ZNF395 is a Hypoxia-Responsive Regulator of Mitochondrial Glutaminolysis in Clear Cell Renal Cell Carcinoma

Abstract

Hypoxia signaling induced by VHL deficiency fuels growth but also imposes metabolic stress on clear cell renal cell carcinomas (ccRCCs). Many ccRCC cells depend on glutamine as the primary source of tricarboxylic acid (TCA) anaplerosis. Hypoxia-inducible factor alpha (HIFα) governs glycolysis but does not directly regulate glutamine metabolism; instead, the factor responsible for orchestrating glutamine metabolism and mitochondrial adaptations to hypoxia remains elusive. Here, we showed that ZNF395 is a hypoxia-responsive factor that regulates glutamine metabolism in the mitochondria. When activated by a HIF2α-modulated super-enhancer, ZNF395 facilitated the transcription of enzymes essential for glutaminolysis, including glutaminase (GLS) and isocitrate dehydrogenase 2 (IDH2). Functionally, ZNF395 depletion resulted in reduced TCA cycle intermediates and their derivatives, including amino acids, glutathione, and pyrimidine nucleotides, leading to impaired mitochondrial respiration. Restoration of mitochondrial complex I function and GLS expression partially rescued the effects of ZNF395 depletion on ccRCC tumor growth. Together, this study underscores the coordinated role of HIFα and ZNF395 in shaping metabolic adaptations in response to hypoxia in VHL-deficient ccRCCs.

Introduction

VHL loss is a mutational driver event in the initiation of clear cell renal cell carcinoma (1,2). VHL loss leads to the stabilization of hypoxia-inducible factor (HIF) (3–5), resulting in a state of constitutive hypoxia that confers pro-tumorigenic traits such as increased angiogenesis (6) but inadvertently poses metabolic stress. Most prominently, HIF increases glycolysis but simultaneously suppresses pyruvate oxidation by activating pyruvate dehydrogenase kinases (PDK) (7). As a result, a large proportion of the glycolytic flux is shunted away from the tricarboxylic acid (TCA) cycle, limiting the carbon source that is necessary to synthesize macromolecules, including nucleotides, amino acids, and lipids. Glutamine supplies TCA cycle intermediates through complementary processes called oxidative anaplerosis and reductive carboxylation (8–12). In both pathways, glutamine is converted into glutamate by glutaminase GLS and then to α-ketoglutarate for further metabolism. Therefore, glutaminolysis is a crucial metabolic process that adapts VHL-deficient tumors for survival under hypoxic stress (9,13).

Despite the elucidation of metabolic changes that occur in hypoxia and VHL deficiency, the transcriptional program that orchestrates a metabolic response to hypoxia is not completely understood. HIF2α is a well-known transcription factor that activates glycolytic genes such as hexokinase 2 HK2 and glucose transporter SLC2A1 (14,15). Unexpectedly, HIF2α does not directly regulate key glutaminolytic targets including GLS and IDH2 (14,15). HIF2α can indirectly repress mitochondrial function by suppressing PGC-1α, a transcription factor that coordinates mitochondrial biogenesis (16). Other transcription factors known to regulate glutamine metabolism include MYC and c-Jun. MYC expression can upregulate GLS, LDHA, and ASCT2 (17), while c-Jun can bind directly to the promoter of GLS, leading to the upregulation of GLS (18). However, neither MYC nor c-Jun are known to be upregulated by hypoxia or VHL loss (19–21). A hypoxia-responsive transcription factor that increases glutamine metabolism remains elusive.

Master regulators important for lineage specification and disease identity are often epigenetically regulated by long stretches of enhancers called super-enhancers (22). These regulatory hubs integrate multiple signals and modulate important cellular processes. We have previously found that HIF2α establishes super-enhancers by recruiting the histone acetyltransferase p300 (21). One key super-enhancer region activated by VHL loss/HIF2α stabilization targets a lesser-known transcription factor, ZNF395, first identified as a binding factor to papillomavirus (23). We subsequently identified ZNF395 as being critical for the tumorigenesis of ccRCC (21). We now report that ZNF395 functions as a HIF2α-activated master regulator that is crucial for maintaining glutaminolysis in response to hypoxia induced by VHL deficiency in ccRCC.

Methods

Cell lines

Commercial cell lines, 786-O (RRID:CVCL_1051), A-498 (RRID:CVCL_1056), HK-2 (RRID:CVCL_0302), ACHN (RRID:CVCL_1067), Caki-1 (RRID:CVCL_0234) and HEK293 (RRID:CVCL_0045) were purchased from ATCC. T98G (RRID:CVCL_0556) was a gift from Dr. Christopher Ang Beng Ti from National Neuroscience Institute. Cell lines 786-O, A-498, HK-2 and Caki-1 were maintained in RPMI (Gibco) and cell lines ACHN, HEK293 and T98G were maintained in DMEM (Gibco). Both media were supplemented with 10% FBS (Gibco). Cell line authentication was performed by short tandem repeat (STR) analysis (Cancer Science Institute of Singapore) against publicly available STR profiles. Mycoplasma testing was performed using the MycoSensor PCR assay kit (Stratagene).

Animal studies

All animal studies were conducted in compliance with animal protocols approved by the Institutional Animal Care and Use Committee (IACUC) of SingHealth (Ref No: 2023/SHS/1785) and University of Texas Southwestern Medical Center (Protocol # 2019–102794). To measure metabolic changes, male NSG mice (6–8 week old) were randomly assigned into experimental groups and implanted with A-498 or 786-O cells transduced with shRNA against non-targeting control or ZNF395 subcutaneously in the flank. Tumor volume was calculated as (length × width × width) × 0.5. Tumors were harvested for metabolic profiling when the tumor volume exceeded 1000 mm³. For the GLS rescue experiment, 1 × 10⁶ A-498 cells from the indicated groups were implanted subcutaneously into male NOD/SCID mice (6–8 weeks of age). Tumor growth was monitored weekly, starting three weeks post-implantation.

siRNA knockdown

ON-TARGETplus SMARTpool siRNA (Dharmacon, UK) were used with Non-Targeting Control Pool as negative control. Refer to Supplementary Information for oligo sequences used in siRNA of ZNF395. A-498 and 786-O cells were trypsinized and diluted to appropriate concentrations. SMARTpool siRNAs were diluted in Opti-MEM to a final siRNA concentration of 50 nM. The diluted Lipofectamine 3000 was added to the diluted siRNA and incubated for 15 min at room temperature to allow complex formation to occur. The siRNA mixtures were aliquoted to wells in a 6-well plate. Forty-eight hours after transfection, cells were re-seeded into 6-well plates for colony formation assays or 96-well plates for cell viability assay and total RNA was harvested for gene expression changes.

Knockdown of ZNF395 and MYC in ccRCC cell lines

Stable knockdown of ZNF395 in ccRCC cell lines was performed using retroviral transduction with either commercially available MISSION shRNA constructs (Sigma Aldrich) or self-assembled shRNA constructs.

MISSION shRNA glycerol stock was cultured and plasmid was extracted using NucleoBond^® Xtra Maxi Plus EF (MACHEREY-NAGEL). Viral particles were packaged by transfecting plasmid into HEK293T at 2 μg DNA/well of a 6-well plate using Lipofectamine 3000 (Life Technologies). A media change was performed 6 hours after transfection. The cell supernatant containing retroviral particles was harvested 48 hours post-transfection, and inoculated onto ccRCC cells, which were then selected with puromycin (2 mg/mL) for 3 days. Cells were analyzed for gene and protein expression and other functional assay post-selection.

The self-assembled constructs were made by ligating phosphorylated and annealed sense and antisense oligos into Bpil digested Tet-pLKO-puro plasmid (Addgene plasmid #21915, RRID:Addgene_21915). Packaging of viral particles and transduction of cells was as described above. These knockdown cells required additional doxycycline treatment (0.10 μg/mL) for 3 days to induce shRNA expression before subsequent analysis. Refer to Supplementary Information for the shRNA oligos used for ZNF395 and MYC knockdown.

Overexpression of NDI1 and MYC in ccRCC cell lines

Platinum-A (PlatA) retroviral packaging cells were seeded at a density of 8 × 10^5 cells in 2 mL of Dulbecco’s Modified Eagle Medium (DMEM) in a 6-well plate overnight. On day 2, DMEM was carefully aspirated and replaced with Opti-MEM. Using Lipofectamine 3000 Transfection Reagent (ThermoFisher), 5 μg of plasmid containing pMXs-NDI1 (Plasmid #72876, Addgene, RRID:Addgene_72876) was transfected into the packaging cells. At 6 hours post-transfection, the media was removed and replaced with 2 mL of fresh DMEM and allowed to incubate for viral production. On day 3, cell lines 786-O and A-498 were seeded at a density of 3 × 10⁵ cells in 2 mL of Roswell Park Memorial Institute (RPMI) media in a 6-well plate. On day 4, the retrovirus was then harvested and filtered. The filtered virus was added directly, dropwise, onto the seeded cells of 786-O and A-498 and the plate was spun at 1,800 RPM for 90 mins. After 2 days post-transduction, the media containing virus were aspirated off, and the cells were passaged with RPMI containing puromycin (2 μg/mL) for 3 days.

For colony formation assay, 1000 cells were seeded into 6-well plates and were grown for 7 days. Colonies were stained with 0.05% Crystal Violet.

MYC was cloned into pCMV-GFP (Addgene Plasmid #11153, RRID:Addgene_11153). Packaging of viral particles and transduction of cells was performed as described above. Colony formation assays were also performed as described above.

CRISPR GFP knock-in

Plasmid expressing guide RNA to insert a LAP tag downstream of the ZNF395 gene was a gift from Kevin White (https://www.encodeproject.org/genetic-modifications/ENCGM446NKU/). The sequence of the guide RNA was CAGCUCAGUCCAGAAAGCGC. Both 786-O and A-498 cells were transfected using the Neon^™ Transfection System (Invitrogen Life technologies) to express a EGFP tag for subsequent TF ChIP. Briefly, 1×10⁶ cells were resuspended in 100 μL of resuspension buffer R along with 10 μg of donor plasmid and 1 μg of guide plasmid. Electroporation was performed at 1100V with 20ms pulse width for 2 pulses. Genomic DNA was harvested using DNeasy Blood & Tissue Kit (Qiagen) to validate the presence of GFP by PCR and sanger sequencing. Refer to Supplementary Information on the primer sequences used for GFP knock-in.

RNA-seq

Ten pairs of normal-tumor tissue matching the ChIP-seq tissues were prepared for RNA-seq. Total RNA was extracted using the Qiagen RNeasy Mini kit. RNA-seq libraries were prepared using Illumina Tru-Seq RNA Sample Preparation v2 protocol, according to the manufacturer’s instructions. Briefly, poly-A RNAs were recovered from 1 μg of input total RNA using poly-T oligo conjugated magnetic beads. The recovered poly-A RNA was chemically fragmented and converted to SuperScript II and random primers. The second strand was synthesized using the Second Strand Master Mix. Libraries were validated with an Agilent Bioanalyzer (Agilent Technologies, Palo Alto, CA), diluted to 11 pM and applied to an Illumina flow cell using the Illumina Cluster Station. Sequencing was performed on HT2000 with 100 bp paired-end reads.

RNA-seq analysis

The analysis of RNA-seq data was conducted using HTSeqGenie. Reads characterized by low nucleotide qualities (defined as having 70% of bases with quality scores below 23) or those matching rRNA and adapter sequences were excluded. The remaining reads were aligned to the human reference genome (human: GRCh38.p10) using GSNAP (RRID:SCR_005483) with a maximum allowance of two mismatches per 75-base sequence (utilizing parameters: ‘-M 2 -n 10 -B 2 -i 1 -N 1 -w 200000 -E 1 --pairmax-rna=200000 --clip-overlap’). Transcript annotation relied on Ensembl release 90. For the quantification of gene expression levels, the count of reads unambiguously mapping to the exons of each gene was computed. Subsequent differential analysis was performed using DESeq2(24) (RRID:SCR_000154) and used for GeneSet Enrichment Analysis (GSEA)(25) with the HALLMARK pathways.

Transcription factor ChIP-Seq

For each transcription factor, ~4×10⁷ cells were cross-linked with 1% formaldehyde for 10 min at room temperature, and stopped by adding glycine to a final concentration of 0.2M. Chromatin was extracted and sonicated to ~500bp (Vibra cell, SONICS). The antibody used for chromatin immunoprecipitation was anti-GFP (Abcam ab290, RRID:AB_303395). The total volume of immunoprecipitation was 0.5 mL and the amount of antibody used was 5 μg. The input DNA was precleared with protein G Dynabeads (LifeTechnologies) for 2 hr at 4°C and then incubated with antibodies conjugated protein G beads overnight at 4°C. The beads were washed with wash buffers 6 times at room temperature. At least 10 ng of the amplified DNA was used with NEBNext ChIP-Seq library prep reagent set (NEB). Each library was sequenced to an average depth of 30–50 million reads on HiSeq2500 using 101bp single end reads.

Transcription factor ChIP-Seq analysis

Sequencing tags were mapped against the human reference genome (hg19) using Burrows-Wheeler Aligner (BWA-mem) (version 0.7.10)(26). Reads were trimmed 10 bp from the front and the back to produce 81 bp. Only reads with mapQ >10 and with duplicates removed by rmdup were used in the subsequent analysis. Significant peaks were called using MACS2(27) (p-value < 1×10⁻⁴).

Immunoblotting

Cells were harvested with cold RIPA lysis buffer (50 mM Tris pH 8, 150 mM NaCl, 0.1% Triton X-100, 0.5% Sodium deoxycholate, 0.1% SDS) with protease inhibitors (Roche) on ice. Cells were lysed by RIPA lysis buffer and centrifuged at 13,000 rpm for 15 min at 4°C. Protein concentration was measured by Pierce BCA protein assay (Life Technologies). Cell lysates were heated at 70°C for 10 min in the sample buffer. Per well, 8 μg of cell lysate was loaded and gel electrophoresis was run at 100V constant for 150 mins. Proteins were transferred to nitrocellulose membranes by transferring at 100V for 100 mins on ice. Western blotting was performed by incubating membranes overnight at 4°C with the following antibodies and dilutions: ZNF395 (1 μg/ml (28)), GFP (1:1000, Abcam ab290, RRID:AB_303395), GLS (1:1000, Abcam 156876, RRID:AB_2721038), IDH2 (1:1000, Cell Signaling Technology 12652, RRID:AB_3720119), HIF2α (1:1000, ProteinTech 26422–1-AP, RRID:AB_2880510), HIF1a (1:1000, Abcam 51608, RRID:AB_880418), and β-actin (1:2000, Santa Cruz sc-47779). Membranes were incubated in secondary antibodies at 1:10,000 dilution for 1 hr at room temperature and developed with SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Scientific).

Quantitative RT-PCR Analysis (qPCR)

Total RNA was extracted from cell lines using the RNeasy Mini Kit (Qiagen). Reverse transcription was performed using the iScript Reverse Transcription Supermix for RT-qPCR (Biorad). qPCR was performed using primers from Integrated DNA Technologies with PowerUp SYBR Green Master Mix (ThermoFisher). B-actin (ACTB) was used to normalize gene expression changes. Refer to the Supplementary Information for primer sequences used in qPCR.

NADP+/NADPH & NAD+/NADH assay

Cells were seeded into 96-well plate at a concentration of 10000 cells per well, with or without doxycycline (0.10 μg/mL), on the day prior to measurement of NADP/NAPDH or NAD/NADH levels using NADP/NADPH-Glo^™ assay (Promega G9081) or NAD/NADH-Glo^™ assay (Promega G9071). Cells were lysed using 0.2N NaOH with 1% DTAB. For NADP+/NAD+ measurement, lysed cells were acid-treated and heat-treated then neutralized with 0.5M Trizma base. For NADPH/NADH measurement, lysed cells were only heat-treated before adding HCl/Trizama solution. Detection reagent was added to both NADP+ and NADPH wells and incubated for an hour. Luminescence intensity was measured using a Tecan plate reader. The assay was then normalized to total protein; BCA analysis was performed on the cells seeded into the clear 96-well plate.

GSH/GSSG assay

Cells were seeded into 96-well plate at a concentration of 10000 cells per well, with or without doxycycline (0.10 μg/mL), on the day prior to carrying out the GSH/GSSG-Glo^™ assay (Promega V6611). Cells were lysed with either Total Glutathione Lysis Reagent to measure total GSH level or Oxidised Glutathione Lysis Reagent to measure oxidized GSH level. Luciferin Generation Reagent and Luciferin Detection Reagent were added to lysed cells according to manufacturer’s protocol. Luminescence intensity was measured using a Tecan plate reader. The assay was then normalized to total protein; BCA analysis was performed on the cells seeded into the clear 96-well plate.

ROS level

Cells were stained with 5uM of MitoSOX^™Red mitochondrial superoxide indicator (ThermoFisher) for 10 min at 37°C and analyzed using FACS.

H₂O₂ assay

Cells were seeded at a density of 3,000 cells per well and treated with H₂O₂ (Sigma-Aldrich, 88597) 24 hours later. After 48 hours of treatment, cell viability was measured using the Cell Counting Kit-8 (Dojindo, CK04).

Drug treatment

Cell lines 786-O and A-498 were seeded at a density of 5,000 cells in 2 mL of either DMSO or drug-containing media per well. On day 4, cells were briefly trypsinized and re-seeded at 50% density for A-498, and 25% density for 786-O in DMSO or drug-containing media for an additional 4 days. On day 8, DMSO or drug was replenished by a complete media change. On Day 10 – 11, cells were fixed in cold methanol for 10 minutes at −20°C and stained with 0.05% crystal violet for 10 minutes. The plates were scanned using the GelCount imager (Oxford Optronix) with the default settings.

Hypoxia treatment

In hypoxia immunoblotting, cell lines were seeded in duplicate 6-well plates to an approximate confluency of 70 – 80% before placing them in a normal CO2 incubator (normoxia) or hypoxic CO2 incubator (hypoxia) (Eppendorf, Galaxy 170R) for 24 hours. The cell lines were seeded with the following densities: HEK293T (1×10⁶ cells/well), T98G and Caki-1 (0.45×10⁶ cells/well), ACHN and HK2 (0.7×10⁶ cells/well), and for 786-O and A-498 (0.3×10⁶ cells/well). After 24 hours post treatment, the protein lysates were harvested as per immunoblotting protocol previously described for both the normoxic and hypoxic condition, except that a hypoxic chamber (Coy Laboratory, O2 Control Glove Box), where O2 was maintained ~1%, was involved for the harvesting of hypoxia protein lysates.

In hypoxia colony formation assay, cell lines were seeded in duplicate 6-well plates before subjecting them to hypoxic or normoxic condition in the incubator for 6 – 10 days. The cell lines were seeded with the following densities: 786-O (3,000 cells), A-498 (5,000 cells), T98G and Caki-1 (7,500 cells). Cells were then fixed in cold methanol for 10 minutes at −20°C and stained with 0.05% crystal violet for 10 minutes. The plates were scanned using GelCount imager (Oxford Optronix) with the default settings.

Mass spectrometry of amino acids and organic acids (TCA cycle intermediates)

Mass spectrometry analysis was performed with the Metabolomics Facility at DUKE-NUS Medical School. Cell pellets from 1 million A-498 cells were solubilized in 50% acetonitrile, 0.3% formic acid. For amino acid extraction, 100 μL of cell lysate was extracted using methanol. The amino acid extracts were derivatized with 3M Hydrochloric acid in butanol (Sigma Aldrich, USA), dried and reconstituted in methanol for analysis in LC-MS. Methods of amino acid analysis were modified from Newgard et al (29) and Sinha et al (30), 2014. The LC-MS analysis was performed using an Agilent 1290 Infinity liquid chromatography system (Agilent Technologies, Santa Clara, CA, USA) and a quadrupole ion trap mass spectrometer (QTRAP 5500, AB Sciex, Olympia, DC, USA). A C18 column (Phenomenex, 100 × 2.1 mm, 1.6 m, Luna^® Omega, Torrance, CA, USA) was used to separate the samples. Chromatography separation was performed using Mobile phase A (Water) and Mobile phase B (Acetonitrile with 0.1% formic acid). The LC run was performed at a 0.4 mL min−1 flow rate for 0.8 min with an initial gradient of 2% B, followed by 15% B in 0.1 min, 20% B in 5.7 min, 50% B in 0.5 min, and 70% B in 0.5 min. The run ended with a 0.9 min re-equilibration of the column to the initial run conditions (2% B).

For organic acid extraction, 300 μL of cell lysate was extracted with ethylacetate, dried and derivatized with N,O-Bis(trimethylsilyl)trifluoroacetamide, with protection of the alpha keto groups using ethoxyamine (Sigma Aldrich, USA). Trimethylsilyl derivatives of organic acids were separated by gas chromatography on an Agilent Technologies HP 7890A and quantified by selected ion monitoring on a 5975C mass spectrometer using stable isotope dilution. The initial GC oven temperature was set at 70 °C, and ramped to 300 °C at a rate of 40 °C/min, and held for 2 min.

Seahorse Assay

The Seahorse XF Cell Mito Stress Test Assay was performed with the Seahorse XFe96 or XFe24 Analyzer (Agilent). A day prior to the assay, cells were seeded at a density of 15,000 cells per 80 μL of media per well for XFe96 plates, or 25,000 cells per 200 μL of media per well for XFe24 plates. On the day of the assay, Seahorse assay medium consisting of sodium pyruvate (1mM), glucose (10mM) and L-Glutamine (2mM) were prepared and filtered. The media in the cell culture microplate were aspirated, washed once with seahorse assay medium, and replaced with 175 μL (XFe96 format) or 500 μL (Xfe24 format) of seahorse assay medium. In addition, the drug compounds were added into the sensor cartridge plate as follows: Port A – final well concentration of 1 μM Oligomycin, Port B – final well concentration of 0.125 μM FCCP, and Port C – final well concentration of 0.5 μM of both Rotenone and Antimycin-A. After the assay run, the cell culture microplate was retrieved from the XFe96 analyzer for BCA analysis of total protein or counting the cell numbers in order to normalize the seahorse assay data.

Isotope tracing

Cells were plated in 6-well plates in quadruplicate (three for tracing experiment and one for determining cell numbers) in a density of 1×106 cells/well the day before experiment. The next day, the cell number was counted for one replicate well per group. For the rest of the three wells, the original media was removed and the cells were washed with PBS one time and then incubated the cells with 2 mM 13C5-glutamine (Cambridge Isotope Laboratories, CLM-1822) in glutamine-free DMEM media for 6 hours. After the incubation, the media was completely aspirated and the cells were quickly rinsed twice with 2 mL of cold normal saline to remove the medium residue completely. The cell plates were then immediately placed on dry ice, and quickly added 1 mL 80% (vol/vol) methanol (cooled to −80 °C) to each well. The plates were incubated at −80 °C for 20 min. Cells were scraped on dry ice and transferred to 1.5 mL pre-cooled Eppendorf tubes. The tubes were centrifuged at 20,000 g for 15 min in a refrigerated centrifuge and a certain amount of supernatant that was normalized against the cell number was transferred into a new Eppendorf tube. Finally, the supernatants were speed vacuum dried at room temperature then stored as dry pellets in −80 °C freezer for LC-MS analysis. The LC-MS analysis was performed as previously described (31). For data analysis, the abundances of the indicated metabolite isotopologues were normalized to the m+5 13C5-glutamine.

Puromycin incorporation assay

Global protein synthesis was assessed using a puromycin incorporation assay as described previously(32). Cells were seeded and allowed to adhere overnight with 80% confluency. Puromycin (Thermo Scientific, Cat# A1113803, 10 μg/mL) was added directly to the culture medium and incubated for 1 hour at 37 °C under standard culture conditions. Following treatment, cells were washed twice with cold PBS and subjected to standard Immunoblotting analysis.

Data Availability

ZNF395 and MYC ChIP-seq and RNA-seq can be accessed at GEO (GSE247118). ChIP-seq of HIF1α (SRR1929392) and RNA-seq were downloaded from GEO (GSE67237). RNA-seq data of 786-O and A-498 treated with sgHIF2α and PT2399 were downloaded from GEO (GSE253327 and GSE253373). RNA-seq data of RCC4 treated with sgHIF21 and/or sgHIF2α were downloaded from GEO (GSE269819). RNA-seq data of PDX models treated with PT2399 were downloaded from Sequence Read Archive (SRP073253). Gene expression of HIF1/2α, HIF2α-only and HIF-negative tumors were downloaded from GEO (GSE11904). ScRNA-seq of CRC tumors were downloaded through the HTAN DCC Portal (https://data.humantumoratlas.org/publications/hta12_2023_nature_nadezhda-v-terekhanova). TCGA counts were downloaded from GDC Data Portal and the VHL status was downloaded from cBioPortal using the R package cBioPortalData. Protein abundance values from CPTAC were downloaded from LinkedOmics(https://kb.linkedomics.org/download). Gene expression of CCLE cell lines was downloaded from the R package Depmap.

Code Availability

Analysis code can be found at https://github.com/xiaosaiyao/ZNF395

Results

ZNF395 is a HIF-activated transcription factor critical for ccRCC tumorigenesis

Previously we showed that the knockdown of ZNF395 abrogated in vivo tumor formation of ccRCC cell line, A-498 and significantly delayed tumor progression of a second ccRCC cell line, 786-O (21). We now systematically explored the expression of ZNF395 across a number of tissues and cell lines. Amongst all the cancer types in the Cancer Genome Atlas (TCGA) dataset, ccRCC exhibits the highest expression of ZNF395 (Figure 1a). We confirmed the overexpression of ZNF395 at the protein level by interrogating mass spectrometry data profiled by the Clinical Proteomic Tumor Analysis Consortium (CPTAC) (Figure 1b). The overexpression of ZNF395 in ccRCC was also consistently observed in cancer cell lines (Figure S1a,b). A ccRCC scRNA-seq dataset showed that ZNF395 is mainly expressed in renal tumor cells but not in other cells within the tumor microenvironment (Figure 1c) (33).

Figure 1:

ZNF395 is a HIF-activated transcription factor critical for ccRCC

a) Gene expression level of ZNF395 expressed as reads per kilobase per million reads (RPKM) in the Cancer Genome Atlas (TCGA) dataset. FDR values calculated by DEseq2 using Benjamini-Hochburg correction of p-values from the Wald test.

b) Relative protein expression of ZNF395 in primary ccRCC and adjacent normal tissues profiled by CPTAC. Relative protein levels were downloaded from linkedOmicsKB. P-values calculated by 2-sided t-test.

c) Single cell RNA-seq of primary ccRCC tumors was obtained from Terekhanova et al (33).

d) CcRCC tumor samples from TCGA were stratified by VHL copy number status and the ZNF395 gene expression of the corresponding groups are shown. FDR values calculated by DEseq2 using Benjamini-Hochburg correction of p-values from the Wald test.

e) Effect of CRISPR-mediated depletion of HIF1α and/or HIF2α on ZNF395 gene expression, based on RNA-seq data from Jiang et al (52) and Lombardi et al (53)

f) Effect of HIF2α inhibitor, PT2399, treatment on ZNF395 expression in vitro, based on RNA-seq data from Jiang et al (52)

g) Effect of PT2399 on ZNF395 expression in PDX models established from ccRCC tumors profiled by RNA-seq (34).

h) Hypoxia signature is calculated by summing up the z-normalized expression from the 48-gene signature from Lombardi et al (43). Hypoxia is then correlated with ZNF395 expression. Refer to Table S1 for correlation values across all TCGA tumor types.

i) Cells were cultured under normoxia and under hypoxia (2% oxygen) for 24 hours and assayed for ZNF395 expression by immunoblotting.

j) Histone ChIP-seq is displayed for a ccRCC tumor along with its adjacent normal tissue. Data from Yao et al (54)

Previously we showed that the re-expression of VHL or HIF2α siRNA-mediated knockdown can reduce the gene expression of ZNF395 (21). We now explored the relationship between VHL loss and ZNF395 expression in tumor samples. In the ccRCC TCGA dataset, increased expression of ZNF395 scales with deeper deletion of the VHL gene (Figure 1d). In contrast, the loss of other common tumor suppressors, such as PBRM1 and SETD2, did not affect the expression of ZNF395 (Figure S1c). HIF2α depletion downregulated ZNF395 in HIF1α-deficient cell lines (786-O and A-498), and HIF1α depletion led to decreased ZNF395 expression in the HIF1α/HIF2α double positive cell line RCC4 (Figure 1e). HIF2α-only and HIF1α/HIF2α double positive tumors both expressed higher levels of ZNF395 compared to HIF negative tumors (Figure S1d). Pharmacological inhibition of HIF2α with PT2399 reduced ZNF395 expression (Figure 1f), and downregulation of ZNF395 was associated with PT2399 sensitivity (FDR = 3×10⁻⁷) (Figure 1g) (34). Taken together, these results demonstrate that high ZNF395 expression is a hallmark of VHL-deficient ccRCC and is regulated by both HIF1α and HIF2α.

Beyond ccRCC, ZNF395 is also overexpressed in glioblastoma and other RCC subtypes (Figure 1a). ZNF395 can be induced by hypoxia in a variety of cell lines including neuroblastoma, glioblastoma (U937 and U87-MG cell lines), breast cancer cell line (MCF7), hepatocellular carcinoma (Hep3B), osteosarcoma and adipocytes (28,35–42). Furthermore, ZNF395 expression is highly correlated with a pan-cancer hypoxia signature (43) in renal cell carcinomas (clear cell and papillary) and brain tumors (glioblastoma, paraganglioma & pheochromocytoma and low grade gliomas) (Figure 1h, Table S1). We observed that the expression of ZNF395 increased under hypoxia in VHL-proficient RCC cell lines (Caki-1, ACHN), normal kidney proximal tubular cell line (HK2) and glioblastoma cell line T98G (Figure 1i). However, the depletion of ZNF395 only impacted cell fitness of VHL-deficient ccRCC cell lines (Figure S1e). Therefore, ZNF395 is a broadly hypoxia-inducible gene but is the most critical for the fitness of VHL-deficient ccRCC cells.

ZNF395 is regulated by an extensive and highly specific cancer-specific super-enhancer. The ZNF395 locus exhibits elevated H3K27ac signals in tumor tissue compared to the adjacent normal tissue, and is enriched for H3K4me1, a histone modification indicative of enhancer identity (Figure 1j). We showed previously that VHL restoration or siRNA-mediated knockdown of HIF2α abolished P300 recruitment at ZNF395 super-enhancer, and dampened ZNF395 super-enhancer activity and gene expression (21).

ZNF395 regulates the expression of key metabolic enzymes in hypoxia

To understand the function of ZNF395 in ccRCC, we mapped out the binding profiles of ZNF395. Because a ChIP-grade antibody of ZNF395 is not commercially available, we utilized CRISPR technology to knock in an eGFP tag at the endogenous locus of ZNF395 and performed ChIP-seq against the eGFP tag (Figure S2a). We further verified that ZNF395 was pulled down using ChIP-mass spectrometry and that none of CRISPR off targets were pulled down (Figure S2b, Table S2). We obtained a set of 5763 highly specific ChIP-seq peaks common to both A-498 and 786-O cell lines (Figure 2a).

Figure 2:

ZNF395 regulates the expression of key metabolic enzymes in hypoxia

a) eGFP was introduced downstream of the ZNF395 locus using CRISPR genome editing. ChIP-seq of eGFP-tagged ZNF395 was performed in 786-O and A-498 cells.

b) Genomic distribution of ZNF395 ChIP-seq peaks profiled by ChIPseeker

c) De novo motif discovery using HOMER identified a motif that resembles CTCFL (aka BORIS), another C2H2 zinc finger protein.

d) Nearest genes were mapped to the ZNF395 binding sites and enrichment of GO biological processes was performed using GREAT

e) qPCR validation of a set of metabolic genes downregulated by ZNF395 shRNA-mediated knockdown

f) Immunoblotting of metabolic enzymes following shRNA-mediated knockdown of ZNF395 in A-498 and 786-O cells

g) Browser view of ChIP-seq profiles at the loci of metabolic enzymes GLS, IDH2 and PDK2

h) Genomic occupancy of ZNF395, HIF2α, HIF1β and MYC at ZNF395 or HIF2α binding sites in 786-O cells

i) Overlap of ZNF395, HIF2α and MYC ChIP-seq peaks in 786-O cells using ChIPpeakAnno

ZNF395 is a promoter-centric transcription factor and 80% of the peaks are located in the promoter regions (Figure 2b). De novo motif analysis identified a motif that most closely resembles CTCFL, also from the C2H2 zinc finger protein (Figure 2c). Gene ontology analysis of the overlapped peaks gave metabolic process as the most significant pathway (hypergeometric q-value = 2×10⁻¹²⁷) (Figure 2d).

We interrogated the common transcriptomic changes when we knocked down ZNF395 by two independent shRNAs (shRNA-1 and shRNA-3). GSEA indicated upregulation of inflammation and downregulation of metabolic pathways including lipid metabolism and oxidative phosphorylation (Figure S3a). We observed enrichment of downregulated genes at the influx of glutamine into the TCA cycle, the TCA cycle and fatty acid synthesis (Figure S3b). These pathways correspond to glutaminolysis, a prominent mechanism for replenishing TCA cycle intermediates, and providing the substrates for synthesis of fatty acids in VHL-deficient or hypoxic cells (8).

Some key downregulated targets include enzymes of glutaminolysis GLS and IDH2, and the suppressor of glycolytic flux into the TCA cycle PDK2 (Figure 2e). The decrease in transcriptomics level also translated to decreased protein levels (Figure 2f). ChIP-seq showed ZNF395 binding at the promoter of GLS, IDH2 and PDK2, suggesting that they are direct targets of ZNF395 (Figure 2g). The regulation of these metabolic genes by ZNF395 is confined to VHL-deficient ccRCC cells since the depletion of ZNF395 did not lead to any consistent changes in HK2 cells (Figure 2e). ZNF395 did not have any effects on angiogenesis associated genes such as VEGFA, ANGPTL4 and ADM (43), further proving its non-overlapping function with HIF2α (Figure S4a–b). Therefore, ZNF395 is a master transcription factor that regulates the expression of hypoxia-associated metabolic genes in VHL-deficient ccRCC cells.

A long-standing conundrum in the ccRCC field is that although HIF2α is a master regulator of the hypoxic response, it neither binds to the promoter regions of glutamine metabolism enzymes nor alter their expression (9). Consistent with this, the binding sites of ZNF395 are largely distinct from HIF2α peaks (Figure 2h), and treatment with HIF2α inhibitors did not reduce GLS expression in vitro (Figure S4c) or in vivo (Figure S4d). In contrast to HIF2α, HIF1α has been shown to regulate GLS expression under hypoxic conditions (44). When introduced into the HIF1α-deficient 786-O cells, HIF1α shares many binding sites with ZNF395 (45) (Figure S4e), upregulates the expression of both ZNF395 and GLS (Figure S4f), and directly binds at the promoters of these genes (Figure S4g). This suggests that HIF1α-deficient ccRCC cells may be especially susceptible to ZNF395 depletion.

We also compared ZNF395 against MYC, a known regulator of glutaminolysis (17). More than half of the ZNF395 binding sites overlapped with MYC in 786-O cells (Figure 2h–i). Expression levels of ZNF395 and MYC showed a moderate correlation in TCGA (Figure S5a). Depletion of MYC and ZNF395 each led to reduced expression of GLS and IDH2 (Figure S5b) and reduced colony formation in A-498 and 786-O cells (Figure S5c). However, co-depletion of MYC and ZNF395 did not consistently reduce GLS and IDH2 levels or clonogenic potential (Figure S5b,c). Furthermore, ZNF395 overexpression did not compensate for the loss of MYC and vice versa (Figure S5d,e), suggesting that ZNF395 and MYC play non-redundant roles despite their ability to regulate glutaminolysis individually.

ZNF395 maintains TCA anaplerosis

We next sought to investigate the metabolic effects of ZNF395 depletion. By performing LC-MS, we observed a significant decrease in the levels of TCA cycle intermediates following ZNF395 depletion in A-498 cells (Figure 3a). Glutamine is a starting substrate for the synthesis of glutamate, proline and alanine and ZNF395 loss led to reduction of these three amino acids (Figure 3b). Reduction in amino acids did not decrease protein synthesis rate, as assayed by puromycin incorporation assay (Figure S6a).

Figure 3:

ZNF395 maintains TCA anaplerosis

a) Levels of TCA cycle intermediates measured by GC-MS in A-498 cells following shRNA-mediated knockdown of ZNF395

b) Levels of amino acids measured by LC-MS in A-498 cells following shRNA-mediated knockdown of ZNF395

c) NADPH/NADP levels measured using NADP/NADPH-Glo^™ assay in A-498 cells following shRNA-mediated knockdown of ZNF395

d) GSH/GSSG measured using GSH/GSSG-Glo^™ assay in A-498 cells following shRNA-mediated knockdown of ZNF395

e) Levels of reactive oxygen species (ROS) in A-498 cells following shRNA-mediated knockdown of ZNF395 determined by staining cells with MitoSOX^™Red mitochondrial superoxide indicator and quantifying by flow cytometry

f) Cell viability measured by CCK8 after 24 hr treatment of H₂O₂ in A-498 cells

g) Isotope tracing using ¹³C₅-glutamine was performed to determine the abundance of TCA cycle intermediates, glutathione, pyrimidine nucleotides and amino acids derived from glutaminolysis

h) Levels of TCA cycle intermediates, glutathione and pyrimidine nucleotides detected in A-498 cells following shRNA-mediated knockdown of ZNF395 from experiment in g) split by oxidative and reductive pathways

i) Levels of GSH and GSSG from experiment g)

j) Levels of amino acids from experiment g)

k) Levels of TCA cycle intermediates measured by GC-MS in 786-O cells following shRNA-mediated knockdown of ZNF395 propagated in vivo

l) Levels of amino acids measured by LC-MS in 786-O cells following shRNA-mediated knockdown of ZNF395 propagated in vivo

Since glutaminolysis is critical for maintaining redox homeostasis, we measured the effect of ZNF395 perturbation on redox potential. Depletion of ZNF395 led to a decrease in redox potential as measured by NADPH/NADP ratio and GSH/GSSG ratio, and a consistent increase in reactive oxygen species as measured by mitoSox (Figure 3c–e). Depletion of ZNF395 sensitized cells to hydrogen peroxide treatment, likely due to the reduced ability to neutralize reactive oxygen species (Figure 3f, Figure S6b).

To conclusively determine the effects of ZNF395 on glutaminolysis, we performed isotope tracing using ¹³C₅-glutamine and compared the abundance of labeled downstream metabolites following ZNF395 depletion (Figure 3g). In A-498 cells, we observed significant reduction of TCA cycle intermediates (malate, citrate and 2-hydroxyglutarate). Interestingly, ZNF395 depletion significantly impacted the glutamine oxidative pathway, while having less effect on reductive carboxylation (Figure 3h, Figure S6c). In addition to TCA cycle intermediates, ZNF395 depletion also led to reduction in glutathione (GSH and GSSG), glutamine-derived amino acids (glutamate, aspartate and oxoproline), and pyrimidine nucleotides (UMP and CMP) (Figure 3h–j). This tracing experiment confirmed the requirement of ZNF395 in maintaining TCA anaplerosis and redox homeostasis.

Finally, we showed that depletion of ZNF395 reduced the levels of lactate, TCA cycle intermediates (fumarate and malate) and nearly all detectable amino acids in 786-O cells in vivo (Figure 3k,l). We observed a similar decrease in TCA cycle intermediates and amino acids in A-498 expressing shZNF395–7 (Figure S6d,e). We failed to measure a consistent change in A-498 expressing ZNF395-sh1. This is because shZNF395–1 induced a significant growth defect in A-498 in vivo (21), and the remaining tissue escaped ZNF395 depletion (Figure S6f). Taken together, ZNF395 is important for maintaining TCA anaplerosis, synthesis of glutamine-derived amino acids and pyrimidine nucleotides and redox balance.

ZNF395 maintains mitochondrial respiration

Because the TCA cycle generates NADH as the substrate for electron transport chain, we reasoned that oxidative phosphorylation may be impacted by the reduced TCA activity. We first showed that ZNF395 depletion reduced the levels of both NADH and NAD+ (Figure 4a). Then by performing the Seahorse mito stress test, we observed that ZNF395 depletion led to a significant reduction in the maximal respiration and spare respiration in both ccRCC cell lines, A-498 and 786-O (Figure 4b–c, Figure S7a–b). ZNF395 depletion further reduced the rates of basal respiration and ATP production in 786-O cells (Figure 4b,c). In contrast, we did not see consistent decreases in glycolysis when these cell lines were subjected to ZNF395 siRNA mediated depletion (Figure S7c–d).

Figure 4:

ZNF395 maintains mitochondrial respiration

a) Levels of NADH, NAD+ and NADH/NAD+ ratio following dox-inducible shRNA-mediated knockdown of non-targeting control (NTC) or ZNF395. ***p-value < 0.001, **p-value < 0.01, two-tailed t-test

b) Seahorse mito stress test was performed in 786-O expressing two dox-inducible shRNA against ZNF395 or non-targeting control (shNTC)

c) Mitochondrial respiration rate from b). ****p-value < 0.0001, **p-value < 0.01, *p-value <0.05, two-tailed t-test

d) Seahorse mito stress test was performed in 786-O cells expressing either empty vector or NDI-1, a yeast mitochondrial NADH dehydrogenase that oxidizes NADH to NAD+ and thus supplements the mammalian complex I. These two cell lines were further subjected to ZNF395 shRNA-mediated knockdown.

e) Mitochondrial respiration rate from d)

f) Same as d) but performed in A-498 cells

g) Same as e) but performed in A-498 cells

h) Colony formation assay of 786-O and A-498 with either empty vector or NDI-1 paired with shRNA against non-targeting control (NTC) or ZNF395

The reduced oxygen consumption rate coupled with reduced glutamine metabolism suggests impaired mitochondrial function caused by ZNF395 depletion. We explored whether restoration of mitochondrial respiration could rescue the growth inhibition caused by ZNF395 depletion. Even though ZNF395 did not lead to decreased gene expression (Table S3) or protein expression of the oxidative phosphorylation (OXPHOS) system (Figure S7e), overexpression of OXPHOS could sometimes compensate for reduced oxygen consumption rate. We introduced NDI-1, the yeast mitochondrial NADH dehydrogenase that oxidizes NADH to NAD+ (46) and supplements the function of the mammalian complex I in the electron transport chain. Expression of NDI-1 restored oxygen respiration level, increasing the rate of basal and maximal respiration in both 786-O and A-498 cells (Figure 4d–g). However, while NDI-1 overexpression rescued cell fitness in 786-O, it failed to rescue A-498 cells (Figure 4h), suggesting the mitochondrial restoration induced by ND-1 may affect cell fitness in a cell line specific manner and that mitochondrial restoration alone might not be sufficient to overcome the growth defect induced by ZNF35 depletion.

GLS rescues growth defects due to ZNF395 depletion

Next we ask whether restoration of glutaminolysis could overcome the loss in cell fitness. Supplementation of glutaminolytic substrates including glutamine, glutamate and alpha-ketoglutarate failed to rescue the growth defects (Figure 5a). Since ZNF395 loss led to the downregulation of glutaminolytic enzymes (Figure 2e,f), we reasoned that restoration of enzymatic function may be necessary to rescue cell fitness. We overexpressed GLS in A-498 cells expressing ZNF395 shRNA-1 (Figure 5b–c). GLS overexpression did not increase ZNF395 expression, suggesting GLS is a downstream target of ZNF395. GLS overexpression restored mitochondrial respiration, increasing basal, ATP-linked, maximal and spare respiration rates (Figure 5d,e). Importantly, GLS overexpression led to partial restoration of clonogenic potential of A-498 cells (Figure 5f). Similarly, overexpression of GLS overcame the in vivo growth defects due to ZNF395 depletion (Figure 5g–i). This suggests that a major function of ZNF395 is to maintain the expression of GLS and support glutaminolytic flux and mitochondrial function necessary for ccRCC fitness under hypoxic stress.

Figure 5:

GLS rescues growth defects due to ZNF395 depletion

a) Supplementation of glutaminolysis substrates - glutamate, glutamine and alpha-ketoglutarate (a-KG) - in A-498 cells expressing shRNA against non-targeting control (NTC) or ZNF395

b) Empty vector (EV) or GLS was introduced into A-498 cells expressing shRNA against non-targeting control (NTC) or ZNF395. Shown are the levels of ZNF395 mRNA measured by RT-qPCR

c) Same as b) but showing the expression of GLS

d) Seahorse mito stress test was performed in A-498 cells expressing empty vector (EV) or GLS coupled with shRNA against non-targeting control (NTC) or ZNF395

e) Rate of respiration measured in e)

f) Colony formation of A-498 cells with expressing empty vector (EV) or GLS coupled with shRNA against non-targeting control (NTC) or ZNF395

g) In vivo tumor formation in mice bearing A-498 cells expressing empty vector (EV) or GLS coupled with shRNA against non-targeting control (NTC) or ZNF395 (n = 8). Tumor volume plots represent mean ± standard error of the mean (s.e.m.)

h) Images of individual tumors in g)

i) Individual tumor weight plots represent mean ± s.d. (two-sided t-test).

Discussion

ZNF395 is a master regulator critical for the tumorigenesis of clear cell renal cell carcinoma; its depletion is incompatible with tumor growth (21). Here we have elucidated the functional role of ZNF395 in enabling transcriptional regulation of metabolic enzymes in hypoxic response. These enzymes, including GLS, IDH2 and PDK2, regulate key components of the glutamine metabolism. As a result, ZNF395 is essential for the mitochondrial function of ccRCC cells by maintaining TCA anaplerosis, synthesis of glutamine-derived amino acids and nucleotides, redox potential and oxidative phosphorylation.

HIF2α is the most well-established oncogenic driver associated with VHL loss but HIF2α itself does not directly regulate the transcription of glutaminolytic enzymes (9). We now show that HIF2α can still propagate hypoxic reprogramming through a regulatory cascade that activates the super-enhancer region governing ZNF395 which in turn regulates the transcription of glutaminolytic enzymes. HIF2α and ZNF395 regulate different components of the hypoxia response, with HIF2α regulating genes in the glycolytic pathway and ZNF395 regulating genes in glutamine metabolism and its associated pathways including TCA anaplerosis, macromolecule synthesis and redox homeostasis. ZNF395 is transcriptionally regulated by both HIF1α and HIF2α. Although ZNF395 is expressed in both HIF2α-only and HIF1/2α RCC tumors, HIF1α-deficient tumors may be more vulnerable to ZNF395 inhibition, as HIF1α can partially compensate for the regulation of glutamine metabolism (44).

The recent approval of belzutifan represents an exciting advance in ccRCC treatment. However, resistance to HIF2α inhibitors remains a major challenge. Glutaminase inhibitors, such as telaglenastat, have shown limited clinical activity (47), likely due to the plasticity and redundancy of metabolic pathways. This underscores the need to identify novel therapeutic targets in ccRCC. Peptides of ZNF395 are presented by HLA-B*5502, HLA-A*0201 and HLA-A*24 in sarcomas (48–50), and T cell receptor engineered T cells directed against a ZNF395 peptide elicited antitumor effects in tumors engineered to express ZNF395 peptides (51). These findings suggest that ZNF395 can be targeted by multiple modalities including peptide-based immunotherapies or adoptive cell therapies.

Our study has several limitations. First, we performed ChIP-seq using antibodies against the GFP-tagged version of ZNF395 as neither commercial nor our custom antibodies could produce reliable peaks. Future studies using ChIP-seq grade antibodies against endogenous ZNF395 are needed to confirm its binding sites. Second, we only examined the effect of ZNF395 on glycolysis using transient siRNA-mediated depletion of ZNF395. Given the interconnected nature of metabolic pathways, the role of ZNF395 on other pathways beyond glutamine metabolism warrants further investigation.

In conclusion, we have shown that ZNF395 is a transcription factor that permits coordinated transcription of hypoxia-responsive enzymes in VHL-deficient ccRCCs. It could become an attractive target in kidney cancers through peptide-based immunotherapy or degraders.

Statement of Significance:

ZNF395 and HIF are complementary mediators of hypoxia-induced metabolic reprogramming and therapeutic targets in VHL-deficient kidney cancer, with the former regulating glutamine metabolism and the latter regulating glucose metabolism.