Cysteine dioxygenase 1 is a metabolic liability for non-small cell lung cancer

NRF2 is emerging as a major regulator of cellular metabolism. However, most studies have been performed in cancer cells, where co-occurring mutations and tumor selective pressures complicate the influence of NRF2 on metabolism. Here we use genetically engineered, non-transformed primary murine cells to isolate the most immediate effects of NRF2 on cellular metabolism. We find that NRF2 promotes the accumulation of intracellular cysteine and engages the cysteine homeostatic control mechanism mediated by cysteine dioxygenase 1 (CDO1), which catalyzes the irreversible metabolism of cysteine to cysteine sulfinic acid (CSA). Notably, CDO1 is preferentially silenced by promoter methylation in human non-small cell lung cancers (NSCLC) harboring mutations in KEAP1, the negative regulator of NRF2. CDO1 silencing promotes proliferation of NSCLC by limiting the futile metabolism of cysteine to the wasteful and toxic byproducts CSA and sulfite (SO32-), and depletion of cellular NADPH. Thus, CDO1 is a metabolic liability for NSCLC cells with high intracellular cysteine, particularly NRF2/KEAP1 mutant cells.


Introduction
NRF2 (Nuclear factor-erythroid 2 p45-related factor 2 or NFE2L2) is a stress-responsive cap'n'collar (CNC) basic region leucine zipper (bZIP) transcription factor that directs various transcriptional programs in response to oxidative stress. Under basal conditions, NRF2 is kept inactive through binding to its negative regulator KEAP1 (Kelch-like ECHassociated protein), which is a redox-regulated substrate adaptor for the Cullin (Cul)3-RING-box protein (Rbx)1 ubiquitin ligase complex that directs NRF2 for degradation (Kobayashi et al., 2004). KEAP1 is the major repressor of NRF2 in most cell types, which is supported by the evidence that disruption of Keap1 in the mouse increased the To examine the immediate consequence of constitutive NRF2 stabilization on cellular metabolism in non-transformed cells, we generated a genetically engineered mouse model expressing the KEAP1 R554Q loss-of-function mutation found in human lung cancer. Using this model, we have examined the control of cellular metabolism by NRF2 in mouse embryonic fibroblasts (MEFs) and find that NRF2 promotes the accumulation of intracellular cysteine (CYS) and sulfur-containing metabolites, including GSH and the intermediates of the taurine biosynthesis pathway cysteine sulfinic acid (CSA) and hypotaurine (HTAU). Entry of CYS into the taurine synthesis pathway was mediated by cysteine dioxygenase 1 (CDO1), which was elevated in KEAP1 R554Q MEFs. Taurine synthesis is initiated by the irreversible metabolism of CYS by CDO1 to CSA, which is then decarboxylated by cysteine sulfinic acid decarboxylase (CSAD) to HTAU. In turn, HTAU is non-enzymatically converted to taurine (TAU), or is transaminated by the cytosolic aspartate aminotransferase (GOT1) to produce β-sulfinyl pyruvate, which spontaneously decomposes to pyruvate and sulfite (SO 3 2-). At the organismal level, decarboxylation of CSA via CSAD predominates over transamination by GOT1 (Weinstein et al., 1988). By contrast, lung cancer cell lines accumulated significant CYS due to epigenetic silencing of the CDO1 locus. CDO1 re-expression antagonized proliferation and promoted the metabolism of CYS to CSA, but surprisingly most CSA was exported from cells or transaminated to produce toxic SO 3 2-. Further, continual (CYS) 2 reduction to replenish the CYS pool impaired NADPH-dependent cellular processes. These results demonstrate that CDO1 antagonizes the proliferation of lung cancer cells with high intracellular cysteine and its expression is selected against during tumor evolution.

NRF2 promotes the accumulation of sulfur-containing metabolites
To evaluate how constitutive NRF2 activity reprograms metabolism, we generated a genetically engineered, conditional knock-in mouse model of the cancer mutation KEAP1 R554Q ( Figure 1A). Mutations at this residue prevent the association of KEAP1 with NRF2, thereby stabilizing NRF2 and inducing the expression of NRF2 target genes (Hast et al., 2014). We inserted a loxP-flanked wild-type Keap1 cDNA upstream of the R554Q mutation in exon 4 in the endogenous Keap1 gene. Prior to exposure to Cre recombinase, wild-type KEAP1 protein is expressed. Following Cre-mediated excision of the loxP-flanked cargo, mutant Keap1 R554Q is expressed at physiological levels, thus recapitulating the genetic events of human NSCLC and allowing for the interrogation of the consequences of KEAP1 R554Q expression in an isogenic system. Mouse embryonic fibroblasts (MEFs) harboring this allele were derived to evaluate the consequence of KEAP1 R554Q expression in primary cells. The expression of homozygous KEAP1 R554Q led to NRF2 accumulation and increased expression of the NRF2 target NQO1 ( Figure   1B). We performed non-targeted metabolomics to identify metabolite alterations in these cells and found that the most abundant metabolites following NRF2 accumulation are sulfur-containing metabolites derived from cysteine (CYS) ( Figure 1C), while infection of wild-type MEFs with adenoviral Cre did not significantly alter metabolite levels ( Figure   S1A). To interrogate cysteine metabolism in more detail, we performed targeted metabolomics to quantify the concentration of intracellular CYS and its downstream metabolites. As expected, NRF2 promoted an increase in intracellular CYS and its downstream metabolite GSH ( Figure 1D-F), consistent with previous observations that NRF2 promotes the uptake of (CYS) 2 , 1999). Surprisingly, we also observed a significant increase in intermediates of the taurine biosynthesis pathway, including cysteine sulfinic acid (CSA), and hypotaurine (HTAU) ( Figure 1G-I). Importantly, HTAU is a highly abundant metabolite and the increase of HTAU was similar to the increase of GSH in the millimolar range ( Figure 1F,H), suggesting that entry into the taurine biosynthesis pathway may represent a significant percentage of total CYS usage. Collectively, these results indicate that NRF2 promotes the accumulation of intracellular cysteine and entry of cysteine into multiple downstream pathways.

NRF2 promotes the entry of cysteine into the taurine synthesis pathway via CDO1 in non-transformed, primary MEFs
The significant accumulation of intracellular CYS and taurine synthesis intermediates led us to hypothesize that NRF2 promotes the accumulation of cysteine dioxygenase (CDO1) protein, which is stabilized following CYS accumulation due to a loss its ubiquitination and degradation (Dominy et al., 2006). We observed a robust increase in CDO1 protein in KEAP1 R554Q MEFs compared to KEAP1 WT MEFs in the absence of an increase in mRNA expression (Figures 2A-B), consistent with the known mechanism of CDO1 regulation. To examine whether CDO1 mediates CYS metabolism to CSA and HTAU, and whether this limits the use of CYS for GSH synthesis, we deleted CDO1 with CRISPR/Cas9, followed by infection with empty or Cre expressing adenovirus to generate CDO1-deficient, isogenic KEAP1 WT and KEAP1 R554Q MEFs. Western analysis of CDO1 protein revealed a significant reduction of CDO1 expression in KEAP1 R554Q MEFs, although the already low CDO1 levels did not change significantly in KEAP1 WT MEFs ( Figure 2C). We performed quantitative 13 C 6 -cystine [(CYS) 2 ] tracing to examine the entry of CYS into GSH and TAU synthesis and found that depletion of CDO1 inhibited HTAU synthesis from CYS ( Figure 2H-I). By contrast, the total CYS and GSH levels as well as GSH labeling from 13 C 6 labeled (CYS) 2 were increased by CDO1 depletion. (Figure 2D-F). These results demonstrate that CDO1 accumulation in KEAP1 R554Q MEFs promotes CYS entry in into the TAU synthesis pathway, thereby limiting CYS accumulation and GSH synthesis.

CDO1 is preferentially silenced in KEAP1 mutant NSCLC and antagonizes proliferation
The limitation of CYS and GSH levels by CDO1 suggests that this enzyme may antagonize NRF2-dependent processes in cancer. Thus, we hypothesized that the CDO1-mediated CYS homeostatic control mechanism might be deregulated in NSCLC, allowing enhanced CYS entry into GSH synthesis and other pathways. To evaluate this possibility, we examined the expression of CDO1 in NSCLC patient samples from The Cancer Genome Atlas (TCGA). CDO1 mRNA expression was significantly lower in lung adenocarcinoma samples compared to normal lung ( Figure 3A), which was associated with CDO1 promoter methylation ( Figure 3B) and poor prognosis ( Figure S3A).
Interestingly, the incidence of CDO1 promoter methylation was significantly higher in KEAP1 mutant lung adenocarcinoma compared to wild-type ( Figure 3B), suggesting that CDO1 expression confers a selective disadvantage in the context of NRF2 accumulation. CDO1 protein expression was undetectable in a panel of NSCLC cell lines with the exception of H1581 cells ( Figure 3C), and treatment with the DNMT inhibitor decitabine restored CDO1 mRNA expression ( Figure S3B). These results indicate that CDO1 epigenetically silenced by promoter methylation in NSCLC cell lines and patient samples.
To investigate the NRF2-dependent regulation of CDO1 protein in NSCLC, we generated a doxycycline-inducible lentiviral expression system to reintroduce GFP, The level of CDO1 protein expression in these cells was similar with the physiological CDO1 levels in mouse lung and liver ( Figure S3D), with liver being one of the highest CDO1-expressing tissues that is responsible for supplying taurine to the body (Stipanuk et al., 2015). We find that CDO1 accumulated to higher levels in KEAP1 MUT cells than KEAP1 WT , although accumulation was observed in many KEAP1 WT cell lines as well ( Figure 3D). We investigated the association with intracellular cysteine levels across the panel of parental cell lines and found a strong association between the level of CDO1 accumulation and intracellular CYS levels but not with the level of CDO1 mRNA expressed from our inducible promoter system ( Figure 3D, S3C), which is consistent with our findings in KEAP1 R554Q MEFs (Figure 2A-B).
To directly examine the effect of NRF2 on CDO1 expression in NSCLC cell lines, we used multiple isogenic cell systems. First, we used NRF2-deficient A549 cells (Torrente et al., 2017), in which we restored NRF2 expression in combination with CDO1 WT , CDO1 Y157F , or GFP. NRF2 restoration in these cells led to higher expression of endogenous xCT and accumulation of ectopically expressed CDO1 compared to GFP control ( Figure 3E). Next, we selected the two KEAP1 WT NSCLC cell lines that had the lowest CDO1 accumulation and low intracellular CYS in our cell line panel, H1299 and H1975. Using a NRF2 T80K mutant that is unable to bind KEAP1 (Berger et al., 2017), the effects of NRF2 on CDO1 accumulation was recapitulated in these KEAP1 WT NSCLC cell lines ( Figure 3F). Interestingly, we observed that NRF2 T80K could also promote the accumulation of endogenous CDO1 in H1299 cells. Consistently, NRF2 depletion in KEAP1 MUT NSCLC cells following KEAP1 WT expression led to CDO1 depletion ( Figure   3G, S3E-F), although the effects were more modest than what was observed with NRF2 activation. Notably, NRF2 expression promoted intracellular CYS accumulation, while NRF2 depletion impaired CYS accumulation ( Figure 3H), supporting a role for intracellular CYS in CDO1 stabilization. To directly asses the requirement for CYS, A549 cells were cultured in high or low (CYS) 2 and CDO1 levels were found to be dependent on CYS availability ( Figure S3G).
Next, we examined the consequence of CDO1 expression on cellular proliferation.
Using the isogenic NRF2 KO A549 cell system, we observed that CDO1 expression significantly impaired the proliferation of NRF2-expressing cells, while no effect was observed on NRF2 KO cells ( Figure 3I). Looking more broadly, we observed that CDO1 expression generally antagonized the proliferation of NSCLC cell lines and proliferation inhibition was strongly correlated with CDO1 protein expression, but not RNA expression ( Figure 3J and S3H-I). Overall, these results demonstrate that NRF2 and other mechanisms of intracellular CYS accumulation promote CDO1 accumulation, which leads to a selective growth disadvantage in lung cancer cells.
CDO1 depletes cysteine, leading to its export as CSA To evaluate the mechanism by which CDO1 expression impaired proliferation we interrogated CYS metabolism following CDO1 expression. CYS has multiple intracellular fates, including the synthesis of GSH. CDO1 metabolizes CYS to CSA, which is then decarboxylated to HTAU ( Figure 4A). CDO1 WT and CDO1 Y157F -expressing A549 cells were fed fresh (CYS) 2 -containing media and sulfur containing metabolites were quantified over time ( Figures 4B-D). CDO1 WT , but not the enzyme-inactive CDO1 Y157F , limited intracellular CYS levels and promoted the accumulation of CSA, which peaked at 4 hours ( Figure 4B). Interestingly, unlike what was observed in MEFs, the levels of GSH, HTAU, and TAU were not changed over the time course of this assay ( Figure 4C). We also interrogated metabolite changes in the media and observed that (CYS) 2 was rapidly depleted from the media by 24 hours, while CSA steadily accumulated ( Figure 4D). Based on this time course, 4 hours was selected for all subsequent experiments to prevent (CYS) 2 starvation by CDO1. To examine the NRF2dependence of these metabolite alterations, we utilized the NRF2 KO  Consistently, NRF2 depletion by KEAP1 WT expression in KEAP1 MUT cells inhibited the CDO1-dependent production of CSA ( Figure 4I-K). Interestingly, KEAP1 WT expression did robustly affect CDO1 protein levels in H322 cells but significantly impaired CSA production by CDO1. Intracellular CYS can also influence CDO1 activity promoting its catalytic efficiency (Dominy et al., 2008), which may explain these results. Collectively, these results demonstrate that CDO1 expression promotes the production of CSA from CYS, leading to CSA accumulation both intracellularly and extracellularly, and enhanced (CYS) 2 consumption.

CDO1 restoration in NSCLC cells promotes sulfite production, thereby depleting cystine via sulfitolysis
We found that CSA only accounted for a fraction of CDO1-dependent (CYS) 2 depletion ( Figure S4), suggesting that CSA is metabolized to an alternative product in NSCLC cell lines. To further characterize the consequence of CDO1 expression on CYS metabolism, we performed untargeted metabolomics and found significant accumulation of sulfite (SO 3 2-) in CDO1-expressing cells ( Figure 5A). Importantly, CSA can be transaminated by the cytosolic aspartate aminotransferase (GOT1) to produce sulfinyl pyruvate, which spontaneously decomposes to pyruvate and SO 3  production may also contribute to (CYS) 2 depletion through disulfide cleavage, also known as sulfitolysis.
We next performed a quantitative analysis of SO 3 2 levels following CDO1 expression.
CDO1 WT -, but not CDO1 Y157F -expressing A549s demonstrated rapid accumulation of extracellular SO 3 2over the 24-hour time course following media replenishment ( Figure   5C). Further, the accumulation of the product of the sulfitolysis reaction, cysteine-Ssulfate (CYS-SO 3 -), was also observed in the media of CDO1 WT -expressing cells ( Figure   5D). We observed that CYS-SO 3 appeared earlier than SO 3 2-, and stopped accumulating once (CYS) 2 levels were depleted, suggesting that SO 3 2reacted with (CYS) 2 in a rapid and complete manner. To test this possibility, we incubated either CSA or sodium sulfite (Na 2 SO 3 ) with culture media in the absence of cells, and observed rapid and robust conversion of (CYS) 2 to CYS-SO 3 by Na 2 SO 3 ( Figure 5E), but not CSA ( Figure S5B), within 5 minutes. To evaluate whether SO 3 2production was a consequence of our CDO1 overexpression system, we transduced KEAP1 WT and KEAP1 R554Q MEFs with our inducible CDO1 vectors ( Figure S5C). Consistent with its regulation by intracellular CYS, ectopic CDO1 expression was significantly higher in KEAP1 R554Q MEFs compared to KEAP1 WT MEFs. While CDO1 overexpression promoted the accumulation of intracellular CSA and HTAU, and the depletion of CYS and GSH ( Figure S5D-I), we did not observe the production of SO 3 2or CYS-SO 3 in MEFs (data not shown). Interestingly, MEFs express lower GOT1 but higher CSAD protein compared to A549 cells ( Figure S5C), and NSCLC cell lines were uniformly low for CSAD and high for GOT1 ( Figure S5J) suggesting that expression of CSA metabolic enzymes may be a key determining factor in the generation of hypotaurine vs. sulfite.
These results demonstrate that SO 3 2is generated downstream of CDO1 and rapidly reacts with (CYS) 2 , thereby depleting (CYS) 2 from the culture media.
To examine the NRF2-dependence of (CYS) 2 depletion via SO 3 2-, we utilized the NRF2 KO cells, KEAP1 WT , and KEAP1 MUT cell lines. NRF2 promoted CYS-SO 3 production following CDO1 WT expression ( Figure 5F), which was accompanied by significant accumulation of both intracellular and extracellular SO 3 2- (Figure S5K-L). These findings were recapitulated in KEAP1 WT and KEAP1 MUT NSCLC cell lines expressing NRF2 T80K or following KEAP1 restoration, respectively ( Figure 5G-K), with the exception of H322, which maintained CYS-SO 3 2production following KEAP1 restoration ( Figure 5K). While KEAP1 restoration in these cells significantly reduced CSA production ( Figure 4K), unlike A549 and H1944, intracellular CSA levels in H322 cells were still in the millimolar range ( Figure S5M-O), suggesting that CSA transamination by GOT1 was saturated.
Next, we examined the toxicity of CDO1 products to NSCLC cell lines. Treatment of cells with CSA and Na 2 SO 3 , but not HTAU, led to cytotoxicity ( Figure S6A). We found that (CYS) 2 starvation and Na 2 SO 3 treatment were universally toxic to NSCLC cell lines, which did not depend on KEAP1 mutation status ( Figure S6B-C). In addition, CDO1, CSA and Na 2 SO 3 sensitized A549 cells to oxidative stress ( Figure S6D-E), consistent with their ability to deplete (CYS) 2 . Collectively, these results indicate that NRF2 induces CDO1-mediated sulfitolysis, thereby depleting extracellular (CYS) 2 in NSCLC cells.

Sulfitolysis is not required for the inhibition of proliferation by CDO1
To evaluate the role of GOT1 in CDO1-dependent sulfitolysis and cell growth inhibition, we generated GOT1 KO A549 and H460 cells ( Figure 6A). Two independent clones of each were generated, and a sgRNA-resistant GOT1 cDNA was expressed in each to restore GOT1 expression (Birsoy et al., 2015). In support of GOT1 mediating the production of SO 3 2and (CYS) 2 depletion, GOT1 KO cells had significantly lower CDO1dependent (CYS) 2 consumption and SO 3 2and CYS-SO 3 production rates compared to each parental line which was rescued by GOT1 restoration ( Figure 6B). Surprisingly, we observed that CDO1 antagonizes cell proliferation independent of GOT1 expression and sulfitolysis, suggesting that the intracellular metabolism of cysteine also contributes to the phenotype. To address this mechanism, we evaluated whether CDO1 could limit CYS-dependent processes in NSCLC cell lines, similar to what was observed in MEFs.
However, unlike in MEFs, CDO1 expression did not affect CYS utilization as the rates of GSH, Coenzyme A (CoA), and protein synthesis were similar ( Figure 6D-G).

CDO1 limits NADPH availability for cellular processes
Next, we examined other consequences of CDO1 expression in cells. After (CYS) 2 enters cells through its transporter xCT, it must be reduced to two CYS molecules using NADPH as the electron donor. As such, we hypothesized that the continual reduction of (CYS) 2 to CYS in CDO1-expressing cells would consume a significant amount of cellular NADPH. Indeed, we observed that the NADPH/NADP+ ratio is depleted following CDO1 expression in both GOT1 KO and GOT1 expressing cells ( Figure 7A).
NADPH is critical for both antioxidant defense and cellular biosynthetic processes. To examine whether the decrease in the NADPH/NADP+ ratio influenced cellular antioxidant responses, we challenged CDO1-expressing cells with the lipid peroxidation inducer cumene hydroperoxide (CuH 2 O 2 ). Expression of CDO1 increased sensitivity to CuH 2 O 2 independent of GOT1 expression ( Figure 7B). Next, we placed cells into detached conditions, which has been shown to increase reliance on IDH1-dependent reductive carboxylation to promote NADPH generation in the mitochondria (Jiang et al., 2016). Consistently, CDO1 expression significantly impaired the ability of NSCLC cell lines to grow in soft agar ( Figure S7). Next, we examined the consequence of the altered NADPH/NADP+ ratio on NADPH-dependent metabolic reactions. Using 13 C 5glutamine tracing, we observed that glutamine readily entered the TCA cycle to produce M+4 citrate, which was unaffected or increased following CDO1 expression, but CDO1 impaired both NADPH-dependent synthesis of proline from glutamate ( Figure 7C), and NADPH-dependent reductive carboxylation of α-ketoglutarate to produce M+5 citrate ( Figure 7D-E). Collectively, these results suggest that CDO1 further inhibits cellular processes by limiting NADPH availability, thereby impairing cellular proliferation.

Discussion
The carbon, nitrogen and sulfur molecules of CYS are used for diverse cellular processes that are required for both homeostasis and proliferation. The carbon, nitrogen and sulfur atoms are incorporated into protein, glutathione, taurine and CoA. Further, the sulfur atom of cysteine is incorporated into iron-sulfur (Fe-S) clusters. CYS is generally thought to be more limiting than glycine or glutamate for GSH synthesis in ). This mechanism of regulation prevents toxicity associated with CYS accumulation. However, the contribution of this regulatory process to CYS availability in cancer was not well understood.
Our findings implicate CDO1 as a metabolic liability for lung tumor cells with high intracellular cysteine levels, particularly those with NRF2/KEAP1 mutations ( Figure 7F).
We find that high intracellular cysteine levels are a common feature of lung cancer cell lines, suggesting that NRF2-independent mechanisms exist to promote cystine/cysteine uptake. Indeed, cystine uptake is regulated by many signaling pathways, including EGFR, mTORC2 and p53 ( . By contrast, CDO1 depleted intracellular CYS but did not limit GSH synthesis or other CYS-dependent processes in NSCLC cell lines. One potential explanation for this difference is that cancer cells are more efficient at maintaining intracellular CYS levels and/or GSH synthesis. While more efficient maintenance of intracellular CYS is an advantage for CYS-dependent metabolism, it is also a liability because considerable resources in the form of reducing power must be committed to continually reduce (CYS) 2 to replenish the CYS pool. Although we do not know the rate of NADPH production in our cell lines, comparison with NADPH production rates in other lines (Fan et al., 2014) suggest that CDO1-dependent (CYS) 2 consumption would consume a significant fraction of the cellular NADPH produced, consistent with our observations.
Interestingly, we find that CSA is differentially metabolized in cancer cells and mouse embryonic fibroblasts, which correlates with differential expression of CSAD and GOT1.
Partitioning between the decarboxylation and transamination reactions is likely influenced by the levels and activities of CSAD and GOT1, but this is not well studied.     For E-K cells were treated with 0.25µg/ml doxycycline for 2 days prior to and during the assay and fresh media was added 4 hr prior to sample collection.        applied as a stationary phase. The mobile phase A was 0.1 % formic acid in water, and the mobile phase B was 100% Acetonitrile (ACN). The column temperature was set to 30°C, and the gradient elution was conducted as following at a flow rate of 0.35 mL/min: 0 to 3 min, 0 % of phase A; 3 to 13 min, linear gradient from 0% to 80% of phase A; 13 to 16 min, 80% of phase A. The MS scan was operated in negative mode and the mass scan range was 58 to 870 m/z. The FT resolution was 120,000, and the AGC target was 3 × 10 6 . The capillary temperature was 320°C, and the capillary voltage was 3.5 kV.
The injection volume was 5µL. The MS peak extraction, the chromatographic peak extraction and deconvolution, the peak alignment between samples, gap filling, putative peak identification, and peak table exportation for untargeted metabolomics were    Soft agar assays. Soft agar assays were performed in triplicate in 6-well dishes. A 1 mL base layer of 0.8% agar in RPMI was plated and allowed to solidify, then 5,000 cells/well were plated in 0.4% agar on top. The following day, 1mL of RPMI was added to each well, and changed as needed. Colonies were allowed to form for 10-14 days, and wells were stained with 0.01% crystal violet in a 4%PFA/PBS solution. Plates were scanned on a flatbed scanner and colonies quantified with Image J.
Statistical analysis. Data were analyzed using a two-sided unpaired Student's t test.
GraphPad Prism 7 software was used for all statistical analyses, and values of p < 0.05 were considered statistically significant (*P < 0.05; **P < 0.01; ***P < 0.001, **** P < 0.0001. In some figures with multiple comparisons # is used in addition to *). The mean ± standard error of experiments performed in at least triplicate is reported. Similar variances between groups were observed for all experiments. Normal distribution of samples was not determined. demonstrates high basal expression, in agreement with the western blot from Figure   3C.

Supplemental information titles and legends
(C) Correlation between CDO1 protein levels (from Figure 3D) and CDO1 mRNA expression in the same cell lines.