; Santa Cruz Biotechnology). All horseradish peroxidase–conjugated secondary antibodies were purchased from Jackson ImmunoResearch Inc. (Jackson ImmunoResearch).

The Protein Synthesis Assay Kit (Cayman Chemical) was used according to manufacturer's protocol. Briefly, cells were cultured with or without the protein synthesis inhibitor cycloheximide (at a final concentration of 50 μg/mL) at 37°C for 30 minutes. After incubation, cells were trypsinized and collected by centrifugation at 400 × g for 5 minutes. Cells were resuspended in 0.5 mL of 20 mmol/L O-propargyl-puromycin (OPP) and incubated at 37°C for 2 hours. OPP is an alkyne analogue of puromycin that is incorporated into the nascent peptide chain and causes premature chain termination (41). After incubation, cells were collected by centrifugation as above. Cells were resuspended in 0.5 mL cell-based assay fixative solution and incubated at room temperature for 5 minutes. Cells were then collected, resuspended in 1 mL cell-based assay wash buffer, and incubated at room temperature for 5 minutes. Cells were collected by centrifugation as above and resuspended in 1 mL FAM-Azide [conjugated with the fluorescein isothiocyanate (FITC) dye] staining solution and incubated in the dark at room temperature for 30 minutes. After incubation, cells were collected as above and resuspended in 1 mL cell-based assay wash buffer and incubated at room temperature for 5 minutes. Cells were collected and resuspended in 1 mL assay buffer and FITC-stained cells quantified by NovoCyte Flow Cytometer (ACEA Biosciences).

Cells were washed with PBS and incubated with 10 μmol/L dihydroethidium (DHE; Santa Cruz Biotechnology) in Hank's Balanced Salt Solution (HBSS) at 37°C for 30 minutes in the dark. In the presence of ROS, DHE is oxidized and becomes fluorescent (42). After incubation, cells were trypsinized and washed with ice-cold PBS three times. The cellular ROS level was quantified by NovoCyte Flow Cytometer (ACEA Biosciences).

Histologic staining of paraffin sections was conducted using previously described methods (19). Collagen deposition was visualized using Picro-Sirius Red Stain Kit (Abcam) following the manufacturer's instruction. The slides were cleared and mounted using Cytoseal Mounting Media-Cytoseal 60 (Thermo Fisher). The sections were evaluated under light microscope Olympus IX81 and the images were analyzed with ImageJ software (NIH, Bethesda, MD).

For VHL knockdown, cells were transfected with plasmids expressing short hairpin RNAs (shRNA) as described (43, 44). The target sequences shVHL1, gagaactgggacgaggccg, and shVHL3, gagcctagtcaagcctgag, were used in this study.

For IRE1α knockdown, two independent siRNA was used: (i) siIRE1α1 is a mixture of four duplex RNAs (from Santa Cruz Biotechnology), consisting of the following target sequences: caaccucucuucuguaucu, ggaaggugaugcacaucaa, cuggaggagacgaaugaua, cuguacucuuggaguaaca; and (ii) siIRE1α2 targets the sequence cugcuuaaugucagucuac, as described previously (45).

Mouse primary renal tubule cells were isolated using the previously described methods (46, 47).

HK-2 cells were transfected with scrambled siRNA or siRNA specific for IRE1α, or vectors expressing scrambled shRNA, shRNA specific for VHL (shVHL), FLAG-IRE1α^wt, or FLAG-IRE1α^K599A (provided by Dr. K. Mori in Kyoto University, Japan; ref. 27) using the BTX Gemini X2 Electroporation System (BTX Harvard Apparatus). Briefly, 5 × 10⁵ cells were electroporated with 10 pmol of siRNA or 4 μg of plasmid in 1 mL serum-free media with one 100 V pulse for 10 ms and plated in 6-well plates.

The cells were collected, resuspended in hypotonic buffer (10 mmol/L HEPES, pH 7.9; 10 mmol/L KCl; 1.5 mmol/L MgCl₂; 0.2 mmol/L PMSF; 20 μg/mL aprotinin; 0.5 mmol/L DTT; and 0.5% NP-40), and incubated on ice for 15 minutes. After centrifugation at 6,000 × g for 15 minutes at 4°C, the supernatant was collected (the cytosolic fraction). The pellet was then washed once with basal buffer (hypotonic buffer without 0.5% NP-40). After centrifugation again at 6,000 × g for 15 minutes at 4°C, the pellet was collected and resuspended in hypertonic buffer (20 mmol/L HEPES, pH 7.9; 400 mmol/L KCl; 1.5 mmol/L MgCl₂; 0.2 mmol/L PMSF; 20 μg/mL aprotinin; 0.5 mmol/L DTT; 0.2 mmol/L Na-EDTA; 10% glycerol) and incubated at room temperature for 30 minutes. After centrifugation at 10,000 × g for 30 minutes at 4°C, the nuclear fraction contained in the supernatant was collected.

Transwell plates (24-well; 8-μm pore size, Corning) were used to measure chemotaxis of the RAW264.7 cells. HK-2 cells were transfected with siRNA or vector-expressing shRNA by electroporation (see above) and seeded into the bottom well. After incubation for 48 hours, cells were treated with solvent (DMSO:PBS = 1:3; or DMSO), tunicamycin, APY29, or 4-PBA as described above, for 24 hours in serum-free media supplemented with 5% BSA (BIONOVAS). RAW264.7 cells (1 × 10⁵) were seeded in the top well and then placed over the bottom well containing the treated or untreated HK-2 cells. In a control experiment, RAW264.7 cells in the top well were pretreated with tunicamycin (at 1 μg/mL) before being placed over HK-2 cells, to assess whether tunicamycin can directly influence the migratory capability of the macrophages. In a separate experiment, the HK-2 cells were pretreated with chemical chaperon 4-PBA at a final concentration of 1 mmol/L for 24 hours. The media containing 4-PBA were replaced with the serum-free media described above, and the top well containing RAW264.7 cells inserted for the chemotaxis assay.

The chemotaxis was allowed to proceed for 24 hours, after which the remaining cells on the upperside of the membrane was wiped off with a cotton swap and the migrated cells on the underside of the membrane were stained with 0.1% crystal violet for 5 minutes and washed with H₂O. The excess crystal violet was destained with methanol for 15 minutes. The stained membrane was imaged with the EPSON V750 PRO scanner. Quantification was performed by colorimetric measurement at OD₅₇₀ using Synergy HT (BioTek).

Data are expressed as means ± SEM. Groups were compared via a one-way or two-way ANOVA followed by Bonferroni post hoc analysis. P < 0.05 was considered statistically significant. The sample size is designed for 80% power and statistical significance set at P < 0.05. On the basis of previous experiences, the normal mean and deviation should give an effective size of 3. We therefore estimate the sample size for our two-way ANOVA to be 3.

In our previous study, inactivation of the mouse VHL gene (designated as Vhlh) in a subpopulation of kidney distal tubule cells, using the HOXB7-Cre-GFP driver, resulted in hyperplastic clear-cell lesions, inflammation (i.e., infiltration of immune cells) and severe fibrosis (Supplementary Fig. S1; ref. 19). As mentioned above, a potential link between VHL mutation and inflammatory response is the metabolic abnormalities in the VHL mutant. We first examined whether metabolic abnormalities exist in noncancerous kidney epithelial cells upon VHL knockdown. As we are primarily interested in cellular conditions before the onset of cancer, in this study, commonly used VHL-mutant cancer cell lines will not be appropriate. We therefore first examined noncancerous human kidney cell lines HK-2 or HEK293 (containing wild-type VHL) and their VHL knockdown counterparts.

As shown in Fig. 1 and Supplementary Fig. S2A and S2B, in human kidney cells, VHL knockdown using either of the shVHL constructs resulted in increased protein synthesis and elevated levels of ROS, consistent with the previous notion (20, 25). This was confirmed using a different kidney cell line HEK293 (Supplementary Fig. S2C and S2D). We therefore hypothesize that the metabolic abnormalities may lead to ER stress and UPR, which in turn can induce inflammatory response.

We next examined ER stress and UPR at the cellular level. In HK-2 cells, VHL knockdown induced ER stress and UPR, as indicated by the increased levels of BiP, phosphorylated IRE1α (p-IRE1α), and IRE1α target XBP1 (Fig. 2A; Supplementary Fig. S3A–S3C). The level of ER stress induction via VHL knockdown is similar to that induced by protein glycosylation inhibitor (and thus ER stress inducer) tunicamycin (Fig. 2A; Supplementary Fig. S3A–S3C). The other UPR pathway, the PERK–eIF2α–ATF4 axis, was also induced by VHL knockdown and tunicamycin treatment (Fig. 2A; Supplementary Fig. S3D–S3F), although the ATF4 induction is not as prominent as BiP, p-IRE1α, and XBP1. This indicates that the eIF2α pathway may be under more stringent feedback control such as that mediated by phosphatase GADD34 (48). In step with IRE1α activation, JNK was phosphorylated (activated) without changes in the total JNK protein levels (Fig. 2A; Supplementary Fig. S3G). Also, the level of NFκB activation was significantly increased in VHL knockdown cells, as shown by the increase of NFκB subunit p65 nuclear localization with concomitant decrease in cytosolic localization (Fig. 2B; Supplementary Fig. S4A and S4B), as well as by the increase in the expression of NFκB target gene products TNFα and IL1β (Fig. 2C; Supplementary Fig. S4C and S4D). The effects of VHL knockdown and tunicamycin treatment are additive in the nuclear localization of NFκB, but the additive effects are modest in ER stress response and in the activation of JNK, suggesting a feedback control in these UPR signaling pathways, but the exact mechanism in this context is unknown. Note that these effects were sustained as long as VHL is reduced or knocked out. We have observed effective knockdown as long as 7 days after transfection (data not shown), but the cells were usually analyzed within 3 days after transfection.

We sought to verify the above results by examining primary kidney tubule cells isolated from wild-type and Vhlh conditional knockout mice (Hoxb7-Cre-GFP-driven; ref. 46). The primary cells were confirmed for their respective VHL protein expression level (Fig. 2D, top). The increased ER stress response was observed in the increased expression of BiP, p-IRE1α, XBP1, p-PERK, p-eIF2α, and ATF4 in Vhlh-mutant cells and after treatment with tunicamycin (Fig. 2D; Supplementary Fig. S5A–S5F). JNK and NFκB were also activated in step with IRE1α activation (Fig. 2D and E; Supplementary Fig. S5G–S5I).

As mentioned earlier, IRE1α is a critical link between ER stress and inflammatory response, possibly via its noncanonical function of JNK and NFκB activation (see Introduction). To test whether VHL inactivation (knockdown or knockout) indeed induces inflammatory response via IRE1α, we used a kinase inhibitor APY29 that showed preference for IRE1α (49). APY29 is a type 1 inhibitor that occupies the IRE1α kinase–active site (50). While it blocks the access of kinase substrates, it also keeps the active site in an open conformation. Because the RNase activity of IRE1α depends on the open conformation induced by autophosphorylation, APY29 can block IRE1α kinase activity yet allosterically activates the IRE1α RNase domain (49). Thus, APY29 will not inhibit, or may actually enhance, the canonical UPR. Specifically inhibiting the IRE1α kinase activity without blocking UPR is advantageous for our assays as the strategy will avoid complicating side effects from further increasing ER stress.

As shown in Fig. 3, treatment with APY29 decreased the level of p-IRE1α while increasing or maintaining the levels of XBP1 in VHL wild-type and knockdown HK-2 cells (Fig. 3A; Supplementary Fig. S6A and S6B). The BiP levels in APY29-treated VHL knockdown cells were reduced compared with untreated VHL knockdown cells (Fig. 3A; Supplementary Fig. S6C), possibly the result of increased XBP1 levels. More importantly, APY29 also significantly decreased the levels of p-JNK in VHL knockdown cells to near the wild-type level without significantly affecting the total IRE1α and JNK protein levels (Fig. 3A; Supplementary Fig. S6D). The same response was reproduced in the mouse primary renal tubule cells comparing wild-type and Vhlh mutant (Fig. 3B; Supplementary Fig. S5J–S5M). Nuclear localization of NFκB and TNFα and IL1β expression in VHL knockdown HK-2 cells were also reduced to the background level upon APY29 treatment (Fig. 3C and D; Supplementary Fig. S6E–S6H).

One key event in tissue inflammation is the infiltration of macrophages, and we have previously observed increased macrophage infiltration into the Vhlh conditional knockout kidney (19). We therefore devised an in vitro system to test whether VHL loss of function in epithelial cells possesses increased capability to induce macrophage chemotaxis. In this assay, HK-2 cells (with or without VHL knockdown) were plated in the bottom well of a Boyden chamber, whereas the mouse monocyte/macrophage cell line RAW264.7 was plated on the porous membrane above the HK-2 cells. The number of RAW264.7 cells migrated toward the HK-2 cells to the underside of the membrane could then be quantified as an index of the chemotactic response. The efficacy of VHL knockdown in HK-2 cells (comparing triplicates each of the scrambled and the two shVHL constructs) was monitored by Western blot analysis for VHL and HIF-1α protein expression (Supplementary Fig. S7A and S7B). In the transwell assay, knockdown of VHL induced a 3- to 6-fold increase in macrophage chemotaxis compared with wild-type cells (Fig. 4A). Treatment of HK-2 cells with tunicamycin increased the macrophage chemotaxis toward wild-type cells to a level comparable with the VHL knockdown cells (3- to 4-fold), and further increased the chemotaxis toward VHL knockdown cells by about 2-fold. Tunicamycin did not directly stimulate macrophage migration as treatment of RAW264.7 cells with tunicamycin did not increase chemotaxis (Supplementary Fig. S7C). Taken together, the results suggest that ER stress in VHL loss-of-function cells can induce macrophage chemotaxis, which is further increased by exacerbating the stress level with tunicamycin.

To further ascertain that the increased chemotaxis of macrophage RAW264.7 cells toward VHL loss-of-function cells is mediated by the IRE1α signaling pathway, we performed the in vitro chemotaxis experiment in the presence of APY29. In this experiment, the primary renal tubule cells were included, in addition to the HK-2 cells. In both cell systems, loss of VHL function significantly increased the chemotactic migration of the RAW264.7 cells (Fig. 4B and C). In the presence of APY29, RAW264.7 cell migration toward VHL loss-of-function HK-2 (Fig. 4B) or primary (Fig. 4C) kidney tubule cells was reduced to the wild-type levels. The specificity of APY29 and the involvement of IRE1α were further verified by knocking down IRE1α in the VHL knockdown background. VHL-IRE1α double knockdown ameliorated the capacity of VHL knockdown HK-2 cells to recruit macrophages (Fig. 4D).

To verify whether ER stress is necessary for inducing macrophage chemotaxis, HK-2 cells were pretreated with a chemical chaperon and ER stress inhibitor 4-PBA that improves protein folding (51). We confirmed that 4-PBA treatment reduced ER stress marker expression including BiP, XBP1, p-IRE1α, and p-PERK, as well as p-JNK (data not shown). Importantly, 4-PBA could also inhibit the chemotaxis of RAW264.7 cells toward VHL knockdown cells (Fig. 4E). Therefore, we conclude that the enhanced ability of VHL loss-of-function cells to attract macrophage is the result of ER stress signaling and is dependent on the IRE1α kinase activity.

We have shown so far that the IRE1α kinase inhibitor APY29 can effectively block the inflammatory response in the VHL knockdown or mutant kidney cells. To further confirm the direct involvement of IRE1α kinase activity, we used an IRE1α kinase–active site mutation K599A (27, 52). As shown in Fig. 5 and Supplementary Fig. S8, the increase of p-JNK level in VHL knockdown HK-2 cells (compare lanes 2 with 1) was ameliorated by the knockdown of IRE1α (compare lanes 3 with 2). Interestingly, the reduction of JNK activation in IRE1α knockdown cells could be restored by expressing a wild-type IRE1α exogenously, but not by reexpressing the kinase-active site mutant K599A (compare lane 3 with lanes 4 with 5). The expression of the exogenous wild-type or mutant IRE1α was monitored by the fusion marker peptide FLAG (top). As controls, without VHL knockdown (lanes 6–10), JNK activation (p-JNK) was very low regardless of the IRE1α status. Note that in these conditions, IRE1α showed low levels of activation (p-IRE1α). This level of activation is likely a constitutive background activity of IRE1α that is under a separate control, as it is not affected by exogenously expressed IRE1α (wild-type or kinase-site mutant). Note that this background level is very low when the extent of IRE1α knockdown is enhanced using a different siRNA construct (Supplementary Fig. S8). We conclude that JNK activation in VHL loss-of-function kidney cells is dependent on the IRE1α kinase function.

The VHL-mutant kidney phenotypes including fibrosis and proteinaceous blockage of tubules are developed at approximately 3–4 months of age in HOXB7-Cre-GFP; Vhlh^fl/fl mice (Supplementary Fig. S1; ref. 19). To verify whether these phenotypes are dependent on the IRE1α kinase activity, we injected 1.3 μg of APY29 (in 50 μL canola oil) intraperitoneally into 2.5-month-old HOXB7-Cre; Vhlh^fl/fl mice (canola oil alone as control), three times a week for 6 weeks. The injection schedule and dosage were determined empirically, taking into account the sublethal effective dosage needed and the age of mouse when the phenotypes are fully developed. As shown in Fig. 6A, mutant kidneys showed extensive fibrosis compared with wild-type (top). With APY29 treatment, the pathologic fibrosis was undetectable (bottom). Quantification of the Sirius red staining showed that APY29 could reduce the fibrosis to the wild-type level (right). It is interesting to point out that, because our injection regimen begins at 2.5 months of age, when the phenotypes are barely detectable (19), the effect of APY29 is likely a preventive one. Concomitant to the reduction of tissue fibrosis, the loss of kidney function observed in knockout mice, as measured by the increased albumin content in equal volume urine (Fig. 6B, compare lanes 1 with 2), was also rescued by APY29 treatment (lanes 2, 3, 4).

Inflammation has been implicated in the initiation of various cancers (10, 11). This is an important finding because if the link can be verified, such cancers are potentially preventable by anti-inflammatory therapeutics. This is particularly relevant for high-risk patients such as family members predisposed to hereditary cancers, and patients of inflammatory diseases such as hepatitis, colitis, and chronic kidney disease. We have previously demonstrated a link between inflammation and development of ccRCC in a knockout mouse model. Genetic and epigenetic inactivation of the VHL tumor suppressor gene has been shown to be the driver in up to 70%–80% of the ccRCC cases (2). We therefore generated conditional knockout of Vhlh in a subset of kidney tubule cells using the HOXB7-Cre-GFP driver, which resulted in hyperplastic lesions containing abnormal clear cells typical of the ccRCC (19). Surprisingly, these hyperplastic lesions were accompanied by severe inflammation and fibrosis. It thus provides support for the notion that inflammatory kidney disease is a risk factor for RCC development (12–15).

One critical question to address is how a single tumor suppressor gene mutation in a subpopulation of tubule epithelial cells can induce a systemic inflammatory response, as the answer to this question may provide an avenue for cancer prevention. Two of the critical inflammation-inducing master switches, NFκB and JNK, have been known to be activated in cells under ER stress via the IRE1α pathway (37, 38). In VHL loss-of-function cells, because of the imbalance in proteostasis induced by metabolic abnormalities (Fig. 1), chronic ER stress is a likely pathophysiologic condition. If the above reasoning is valid, such condition could lead to activation of an inflammatory response beyond the epithelia that exhibit reduced VHL function. This study provides the proof for such a model.

In VHL knockdown or mutant cells, ER stress–responsive pathway emanating from IRE1α is activated, which is accompanied by increased levels of p-JNK and activation of NFκB (Fig. 2). Significantly, these responses can be blocked by an IRE1α kinase inhibitor APY29 (Fig. 3). Functionally, ER stress and VHL loss of function exert additive chemotactic attraction for macrophages, which can also be ameliorated by IRE1α knockdown and treatment with APY29 (Fig. 4). The involvement of IRE1α kinase activity was further confirmed using a kinase-active site mutant of IRE1α, which could not rescue the loss of JNK activation after the endogenous IRE1α was knocked down (Fig. 5). In vivo, APY29 can prevent the fibrotic phenotype and improve the kidney function of the knockout mice (Fig. 6).

We also observed certain degrees of additive effects of VHL inactivation and tunicamycin treatment, which is more prominent in NFκB activation and macrophage recruitment but less so in ER stress marker expression. There are three possibilities for the additive effect: (i) the extent of ER stress has different gradations, ranging from maintenance of proteostasis to induction of apoptosis (28, 29). It is possible that VHL inactivation does not induce ER stress to a maximum extent; therefore addition of tunicamycin can exacerbate the stress. (ii) VHL inactivation can elicit inflammatory response via an ER stress-independent pathway. Indeed VHL loss of function has been shown to result in NFκB and JNK activation via stabilizing a scaffold complex containing Card9 (53, 54), although these were shown in malignant ccRCC cell lines, not in premalignant kidney cells (3). The IRE1α pathway can be activated by an ER stress–independent mechanism, such as membrane lipid saturation (55). We favor the first possibility as our result indicates that knocking down IRE1α or reducing ER stress could completely ameliorate the inflammatory response after VHL inactivation.

APY29 has been known to target the IRE1α-TRAF2 scaffold activity without inhibiting XBP1 activation, thus reducing the UPR (49, 50). As such, APY29 was chosen as an IRE1α kinase inhibitor because we could avoid the complex side effects associated with further increasing ER stress. It is also important to note that in our in vivo experiment, the mice treated with APY29 did not show drug-related lethality during and after the treatment, indicating that the compound may have relatively low cytotoxicity at our chosen dosage. We have previously shown that ruxolitinib, the inhibitor of JAK1/2 that is a main mediator of the inflammatory cytokine signaling pathway, could reduce fibrosis and hyperplasia in the HOXB7-Cre-GFP–driven Vhlh knockout kidney (19). Here we demonstrate that rescue of inflammatory and fibrotic phenotypes can be achieved by intervening at further upstream of the pathologic chain of events—that is, before the inflammatory pathway is activated in the epithelial cells.

Two key events in early cancer progression are hypoxic response (the Warburg effect) and tumor-associated inflammation (“tumors are wounds that do not heal”), but the potential link between these two events has not been elucidated. This study provides a mechanistic view of the origin of inflammatory response induced by hypoxia-induced metabolic imbalance in epithelial cells. The finding can suggest a strategy for early cancer detection and prevention in high-risk cohorts. One high-risk group that may benefit from the knowledge is familial VHL disease patients. Up to 60% of the carriers of germline VHL mutations develop ccRCC in their life time (2). These high-risk patients should be considered for preventive monitoring or treatment for inflammatory kidney conditions. It is also worth noting that the IRE1α-TRAF2–mediated inflammatory pathway identified in the VHL loss-of-function cells is induced by the ER stress, but this IRE1α–TRAF2 pathway does not contribute to the modulation of ER stress, as APY29 treatment could block the inflammatory response but did not reduce the UPR. This provides a therapeutic strategy that can avoid altering the status of ER stress and the associated UPR. Therapeutics that aim to reduce ER stress in noncancerous tissues may decrease inflammatory response but may also cause unwanted side effects, as ER stress and UPR are an important protective mechanism in normal cellular physiology. In this sense, the type of IRE1α kinase inhibitors, such as APY29, that specifically block the inflammatory function of IRE1α offers a good avenue for cancer prevention. Future studies should further delineate the exact roles of ER stress and inflammation at different stages of cancer development.

No potential conflicts of interest were disclosed.

Conception and design: C.-Y. Kuo, T. Hsu

Development of methodology: C.-Y. Kuo, T. Hsu

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C.-Y. Kuo, C.-H. Lin

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C.-Y. Kuo, C.-H. Lin, T. Hsu

Writing, review, and/or revision of the manuscript: C.-Y. Kuo, T. Hsu

Study supervision: T. Hsu

We are indebted to Dr. Ann Chen, Chief of Pathology, Triservice General Hospital, Taiwan, for analysis of the mouse kidney sections. We also thank the Taiwan Bio-development Foundation (TBF) for support of the T. Hsu laboratory.

This work was supported by a grant from the NIH (#R01CA109860 to T. Hsu), a grant from the Ministry of Science and Technology, Taiwan (MOST103-2320-B-008-002-MY3 to T. Hsu), a grant from National Health Research Institute, Taiwan (EX105-0501BI to T. Hsu), and a grant from Cathay General Hospital (CGH-MR-A10304 to C.-H. Lin).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.