Mutational Activation of the NRF2 Pathway Upregulates Kynureninase Resulting in Tumor Immunosuppression and Poor Outcome in Lung Adenocarcinoma

Simple Summary Activation of the Nuclear factor-erythroid factor 2-related factor 2 (NRF2) pathway through gain-of-function mutations or loss-of-function of its suppressor Kelch-like ECH-associated protein 1 (KEAP1) is frequent in lung cancer. NRF2 activation has also been reported to alter the tumor microenvironment. Proteomic profiles of 47 lung adenocarcinoma (LUAD) cell lines (11 KEAP1 mutant and 36 KEAP1 wild-type) revealed the tryptophan-kynurenine enzyme kynureninase (KYNU) as a top overexpressed protein associated with activated NRF2. Mechanistic studies demonstrated that NRF2 is a regulator of enzymatically functional KYNU in LUAD. Analysis of multiple independent gene expression datasets of human lung cancer and a LUAD tumor microarray demonstrated that elevated tumor KYNU expression was associated with immunosuppression, including potent induction of T-regulatory cells, increased levels of PD1 and PD-L1, and poorer overall survival. Our findings indicate a novel mechanism of NRF2 tumoral immunosuppression through upregulation of KYNU. Abstract Activation of the NRF2 pathway through gain-of-function mutations or loss-of-function of its suppressor KEAP1 is a frequent finding in lung cancer. NRF2 activation has been reported to alter the tumor microenvironment. Here, we demonstrated that NRF2 alters tryptophan metabolism through the kynurenine pathway that is associated with a tumor-promoting, immune suppressed microenvironment. Specifically, proteomic profiles of 47 lung adenocarcinoma (LUAD) cell lines (11 KEAP1 mutant and 36 KEAP1 wild-type) revealed the tryptophan-kynurenine enzyme kynureninase (KYNU) as a top overexpressed protein associated with activated NRF2. The siRNA-mediated knockdown of NFE2L2, the gene encoding for NRF2, or activation of the NRF2 pathway through siRNA-mediated knockdown of KEAP1 or via chemical induction with the NRF2-activator CDDO-Me confirmed that NRF2 is a regulator of KYNU expression in LUAD. Metabolomic analyses confirmed KYNU to be enzymatically functional. Analysis of multiple independent gene expression datasets of LUAD, as well as a LUAD tumor microarray demonstrated that elevated KYNU was associated with immunosuppression, including potent induction of T-regulatory cells, increased levels of PD1 and PD-L1, and resulted in poorer survival. Our findings indicate a novel mechanism of NRF2 tumoral immunosuppression through upregulation of KYNU.


Introduction
Activation of the nuclear factor erythroid 2-related factor 2 (NRF2) pathway, either through gain-of-function mutation or loss-of-function of its suppressor, Kelch-like ECHassociated protein 1 (KEAP1), is one of the most dysregulated pathways in lung adenocarcinoma (LUAD) [1,2]. NRF2 is a critical stress response mediator in mammalian cells. NRF2 is regulated by KEAP1, which binds the N-terminal Neh2 regulatory domain of NRF2, mediating its degradation via polyubiquitination and thereby inhibiting NRF2 nuclear translocation and subsequent target gene expression [3]. Nuclear translocation of NRF2 due to loss of KEAP1 expression by biallelic inactivation of the gene via mutation, loss of heterozygosity or promoter methylation has been shown to frequently occur in KRAS mutant lung adenocarcinoma [2]. NRF2 regulates basal and inducible expression of hundreds of genes that contain antioxidant response elements (AREs) in their regulatory regions by heterodimerizing with small MAF proteins. NRF2-target genes are involved in multiple cellular pathways, including those that control reduction/oxidation (redox) homeostasis, drug metabolism and excretion, energetics, amino acid metabolism, iron metabolism, and mitochondrial physiology [4]. Activation of NRF2 has been associated with induction of chemoresistance and disease progression in several cancer types [5][6][7][8][9][10][11]. Recent evidence further demonstrates a pivotal role of NRF2 activation in modulating tumor metabolism and the tumor immunophenotype [12][13][14][15]. For example, NRF2 has been shown to regulate expression of key serine/glycine biosynthesis enzymes via activating transcription factor 4 (ATF4) to support glutathione and nucleotide production [15]. Prior studies have also shown that lung cancer cells harboring KRAS mutations reprograms cancer cell metabolite towards glutamine dependence through NRF2-mediated signaling activities that increase expression of enzymes involved in glutaminolysis [16][17][18]. Nevertheless, the extent to which NRF2 activation regulates tumor metabolism and how these changes impact tumor-immune interaction remains incomplete.
Here, we performed an initial screen of 47 lung adenocarcinoma (LUAD) cell lines to identify protein signatures related to KEAP1 mutational status and activated NRF2, the results of which revealed the kynurenine-metabolizing enzyme kynureninase (KYNU) as a top differential overexpressed protein associated with activated NRF2. KYNU is a pyridoxal-5 -phosphate (pyridoxal-P)-dependent enzyme that catalyses the hydrolysis of kynurenine and 3-hydroxykynurenine into anthranilic and 3-hydroxyanthranilic acids, respectively. In mammalian cells, KYNU is involved in the biosynthesis of NAD cofactors from tryptophan through the kynurenine pathway [19]. Here, we show that the overexpression of KYNU was independent of induction of the entire tryptophan-kynurenine pathway (KP). We then evaluated the functional relevance of KYNU overexpression in LUAD and its impact on the tumor immunophenotype, showing marked immunmodulation towards a suppressive inflammatory infiltrate.

Materials and Methods
Detailed information regarding methodologies is provided in Appendix A.

RT-PCR Analysis
RNA was extracted using the RNeasy Extraction Kit (Qiagen, Germantown, MD, USA) according to manufacturer's protocol. TaqMan PCR assay was performed with a 7500 Fast Real-Time PCR System using universal TaqMan PCR master mix (ThermoFisher, Waltham, MA, USA) and FAM TM -labeled probes for KYNU (Hs00187560_m1) and KEAP1 (Hs00202227_m1) and VIC TM -labeled probes for ß2M (Hs_00187842_m1). PCR was carried out using a BioRad CFX Connect RT System (Hercules, CA, USA). Values are reported as 2 −∆∆Ct .

Proteomic Analysis
For proteomic analyses, each cancer cell line (n = 47 cell lines) was analyzed as a singular replicate. For these experiments, cancer cell lines (Table A1) were grown for seven passages in RPMI-1640 supplemented with 13 C-lysine and 10% dialyzed FBS according to the standard SILAC protocol [20]. The purpose of SILAC labeling was to discriminate the FBS derived proteins which may affect the identification of the cell surface protein list.
For siKYNU experiments, media (RPMI + 10% FBS) were collected 24 h post conditioning. Conditioned media was centrifuged at 2000× g for 10 min to remove residual debris and the supernatants were transferred and stored in −80 • C until use for metabolomics analysis.

Assessment of KP-Related Metabolites
Metabolomics analysis for KP-related metabolites was conducted on Waters Acquity™ 2D/UPLC system (Milford, MA, USA) with parallel column regeneration configuration using H-class quaternary solvent manager and I-class binary solvent manager coupled to a Xevo G2-XS quadrupole time-of-flight (qTOF) mass spectrometer. Mass spectrometry data was acquired on Xevo G2 XS qTOF in 'sensitivity' mode for positive electrospray ionization modes within a 50-1200 Da range.

Data Processing
Liquid chromatography mass spectrometry (LC-MS) and LC-tandem mass spectrometry (MSe) data were processed using Progenesis QI (Nonlinear, Waters, Milford, MA, USA) and values were reported as area units. Annotations for tryptophan, kynurenine, anthranilate and 3-hydroxyanthranilate were determined by matching accurate mass and retention times using authentic standards and by matching experimental tandem mass spectrometry data against the NIST MSMS or HMDB v3 theoretical fragmentations.

Data Normalization
To correct for injection order drift, each feature was normalized using data from repeat injections of quality control samples collected every 10 injections throughout the run sequence. Measurement data were smoothed by Locally Weighted Scatterplot Smoothing (LOESS) signal correction (QC-RLSC), as previously described [22].

TCGA and Oncomine Datasets
Gene expression data, mutational information, and clinical data from The Cancer Genome Atlas (TCGA) network project was download from cBioPortal [23] (http://www. cbioportal.org/ (accessed on: 15 June 2021)). Gene expression data and associated clinical information for the Okayama Lung Study [24] was downloaded from the Oncomine database [25].
Immunohistochemical expression of KYNU was evaluated in the cytoplasm of malignant cells using an H-score (% MCs with mild staining × 1 + % MCs with moderate staining × 2 + % MCs with strong staining × 3; range: 0-300). In this study, tumors which showed a >100 score for KYNU were considered positive for expression of the protein. The densities of cells expressing CD3, CD4, CD8, FOXP3, and PD-1 were evaluated using the Aperio nuclear algorithm and CD68 using Aperio cytoplasmic algorithm and counting the cells positive for them in five random square areas (1 mm 2 each) in both intratumoral and peritumoral compartments as described elsewhere [26]. Histologic assessment of each 1 mm 2 was performed to ensure that tumor tissue (at least 80% malignant cells and tumor stroma) was included in the selected intratumoral region, and only non-malignant cells were included in the peritumoral compartment. For this analysis, each area examined was overlapped with the sequential IHC slides to quantify each marker at the same location of the tumor specimens. The average total number of cells positive for each marker in the 5 square areas was expressed in density per mm 2 [26]. Membranous PD-L1 expression in malignant epithelial cells and macrophages was analyzed using a cell membrane staining algorithm, and the staining intensity scored as 0 (no staining), 1+ (weak staining), 2+ (moderate staining), or 3+ (strong staining) and extension (percentage) of expression were determined. The PD-L1 H-scores for tumor tissues were determined by multiplying the staining intensity and reactivity extension values (range, 0-300).

Statistics
Statistical significance was determined using Kruskal-Wallis multiple comparison tests unless otherwise specified. Spearman correlation heatmaps, Cox proportional hazard models, and construction of Kaplan-Meier survival curves were carried out in R statistical software. Significance in survival distributions was determined by Mantel-Cox log-rank t-test. For survival analyses, we used the method described by Contal and O'Quigley [27] to derive an optimal change point for KYNU expression that yielded the largest difference between individuals in the two already defined groups (alive/dead) [22,28]. Variables included into the multivariable Cox proportional hazard models were based on the backward stepwise method (likelihood ratio).
Figures were generated in either GraphPad Prism v6 or R statistical software. We note that we chose to use quartiles in comparison of the bottom 25th and top 25th percentiles of KYNU mRNA expression with tumor immunophenotype as to highlight the effect between the most differential populations; Spearman correlation analyses based on the TCGA LUAD dataset using continuous variables for KYNU mRNA expression and gene signatures of tumor immune cell infiltrates.

Analysis of the Protein Signature of KEAP1 Mutation in LUAD Cell Lines
We evaluated the proteomes of 47 LUAD (11 KEAP1 mutant (mut) and 36 KEAP1 wild-type (wt)) cell lines to identify protein signatures related to KEAP1 mutation status (Table A1 in the Appendix B). Of the 3892 quantified proteins, 296 exhibited raw p-values < 0.05 ( Figure 1A; Table S1 online). Differential analyses revealed kynureninase (KYNU), a downstream enzyme in the tryptophan-kynurenine pathway (KP), as one of the top overexpressed proteins in KEAP1mut LUAD cell lines (Figure 1A-C; Table S1 online). Other quantified enzymes in the KP were non-statistically significant between KEAP1mut and KEAP1wt LUAD cell lines, which includes the rate-limiting enzyme in de novo NAD+ biosynthesis, quinolinate phosphoribosyltransferase (QPRT), that is downstream of KYNU (Table S1 online). Analysis of gene expression data from the Cancer Cell Line Encyclopedia (CCLE) for LUAD yielded concordant findings at the RNA level ( Figure A2A,B in the Appendix B). The overall findings suggest that the KEAP1mut-associated increase in KYNU expression is independent from induction of the entire kynurenine pathway, as would be observed with increased de novo NAD+ biosynthesis. Therefore, we focused our efforts toward the mechanism and biological consequence of KYNU upregulation in LUAD.

KYNU Protein Expression Is Regulated by NRF2 Activation in Lung Adenocarcinom
KRAS, EGFR, TP53, KEAP1, and STK11 are the most prevalent mutations in and mutations in STK11 frequently co-occur with KEAP1 mutations [29,30]. Stratif of LUAD cell lines based on the occurrence of these mutations revealed a positive ation between elevated KYNU protein expression and KRASmut LUAD cell lines th bor mutations in KEAP1 and STK11 ( Figure 1C; Figure A2C,D in the Appendix B) ysis of The Cancer Genome Atlas (TCGA) LUAD gene expression dataset reve

KYNU Protein Expression Is Regulated by NRF2 Activation in Lung Adenocarcinoma
KRAS, EGFR, TP53, KEAP1, and STK11 are the most prevalent mutations in LUAD and mutations in STK11 frequently co-occur with KEAP1 mutations [29,30]. Stratification of LUAD cell lines based on the occurrence of these mutations revealed a positive association between elevated KYNU protein expression and KRASmut LUAD cell lines that harbor  Figure 1C; Figure A2C,D in the Appendix B). Analysis of The Cancer Genome Atlas (TCGA) LUAD gene expression dataset revealed a strong association between increased KYNU mRNA expression and occurrence of KEAP1 mutations ( Figure 1D; Figure A3A,B in the Appendix B).
To determine if NRF2 activation regulates KYNU expression, we performed siRNAmediated knockdown of NRF2 in KRAS/KEAP1/STK11 mutant LUAD cell lines H2030 and DFCI024, the results of which demonstrated reduced KYNU mRNA and protein expression (Figure 2A,B). No difference in KYNU protein expression was observed following siRNAmediated knockdown of STK11 ( Figure A3C in the Appendix B).
To elucidate if KYNU was enzymatically functional, we performed metabolomic analyses using mass spectrometry on sequentially collected conditioned media (baseline, 30 min, 1 h, 2 h, 4 h and 6 h) from a subset of 18 LUAD cell lines. Analyses revealed a statistically significant positive correlation between the rate (area units/hour/100 µg of protein) of KYNU-derived anthranilate accumulation into conditioned media of LUAD cell lines with whole cell lysate KYNU protein concentration (Spearman ρ = 0.68 (95% CI 0.30-0.87); 2-sided p-value = 0.002) ( Figure 2G,H, Table A3 in the Appendix B). Knockdown of KYNU in KRAS/KEAP1/STK11 mutant LUAD cell lines H2030 and DFCI024 decreased media anthranilate ( Figure 2I), thereby confirming that KYNU is enzymatically functional in LUAD cell lines.
Next, we performed immunohistochemistry for KYNU using a tissue microarray consisting of 124 LUAD tumors and compared KYNU protein levels ( Figure 3C) with CD4+, CD8+, CD3+, and FOXP3+ immune-cell infiltrates, as well as PD-1 and PD-L1 staining positivity. Elevated KYNU protein expression was significantly positively associated with high numbers of CD8+ tumor infiltrating lymphocytes (TILs) (stratified by median, twosided χ 2 test for Trend p = 0.007), as well as PD-L1 cell positivity (stratified by median, two-sided χ 2 test for Trend p = 0.02) ( Figure 3D). No statistically significant association was observed between KYNU protein expression and staining positivity for CD3+, CD4+, FOXP3+, or PD-1. Collectively, these findings revealed that elevated KYNU expression is highly associated with an immunosuppressive tumor microenvironment.   and strong positivity (>. 150)) and PD-L1 staining positivity and CD8+ TILs (E) in LUAD TMAs. PD-L1 was stratified based on whether PD-L1 was > median of the H-Score (high) or ≤ median of H-scores (low). CD8+ levels were stratified into either high (>. median of positive cells per millimeter (mm) 2 tissue) or low (≤ median of positive cells per mm 2 tissue). Statistical significance was determined by two-sided χ2 test for trend.
Next, we performed immunohistochemistry for KYNU using a tissue microarray consisting of 124 LUAD tumors and compared KYNU protein levels ( Figure 3C) with CD4+, CD8+, CD3+, and FOXP3+ immune-cell infiltrates, as well as PD-1 and PD-L1 staining positivity. Elevated KYNU protein expression was significantly positively associated with high numbers of CD8+ tumor infiltrating lymphocytes (TILs) (stratified by median, two-sided χ 2 test for Trend p = 0.007), as well as PD-L1 cell positivity (stratified by median,   and strong positivity (>150)) and PD-L1 staining positivity and CD8+ TILs (E) in LUAD TMAs. PD-L1 was stratified based on whether PD-L1 was >median of the H-Score (high) or ≤median of H-scores (low). CD8+ levels were stratified into either high (>median of positive cells per millimeter (mm) 2 tissue) or low (≤median of positive cells per mm 2 tissue). Statistical significance was determined by two-sided χ 2 test for trend.

Association between KYNU Tumor Expression and Overall Survival
We next evaluated the association between KYNU protein levels in a LUAD tissue microarray (TMA) and overall survival. In multivariate analyses, adjusted for stage, subjects with an optimal KYNU staining positivity cutoff of >120 exhibited statistically significantly worse overall survival compared to those subjects with KYNU staining positivity ≤ 120 (Hazard Ratio (HR) = 3.11 (95% CI 1.41-6.89), two-sided p = 0.005, Table 1, Kaplan-Meier survival curve in Figure 4A). A similar association between high KYNU expression and poorer overall survival was observed in LUAD gene expression datasets from Okayama lung study and TCGA ( Figure 4B,C; Tables A4 and A5 in the Appendix B. Thus, elevated tumor KYNU expression is an independent prognostic marker of poor overall survival in LUAD. two-sided χ 2 test for Trend p = 0.02) ( Figure 3D,E). No statistically significant association was observed between KYNU protein expression and staining positivity for CD3+, CD4+, FOXP3+, or PD-1. Collectively, these findings revealed that elevated KYNU expression is highly associated with an immunosuppressive tumor microenvironment.

Association between KYNU Tumor Expression and Overall Survival
We next evaluated the association between KYNU protein levels in a LUAD tissue microarray (TMA) and overall survival. In multivariate analyses, adjusted for stage, subjects with an optimal KYNU staining positivity cutoff of >120 exhibited statistically significantly worse overall survival compared to those subjects with KYNU staining positivity ≤ 120 (Hazard Ratio (HR) = 3.11 (95% CI 1.41-6.89), two-sided p = 0.005, Table 1, Kaplan-Meier survival curve in Figure 4A). A similar association between high KYNU expression and poorer overall survival was observed in LUAD gene expression datasets from Okayama lung study and TCGA ( Figure 4B

Discussion
While both activation of the NRF2 pathway and aberrant tryptophan catabolism are strongly linked to an immunosuppressive milieu in tumors, these pathways have not been directly linked [34,[37][38][39][40]. Here, we demonstrate a novel finding of NRF2-mediated KYNU upregulation in lung adenocarcinoma that is prognostic for poor overall survival. Mechanistic studies revealed that KYNU is activated by NRF2 signaling and that KYNU overexpression is associated with an immunosuppressive tumor microenvironment characterized by elevated tumor T-cell infiltration, including T regulatory TILs, and concordant increases in protein expression of immune checkpoint blockade-related PD1 and PD-L1 ( Figure 5).
A growing body of literature supports our observation that lung adenocarcinomas mutated for KRAS and with an activated NRF2 pathway have an altered tumor microenvironment that can, in part, be attributed to changes in tumor metabolism [16][17][18]. For instance, cancer cells harboring KRAS mutations have been shown to reprogram cancer cell metabolism towards an increase in uptake and catabolism of amino acids, such as glutamine and tryptophan, with pro-tumoral effects [41][42][43]. Reduced bioavailability of glutamine and tryptophan are reported to promote tumor immune suppression [35,44- A growing body of literature supports our observation that lung adenocarcinomas mutated for KRAS and with an activated NRF2 pathway have an altered tumor microenvironment that can, in part, be attributed to changes in tumor metabolism [16][17][18]. For instance, cancer cells harboring KRAS mutations have been shown to reprogram cancer cell metabolism towards an increase in uptake and catabolism of amino acids, such as glutamine and tryptophan, with pro-tumoral effects [41][42][43]. Reduced bioavailability of glutamine and tryptophan are reported to promote tumor immune suppression [35,[44][45][46][47]. For example, depletion of tryptophan bioavailability in the tumor microenvironment triggers control non-derepressible 2 (GCN2)-mediated T-cell apoptosis [48] and attenuates T-cell proliferation, whereas increased accumulation of its downstream catabolite, kynurenine, promotes immune tolerance by inhibiting proliferation of T cells and natural killer cells and increasing proliferation of Tregs and myeloid derived suppressor cells [49,50]. These insights have thus led to several clinical trials specifically targeting tumor metabolism in KRAS/NRF2 tumors. For instance, the KEAPSAKE trial (Clinicaltrials.gov; NCT04265534) evaluated addition of a glutaminase inhibitor (telaglenastat) to standard-of-care immunotherapy and chemotherapy in advanced lung cancer. Similarly, intensive investigation into the kynurenine pathway has led to the development of several inhibitors including epacodostat and indoximod, which target the rate-limiting enzymes in tryptophan metabolism IDO1, IDO2, and TDO [51]. Despite early success, a recent phase III double-blinded randomized trial of epacadostat/pembrolizumab versus pembrolizumab (ECHO-301) did not note additional benefit of epacadostat [52]. Plausible explanations for lack of additional efficacy can be attributed to insufficient inhibition of tryptophan catabolism or due to disparity in expression of IDO and TDO amongst different cancer types [20]. Additionally, other tryptophan catabolites in the kynurenine pathway, such as 3-hydroxyanthranilate, can exert immunosuppressive functions by directly inhibiting T-lymphocyte activation, promoting regulatory T cell (Treg) differentiation, and in mitigating non-antigen stimulated T-cell proliferation [39,53]. There remains interest in development of pleiotropic tryptophan pathway inhibitors or a combination of inhibitors that act on multiple enzymes within the kynurenine pathway. Our findings reported are therefore of direct relevance, by identifying a potential alternative target to attenuate immunosuppression. Further investigations exploring the targetability of KYNU as a 'immuno-metabolic' adjuvant in LUAD harboring KEAP1 mutations are thus warranted.
Elevated levels of circulating tryptophan and kynurenine-pathway related metabolites have also been reported to be associated with an increase in risk of developing lung cancer [54,55] and poor overall survival [56,57]. A recent report demonstrated that elevated the plasma levels of 3HA, a metabolite derived via the catabolism of 3-hydroxykynurenine by KYNU, was associated with significantly worse progression free survival in NSCLC subjects [58]. Notably, in this study, the combination of high tumor PDL-1 expression with elevated plasma 3HA had the highest predictive accuracy of objective response to immune checkpoint inhibitors (ICI). Here, we report that KYNU is an independent prognostic indicator of poor overall survival in lung adenocarcinoma with increased CD8+ and T-regulatory lymphocyte infiltration into tumors. It has been demonstrated that in tumor-bearing immune competent mice, administration of pharmacologically optimized PEGylated kynureninase (PEG-KYNase) promoted anti-cancer effects via increases in tumor infiltration and expansion of CD8+ lymphocytes [20]. Discrepancies may be attributed to our finding that KYNU is regulated through NRF2 activation, which is a known predictor of poor patient survival [17,59,60]. Additionally, lung cancer commonly occurs in a background of a chronically inflamed lung, which could suppress T-cell function through immune checkpoint blockade or through exhaustion. This may reflect the dynamic nature of T-cell activation and suppression in the tumor microenvironment.

Conclusions
We have identified a distinct signature of perturbed tryptophan catabolism in subsets of lung adenocarcinomas with activated NRF2 characterized by elevated KYNU expression. Protein expression of KYNU serves as a promising prognostic marker for lung adenocarcinoma and may yield capacity to identify subjects who are likely to receive benefit from ICI therapy. Further exploration of KYNU in the context of lung adenocarcinoma and immunotherapy is warranted.

Supplementary Materials:
The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/cancers14102543/s1, Table S1. Differential proteins between KEAP1mut and KEAP1wt lung adenocarcinoma cell lines; Table S2  PCR cDNA preparation was performed using an Eppendorf Thermal Cycler. Cycling conditions were 25 • C/10 min, 37 • C/120 min, 85 • C/5 min, followed by returned to 4 • C. TaqMan PCR assay was performed with a 7500 Fast Real-Time PCR System using universal TaqMan PCR master mix (ThermoFisher, Waltham, MA, USA) and FAM TM -labeled probes for KYNU (Hs00187560_m1) and KEAP1 (Hs00202227_m1) and VIC TM -labeled probes for ß2M (Hs_00187842_m1). PCR was carried out using a BioRad CFX Connect RT System. Cycling conditions were 50 • C/2 min, 95 • C/10 min followed by 40 cycles at 95 • C/15 s to 60 • C/1 min. Each sample was run in duplicate. Ct values for each gene were calculated and normalized to CT values for ß2M (∆CT). The ∆∆Ct values were then calculated by normalization to the Ct values for control. Values are reported as 2 −∆∆Ct .

Appendix A.4. Proteomic Analysis
Proteomic analyses were conducted on 47 lung cancer cell lines; each cell line was analyzed as a single replicate. To obtain whole cell extract,~2 × 10 7 cells were lysed in 1 mL of PBS containing the detergent octyl-glucoside (OG) (1% w/v) and protease inhibitors (complete protease inhibitor cocktail, Roche Diagnostics), followed by sonication and centrifugation at 20,000× g with collection of the supernatant. Protein extraction was performed in a solution containing 2% (v/v) Igepal CA-630 (Sigma-Aldrich, St. Louis, MO, USA) with cell disruption by sonication followed by centrifugation at 20,000× g. Two milligrams of whole cell extracts were reduced in dithiothreitol (DTT) and alkylated with iodoacetamide before fractionation with reverse-phase high performance liquid chromatography (RP-HPLC). A total of 84 fractions were collected at a rate of 3 fractions/minute. The mobile phase A was: H 2 O: Acetonitrile (95:5, v/v) with 0.1% trifluoracetic acid (TFA); mobile phase B was: Acetonitrile:H 2 O (95:5) with 0.1% TFA. The collected fractions from RP-HPLC were dried by lyophilization and subjected to in-solution digestion with trypsin (Trypsin Gold, Mass Spectrometry Grade, Promega). Trypsin powder was dissolved in the Digestion Buffer (100 mM amino bicarbonate, 2% Acetonitrile) to make 8 ng/µL of Trypsin solution. Then, 50 µL of Trypsin solution was added to each of the dried fractions, capped and mixed thoroughly, followed by digestion for 5 h at 37 • C. The digestion was quenched by adding 10 µL of quench solution (1%TFA in H2O). Based on the chromatogram profile, 84 fractions were grouped into 24-pool fractions for LC-MS/MS analysis by RPLC-MS/MS using a nanoflow LC system (Eksigent) coupled on-line with mass spectrometry (MS). Separations were performed using 75 µm inner diameter × 360 µm outer diameter × 25 cm long fused silica capillary column (New Objective) slurry packed in house with 5 µm, 200 Å pore size C18 silica-bonded stationary phase (Magic C18 AQ, New Objective). Tryptic peptides corresponding~2 µg amount of protein from each fraction was individually injected to on-line connected C18 trap column (Waters, 180 µm ID × 20 mm), washed for 5 min with mobile phase A (2% acetonitrile, 0.1% formic acid) at a flow rate of 10 µL/min. After the trap column desalting, peptides were eluted using a linear gradient up to 35% mobile phase B (0.1% formic acid in 95% acetonitrile) for 90 min, then increased to 95% B in 5 min, with a further additional wash with 95% B for 10 min at a constant flow rate of 300 nL/min. Eluted peptides were analyzed by LTQ Orbitrap mass spectrometer (Thermo Scientific, Waltham, MA, USA) in data dependent acquisition (DDA) mode. Each full MS scan (m/z 400-1800) was followed by 10 MS/MS scans (normalized collision energy of 35%) for the 10 most abundant precursor ions in a~1.5 s of duty cycle. Dynamic exclusion was enabled to minimize redundant selection of peptides previously selected for MS/MS analysis. Parameters for MS1 were 60,000 for resolution, 1×106 for automatic gain control (AGC) target, and 150 ms for maximum injection time. MS/MS was done by collision-induced dissociation (CID) fragmentation with 3 × 104 for AGC, 10 ms for maximum injection time, 35 for normalized collision energy (NCE), 2.0 m/z for isolation width, 0.25 for activation q-value, and 10 ms for activation time.
MS/MS spectra were searched against the Uniprot proteome database (Human and Bovine, January 2017) using the X!Tandem search engine through Trans-Proteomic Pipeline (TPP 4.8) and processed with the peptide and protein prophet. For the modifications, one fixed modification of carbamidomethylation (57.04304 Da) or propionamide (71.037114) at cysteine and two variable modifications, oxidation at methionine (15.9949 Da) and SILAC 13C6 at lysine (6.0201 Da) were chosen. The mass error allowed was 20 ppm for parent monoisotopic and 0.5 Da for MS2 fragment monoisotopic ions. Trypsin was specified as protein cleavage site, with possibility of two missed cleavages allowed. A false protein discovery rate of approximately 5% was determined by searching the primary tandem MS data using the same criteria against a decoy database wherein the protein sequences are reversed. Any stable isotope unlabeled lysine containing peptides with bovine homology were discarded. The total number of spectral counts for each protein group was normalized to the total spectral counts of the sample and subsequently multiplied by a constant which was set as 50,000; scaled data was used for subsequent downstream analysis. Measurements were retained based on spectral counts > 5 for at least one cancer cell line type, resulting in a total of 3892 proteins that were kept for statistical analysis.

. Exometabolome Experiments
Each cell line was analyzed in biological triplicate or quadruplicate unless otherwise specified. Cells were grown in 1ml of RPMI 1640 + 10% FBS in 12-well dishes (Costar, Houston, TX, USA) to reach a 70% (50-80%) confluency, 24 h post initial seeding. On the day of the experiment, the cells were washed twice with 500 µL serum free RPMI (Fisher Scientific) containing 5 mM glucose and 0.5 mM glutamine. Next, serum free RPMI (300 µL) containing 5 mM glucose and 0.5 mM glutamine was added to each well and the cells were incubated. After predetermined incubation time (1, 2, 4 and 6 h), 250 µL of the conditioned media was collected. For baseline (T0), 250 µL of media was collected directly after the addition of 300 µL. Blank samples containing media only were included and collected at T0 and T6. The 6 h samples were used to count cell numbers for data normalization. Once all the media samples were collected, the tubes were centrifuged at 2000× g for 10 min to remove residual debris and the supernatants transferred to 1.5 mL tubes (Eppendorf, Hamburg, Germany), and stored in −80 • C until use for metabolomics analysis.

Appendix A.5.2. Mass Spectrometry Data Acquisition
Frozen media samples were thawed on ice and 30 µL transferred to a 96-well microplate (Eppendorf) containing 30 µL of 100 mM ammonium formate, pH 3.0. The microplates were heat sealed, vortexed for 5min at 750 rpm, and centrifuged at 2000× g for 10 min at room temperature. For hydrophilic interaction liquid chromatography (HILIC) analysis, 25 µL of sample was transferred to a new 96-well microplate containing 75 µL acetonitrile, whereas samples for C18 analysis were transferred to a new 96-well microplate containing 75 µL water (GenPure ultrapure water system, Thermofisher, Waltham, MA, USA). Each sample solution was transferred to a 384-well microplate (Eppendorf, Hamburg, Germany) for LCMS analysis. Samples were randomized during LCMS analysis and a matrix-matched reference quality controls and batch-specific pooled quality controls were included.
Quaternary solvent system mobile phases were (A) 0.1% formic acid in water, (B) 0.1% formic acid in acetonitrile and (D) 100 mM ammonium formate, pH 3. Samples were eluted using the following gradient profile: for the HILIC analysis, a starting gradient of 95% B and 5% D was increased linearly to 70% A, 25% B and 5% D over a 5 min period at 0.4 mL/min flow rate, followed by 1 min isocratic gradient at 100% A at 0.4 mL/min flow rate. For the reverse phase C18 analysis, the chromatography gradient was as follows: starting conditions, 100% A, with linear gradient to final conditions of 5% A, 95% B followed by isocratic gradient at 95% B, 5% D for 1 min.
Mass spectrometry data was acquired on Xevo G2 XS qTOF in 'sensitivity' mode for positive electrospray ionization mode within a 50-1200 Da range. For the electrospray acquisition, the capillary voltage was set at 1. To correct for injection order drift, each feature was normalized using data from repeat injections of quality control samples collected every 10 injections throughout the run sequence. Measurement data were smoothed by Locally Weighted Scatterplot Smoothing (LOESS) signal correction (QC-RLSC) as previously described [28]. Feature values between quality control samples were interpolated by a cubic spline. Metabolite values were rescaled by using the overall median of the historical quality control peak areas across all samples. Only detected features exhibiting a relative standard deviation (RSD) less than 30% for either the historical or pooled quality controls samples were considered for further statistical analysis. Calculated rates were adjusted to total cell protein.
Appendix A.6. Immunohistochemical Analysis The immunohistochemistry used in this study comprised 124 surgically resected LUAD tumor specimens collected under an institutional review board protocol and archived as formalin-fixed, paraffin-embedded specimens in The University of Texas Specialized Program of Research Excellence thoracic tissue bank at The University of Texas MD Anderson Cancer Center. Patient characteristics for the analyzed cohort are provided in Table A1 in the Appendix B. Tumor staging was performed using the staging system from the 7th American Joint Committee on Cancer. For adenocarcinoma, predominant histologic pattern (solid, lepidic, acinar, papillary, and micropapillary) was determined according to the 2015 World Health Organization classification. TMA sections were stained in a Leica Bond Max autostainer (Leica Biosystems Nussloch GmbH, Nußloch, Germany). Four micrometer-thick tissue sections (4 µm) were deparaffinized and rehydrated following the Leica Bond protocol. Antigen retrieval was performed for 20 min with Bond Solution # 1 (Leica Biosystems, equivalent Citrate Buffer, pH6, AR9961). The slides were then incubated with the primary antibody for KYNU (E-5, Santa Cruz, sc-390360, at 1:1000 dilution), PD-L1 (clone E1L3N, dilution 1:100; Cell Signaling Technology), CD3 (T-cell lymphocytes; dilution 1:100; Dako), CD4 (helper T cell; Novocastra; clone 4B12, dilution 1:80; Leica Biosystems), CD8 (cytotoxic T cell; clone CD8/144B, dilution 1:20; Thermo Fisher Scientific), PD-1 (clone EPR4877-2, dilution 1:250; Abcam), and FOXP3 (regulatory T cell; clone 206D, dilution 1:50; BioLegend) [26] As an external positive control for KYNU staining, we used non-neoplastic liver tissue stained with KYNU antibody, but we substituted the KYNU antibody with diluent for negative control ( Figure A5 in the Appendix B).
The IHC staining was revealed using the Bond Polymer Refine detection kit (Leica Biosystems DS9800), including 3 3 -diaminobenzidine as chromogen and hematoxylin as counterstaining. All slides where cover-slipped, and then scanned in an Aperio AT2 scanner. Three tissue microarray cores from each patient were evaluated, and the percentages of staining were averaged to provide the final scoring. For the immunostaining scoring, two board-certified surgical pathologists participated in the analysis of this tissue microarray. Before the tissue microarray scoring, the two pathologists evaluated together, at a doubleheaded microscope, the immunostaining pattern for KYNU in control slides and in the tissue microarray slides to agree on the criteria for intensity and percentage of staining. After scoring, the two pathologists met again to review the tissue microarray scoring together in the double-headed microscope and resolve any question or discrepancy in the evaluation. Immunohistochemical expression of KYNU was evaluated in the cytoplasm of malignant cells using an H-score (% MCs with mild staining ×1 + % MCs with moderate staining ×2 + % MCs with strong staining ×3; range: 0-300). In this study, tumors, which showed more than 100 score for KYNU were considered positive for expression of the protein. The densities of cells expressing CD3, CD4, CD8, FOXP3 and PD-1 were evaluated using the Aperio nuclear algorithm and CD68 using Aperio cytoplasmic algorithm and counting the cells positive for them in five random square areas (1 mm 2 each) in both intratumoral and peritumoral compartments, as described elsewhere [26]. Histologic assessment of each 1 mm 2 was performed to ensure that tumor tissue (at least 80% malignant cells and tumor stroma) was included in the selected intratumoral region, and only non-malignant cells were included in the peritumoral compartment. For this analysis, each area examined was overlapped with the sequential IHC slides to quantify each marker at the same location of the tumor specimens. The average total number of cells positive for each marker in the five square areas was expressed in density per mm 2 [26]. Membranous PD-L1 expression in malignant epithelial cells and macrophages was analyzed using a cell membrane staining algorithm, and the staining intensity scored as 0 (no staining), 1+ (weak staining), 2+ (moderate staining), or 3+ (strong staining), and extension (percentage) of expression was determined. The PD-L1 H-scores for tumor tissues were determined by multiplying the staining intensity and reactivity extension values (range, 0-300).

Appendix A.7. Statistical Analysis
For survival analyses, we used the method described by Contal and O'Quigley [27] to derive an optimal cut-off point for KYNU protein staining intensity that yielded the largest difference between individuals in the two already defined groups (alive/dead) [22,28]. Using log rank statistic based on the groups defined by cutoff yielded: where D is the total number of distinct death times, d i is the total number of deaths at each event time (t i ), d + i is the total number of death when KYNU staining value is bigger than the cut-off point. r i and r + i are also defined as the total number at risk for all KYNU staining values and KYNU staining value larger than cut-off point, respectively. We calculated S k for all possible cut points in the KYNU staining column and the estimated cut point as the value that yields the maximum S k . Our analysis yielded a resultant KYNU protein staining positivity cutoff value of 120.  Table A3. Secretion or uptake rates for anthranilate and 3-hydroxyanthranilate for 18 lung adenocarcinoma cell lines.      Best et al. [13] in TCGA-LUAD tumors stratified by presence or absence of STK11 and KEAP1 mutations. (C) Immunoblots for KYNU protein expression following siRNA-mediated knockdown of STK11 in KYNU-low STK11-wild type LUAD cell lines H2009 and H3255. DFCI024 and H2030 lysates were used as positive controls for KYNU protein expression.  [13] amongst TCGA LUAD tumors. Columns are ranked (lowest to highest) by KYNU mRNA expression. (B) Spearman correlation coefficients (95% confidence interval) between KYNU mRNA expression and the gene-based NRF2 consensus signature described by Best et al. [13] in TCGA-LUAD tumors stratified by presence or absence of STK11 and KEAP1 mutations. (C) Immunoblots for KYNU protein expression following siRNA-mediated knockdown of STK11 in KYNU-low STK11-wild type LUAD cell lines H2009 and H3255. DFCI024 and H2030 lysates were used as positive controls for KYNU protein expression.