The Combination of Loss of ALDH1L1 Function and Phenformin Treatment Decreases Tumor Growth in KRAS-Driven Lung Cancer

Lung adenocarcinoma cells express high levels of ALDH1L1, an enzyme of the one-carbon pathway that catalyzes the conversion of 10-formyltetrahydrofolate into tetrahydrofolate and NAD(P)H. In this study, we evaluated the potential of ALDH1L1 as a therapeutic target by deleting the Aldh1l1 gene in KrasLA2 mice, a model of spontaneous non-small cell lung cancer (NSCLC). Reporter assays revealed KRAS-mediated upregulation of the ALDH1L1 promoter in human NSCLC cells. Aldh1l1−/− mice exhibited a normal phenotype, with a 10% decrease in Kras-driven lung tumorigenesis. By contrast, the inhibition of oxidative phosphorylation inhibition using phenformin in Aldh1l1−/−; KrasLA2 mice dramatically decreased the number of tumor nodules and tumor area by up to 50%. Furthermore, combined treatment with pan-ALDH inhibitor and phenformin showed a decreased number and area of lung tumors by 70% in the KrasLA2 lung cancer model. Consistent with this, previous work showed that the combination of ALDH1L1 knockdown and phenformin treatment decreased ATP production by as much as 70% in NSCLS cell lines. Taken together, these results suggest that the combined inhibition of ALDH activity and oxidative phosphorylation represents a promising therapeutic strategy for NSCLC.


Introduction
Although KRAS is the gene most frequently mutated in lung adenocarcinoma, an effective KRAS-targeted therapy has not yet been developed [1,2]. Developing drugs to target KRAS mutations was known as "undruggable". For years, alternative approaches have developed targeted agents affecting the signaling cascades downstream of RAS, such as MAPK and PI3K pathways. BRAF inhibitors (vemurafenib and dabrafenib) [3] and dual specificity MEK1/MEK2 inhibitors (trametinib and combimetinib) [3] have been approved as single agents. Recent promising advances in targeting KRAS G12C, such as AMG510 [4], have increased hope for approval in clinical trials as a clinical agent.
Oncogenic KRAS promotes cellular survival, proliferation, migration, autophagy, anabolic metabolism, and changes in the microenvironment [5,6]. The metabolic flux into the non-oxidative pentose phosphate pathway increases nucleic acid biosynthesis and activates hexosamine biosynthesis and the glycolytic pathway [7]. Bioinformatics analysis of metabolic enzymes in non-small cell lung cancer (NSCLC) revealed upregulation of aldehyde dehydrogenase (ALDH) isoforms including showed that NSCLC cancer patients showed a higher expression level than normal control showed [10]. In folate metabolism, ALDH1L1 (10-formyltetrahydrofolate dehydrogenase; EC 1.5.1.6), one of the most abundant folate-binding proteins [11], plays a role in converting 10-formyl-tetrahydrofolate (THF) to THF and CO2 with production of NAD(P)H, resulting in reduced purine synthesis in normal cells [12]. However, when mitochondrial folate metabolism is dysregulated, the cytosolic folate pathway in which ALDH1L1 participates can support purine synthesis and proliferation in the opposite manner [12]. Formation of THF through oxidation of 10-formyl-THF, catalyzed by ALDH1L1, contributes to recycling of THF for purine synthesis [13]. Folate metabolism is an important metabolic pathway that produces one-carbon units for nucleic acid synthesis [14]. In addition, ~50% of NAD(P)H production in cancer cells depends on the 10-formyl-THF-pathway [15]. Hence, we investigated whether KRAS induces ALDH1L1 to promote tumor growth. Subsequently, we assessed whether ALDH1L1 could have the potential to be a therapeutic target by analyzing the effect of Aldh1l1 deletion in Kras LA2 mice, a surrogate model of human NSCLC.

KRAS Response Region in the ALDH1L1 Promoter
To determine whether ALDH1L1 transcription is regulated by KRAS, we constructed a reporter vector for measuring ALDH1L1 promoter activity using the dual-luciferase system ( Figure 2a). A549, H23, H460, and HOP-62 cells stably expressing the ALDH1L1 promoter reporter were transfected with KRAS siRNA. ALDH1L1 promoter activity was decreased~35% by KRAS knock-down (Figure 2b), whereas when KRAS was over-expressed, ALDH1L1 promoter activity increased~1.15-and 1.21-fold in H1299 and H1975, respectively (Figure 2c). An increase in ALDH1L1 expression as the result of KRAS wild type overexpression is not very relevant because KRAS is not transcription factor that direct affect gene expression. However, KRAS triggers signaling cascades including Raf-MEK-Erk signaling for Myc and Max transcription factor activation or Rho-Rac-JNK for Elk1 transcription factor activation [16]. Therefore, the knockdown of KRAS showed stronger effect of ALDH1L1 expression compared to over expression of KRAS in cancer cells, because all transcription factors, as well as downstream signaling molecules, may be required to increase the transcription activity of ALDH1L1.

KRAS Response Region in the ALDH1L1 Promoter
To determine whether ALDH1L1 transcription is regulated by KRAS, we constructed a reporter vector for measuring ALDH1L1 promoter activity using the dual-luciferase system ( Figure 2a). A549, H23, H460, and HOP-62 cells stably expressing the ALDH1L1 promoter reporter were transfected with KRAS siRNA. ALDH1L1 promoter activity was decreased ~35% by KRAS knock-down ( Figure  2b), whereas when KRAS was over-expressed, ALDH1L1 promoter activity increased ~1.15-and 1.21fold in H1299 and H1975, respectively (Figure 2c). An increase in ALDH1L1 expression as the result of KRAS wild type overexpression is not very relevant because KRAS is not transcription factor that direct affect gene expression. However, KRAS triggers signaling cascades including Raf-MEK-Erk signaling for Myc and Max transcription factor activation or Rho-Rac-JNK for Elk1 transcription factor activation [16]. Therefore, the knockdown of KRAS showed stronger effect of ALDH1L1 expression compared to over expression of KRAS in cancer cells, because all transcription factors, as well as downstream signaling molecules, may be required to increase the transcription activity of ALDH1L1.  Next, we performed an in silico analysis to predict which transcription factors acted downstream of KRAS to induce expression of the ALDH1L1. Previous studies reported that KRAS activates ELK1, ETS1, SP1, SP3, and RREB1 [17]. Based on our analysis, we identified possible SP1 and ELK1 binding sites in the ALDH1L1 promoter region (Figure 2d), suggesting that the regulation of ALDH1L1 by KRAS is mediated by activation of SP1 and ELK1. ALDH1L1 expression was highly increased in tumor nodules of the lung tissue from the Kras LA2 murine mouse model (Figure 2e).

Generation of Aldh1l1-Deficient Mice
To investigate the role of ALDH1L1 in vivo, we generated Aldh1l1-knockout mice. Mice with a knockout one allele (tm1a) of Aldh1l1 was received from IMPC (International Mouse Phenotyping Consortium). The Aldh1l1 knockout allele was generated by crossing the Aldh1l1 tm1a allele with a Cre deleter strain (zp3-Cre, Jackson laboratory strain 003651). After Cre-loxP recombination, exons 3 and 4 of Aldh1l1 were deleted ( Figure 3a). The mutant alleles could be transmitted from both male and female Aldh1l1 heterozygous mice, and homozygous mice could be obtained by interbreeding the heterozygotes. Aldh1l1 heterozygous and homozygous progeny were identified by PCR genotyping (Figure 3b). As a result of the deletion of exons 3 and 4, only the first 42 of the 902 amino acids of ALDH1L1 were correctly translated, and normal full-length ALDH1L1 protein could not be detected in knockout mouse tissues ( Figure 3c). Homozygotes did not exhibit abnormal phenotypes in most tissues, including muscle, respiratory, immune/hematopoietic, neurological/nervous, and reproductive tissues, consistent with previously reported phenotypic characterization of this mutant (https://www. mousephenotype.org/data/genes/MGI:1340024).
Next, we performed an in silico analysis to predict which transcription factors acted downstream of KRAS to induce expression of the ALDH1L1. Previous studies reported that KRAS activates ELK1, ETS1, SP1, SP3, and RREB1 [17]. Based on our analysis, we identified possible SP1 and ELK1 binding sites in the ALDH1L1 promoter region (Figure 2d), suggesting that the regulation of ALDH1L1 by KRAS is mediated by activation of SP1 and ELK1. ALDH1L1 expression was highly increased in tumor nodules of the lung tissue from the Kras LA2 murine mouse model ( Figure 2e).

Generation of Aldh1l1-Deficient Mice
To investigate the role of ALDH1L1 in vivo, we generated Aldh1l1-knockout mice. Mice with a knockout one allele (tm1a) of Aldh1l1 was received from IMPC (International Mouse Phenotyping Consortium). The Aldh1l1 knockout allele was generated by crossing the Aldh1l1 tm1a allele with a Cre deleter strain (zp3-Cre, Jackson laboratory strain 003651). After Cre-loxP recombination, exons 3 and 4 of Aldh1l1 were deleted ( Figure 3a). The mutant alleles could be transmitted from both male and female Aldh1l1 heterozygous mice, and homozygous mice could be obtained by interbreeding the heterozygotes. Aldh1l1 heterozygous and homozygous progeny were identified by PCR genotyping (Figure 3b). As a result of the deletion of exons 3 and 4, only the first 42 of the 902 amino acids of ALDH1L1 were correctly translated, and normal full-length ALDH1L1 protein could not be detected in knockout mouse tissues ( Figure 3c). Homozygotes did not exhibit abnormal phenotypes in most tissues, including muscle, respiratory, immune/hematopoietic, neurological/nervous, and reproductive tissues, consistent with previously reported phenotypic characterization of this mutant (https://www.mousephenotype.org/data/genes/MGI:1340024).

The Combination of Aldh1l1 Deficiency and Phenformin Treatment Suppresses KRAS-Driven Lung Tumorigenesis
For subsequent experiments, we used the Kras LA2 murine lung cancer model, which harbors a targeted, latent Kras G12D allele that is activated by recombination [18]. The in vivo recombination event generates in an activated allele (Kras G12D ) that induces varying grades of tumors, from hyperplasia to carcinomas similar to human NSCLC ( Figure 4). With regard to histopathology, Kras LA2 mice developed tumor progression and increased the number of nodules with age. Mice showed, predominantly, hyperplasia and adenoma, up to 32 weeks of age. As the tumor progressed, the hyperplasia ratio in nodules was decreased and the adenoma ratio in nodules was increased ( Figure  4).

The Combination of Aldh1l1 Deficiency and Phenformin Treatment Suppresses KRAS-Driven Lung Tumorigenesis
For subsequent experiments, we used the Kras LA2 murine lung cancer model, which harbors a targeted, latent Kras G12D allele that is activated by recombination [18]. The in vivo recombination event generates in an activated allele (Kras G12D ) that induces varying grades of tumors, from hyperplasia to carcinomas similar to human NSCLC ( Figure 4). With regard to histopathology, Kras LA2 mice developed tumor progression and increased the number of nodules with age. Mice showed, predominantly, hyperplasia and adenoma, up to 32 weeks of age. As the tumor progressed, the hyperplasia ratio in nodules was decreased and the adenoma ratio in nodules was increased ( Figure 4). To explore the role of ALDH1L1 in tumorigenesis, we genetically targeted the Aldh1l1 gene as described above, and then crossed Kras LA2 and Aldh1l1 −/− mice to generate Kras LA2 ; Aldh1l1 +/+ and Kras LA2 ;Aldh1l1 −/− mice ( Figure 5). The drug was administered once a day, 5 days a week, for 10 weeks from 6 to 16 weeks after birth of Kras LA2 ;Aldh1l1 −/− mice. The phenotype of hyperplasia and adenoma were differentially observed between groups (Figure 5b-d). Cytokeratin 19 (CK-19) was studied as a diagnostic marker of adenocarcinoma, which is known as a marker of endometrioid adenocarcinomas, pancreatic adenocarcinoma and head and neck cancer [19] (Figure 5d). Number of nodule was decreased about 53% in Kras LA2 ;Aldh1l1 −/− mice with phenformin treatment, while that was decreased about 9% in Kras LA2 ;Aldh1l1 −/− mice (Figure 5e). Tumor area was also reduced about 54% in Kras LA2 ;Aldh1l1 −/− mice with phenformin treatment while that was decreased about 14% in Kras LA2 ;Aldh1l1 −/− mice (Figure 5f). During the experiment, we did not observe the weight loss of the mice (Figure 5g). To explore the role of ALDH1L1 in tumorigenesis, we genetically targeted the Aldh1l1 gene as described above, and then crossed Kras LA2 and Aldh1l1 −/− mice to generate Kras LA2 ; Aldh1l1 +/+ and Kras LA2 ; Aldh1l1 −/− mice ( Figure 5). The drug was administered once a day, 5 days a week, for 10 weeks from 6 to 16 weeks after birth of Kras LA2 ; Aldh1l1 −/− mice. The phenotype of hyperplasia and adenoma were differentially observed between groups (Figure 5b-d).  was studied as a diagnostic marker of adenocarcinoma, which is known as a marker of endometrioid adenocarcinomas, pancreatic adenocarcinoma and head and neck cancer [19] (Figure 5d). Number of nodule was decreased about 53% in Kras LA2 ; Aldh1l1 −/− mice with phenformin treatment, while that was decreased about 9% in Kras LA2 ; Aldh1l1 −/− mice (Figure 5e). Tumor area was also reduced about 54% in Kras LA2 ; Aldh1l1 −/− mice with phenformin treatment while that was decreased about 14% in Kras LA2 ; Aldh1l1 −/− mice (Figure 5f). During the experiment, we did not observe the weight loss of the mice (Figure 5g).

Treatment with Gossypol and Phenformin Suppresses KRAS-Driven Lung Tumorigenesis
Previously, we have shown that treatment of an NSCLC mouse xenograft model with the alcohol dehydrogenase inhibitor gossypol or mitochondrial complex I inhibitor phenformin led to only modest tumor regression, but combined treatment with both drugs led to marked synergistic tumor regression [10]. To test the therapeutic effects of gossypol and phenformin, we used Kras LA2 mice, Tumor area (expressed as %) was calculated by dividing the total tumor area by the total area of the lung. p-values were obtained by one-way ANOVA test and are indicated by asterisks (* p < 0.05; ** p < 0.01; *** p < 0.001). (g) Body weight was measured once a week (n = 7).

Treatment with Gossypol and Phenformin Suppresses KRAS-Driven Lung Tumorigenesis
Previously, we have shown that treatment of an NSCLC mouse xenograft model with the alcohol dehydrogenase inhibitor gossypol or mitochondrial complex I inhibitor phenformin led to only modest tumor regression, but combined treatment with both drugs led to marked synergistic tumor regression [10]. To test the therapeutic effects of gossypol and phenformin, we used Kras LA2 mice, which are commonly used as a model of human NSCLC. Oral administration of vehicle or gossypol (40 mg/kg) combined with phenformin (100 mg/kg) five days per week was initiated when Kras LA2 mice were at 6 weeks old and continued until the age of 16 weeks (Figure 6a). After the mice were sacrificed at the indicated times, the area and the number of tumor nodules were lower in the combined treatment group than in the control group. Sixteen-week-old Kras LA2 mice exhibited hyperplasia in both the control and combined treatment groups; adenomas were detected in these animals, but adenocarcinoma lesions were not observed (Figure 6b,c). Immunohistochemical staining of cytokeratin 19 (CK-19) revealed a dramatic increase in tumor nodules, followed by tumor progression, in the Kras LA2 control group (Figure 6d). Compared with Kras LA2 littermates, the combined treatment group developed significantly fewer visible lung tumors from eight weeks after birth. At 16 weeks, the ratio of tumor area to normal area in the combined treatment group decreased to about one third of that the non-treated control group (Figure 6e,f), as did the number of nodules (Figure 6e,f). During the experiment, we did not observe any weight loss in the mice (Figure 6g). Together, these results demonstrate that combined treatment with gossypol and phenformin reduces lung tumor development in vivo. Oncogenic KRAS plays a key role in controlling tumor metabolism, by changing multiple metabolic pathways to give a favor to cancer cells, including the stimulation of glucose uptake, differential channeling of glucose intermediates, reprogrammed glutamine metabolism, increased autophagy, and micropinocytosis [20]. Cells require one-carbon units for nucleotide synthesis, methylation and reductive metabolism, and these pathways support the high proliferative rate of cancer cells [21]. In this study, KRAS also induces ALDH1L1 in the one-carbon pathway.

Discussion
In cancer cells, KRAS mutation induces transcription of genes encoding key enzymes of anabolic glucose metabolism, including glucose transporter 1, hexokinases, phosphofructokinase 1, and lactate dehydrogenase A [7]. KRAS mutation also promotes ribose biosynthesis through the non-oxidative pentose phosphate pathway (PPP), thereby decoupling ribose biogenesis without affecting cellular redox balance (NADP/NADPH ratio) [22]. One glycolytic intermediate, 3-phosphoglycerate, drives carbon into serine synthesis. Serine is required in biosynthesis of other molecules, such as glycine, cysteine, glutathione, and sphingolipids. Serine is also a major donor of one-carbon units to the folate metabolism cycle [12]. Thus, the induction of ALDH1L1 by KRAS promotes the one-carbon pathway in NSCLC (Figure 7).
Oncogenic KRAS plays a key role in controlling tumor metabolism, by changing multiple metabolic pathways to give a favor to cancer cells, including the stimulation of glucose uptake, differential channeling of glucose intermediates, reprogrammed glutamine metabolism, increased autophagy, and micropinocytosis [20]. Cells require one-carbon units for nucleotide synthesis, methylation and reductive metabolism, and these pathways support the high proliferative rate of cancer cells [21]. In this study, KRAS also induces ALDH1L1 in the one-carbon pathway.

Discussion
In cancer cells, KRAS mutation induces transcription of genes encoding key enzymes of anabolic glucose metabolism, including glucose transporter 1, hexokinases, phosphofructokinase 1, and lactate dehydrogenase A [7]. KRAS mutation also promotes ribose biosynthesis through the nonoxidative pentose phosphate pathway (PPP), thereby decoupling ribose biogenesis without affecting cellular redox balance (NADP/NADPH ratio) [22]. One glycolytic intermediate, 3-phosphoglycerate, drives carbon into serine synthesis. Serine is required in biosynthesis of other molecules, such as glycine, cysteine, glutathione, and sphingolipids. Serine is also a major donor of one-carbon units to the folate metabolism cycle [12]. Thus, the induction of ALDH1L1 by KRAS promotes the one-carbon pathway in NSCLC (Figure 7). Figure 7. KRAS promotes glycolysis, protein and lipid synthesis and one carbon pathway [23]. Induction of ALDH1L1 by KRAS promotes the one-carbon pathway in NSCLC. Gossypol and phenformin reduces NADH and ATP production, respectively.

ALDH1L1
(10-formyltetrahydrofolate dehydrogenase, EC 1.5.1.6) converts 10formyltetrahydrofolate (10-formyl-THF) to tetrahydrofolate (THF) and CO2 in an NADP + -dependent reaction [24]. The ALDH1L1 protein is the product of a natural fusion of three unrelated genes and consequently consists of three distinct domains: formyl dehydrogenase, 10-formyl-THF hydrolase, and ALDH [24]. Like other ALDH isotypes, ALDH1L1 performs the aldehyde dehydrogenase reaction using NADP + or NAD + , although the Km for NAD + is three orders of magnitude higher [25]. However, in the cytosol, the NADP + /NADPH ratio [26] is as much as three orders of magnitude lower than the NAD + /NADH ratio [27], because NADPH is abundantly supplied for anabolism, whereas NADH is rapidly oxidized to NAD + for catabolism. Exact measurements of NADH or NADPH Figure 7. KRAS promotes glycolysis, protein and lipid synthesis and one carbon pathway [23]. Induction of ALDH1L1 by KRAS promotes the one-carbon pathway in NSCLC. Gossypol and phenformin reduces NADH and ATP production, respectively.

ALDH1L1
(10-formyltetrahydrofolate dehydrogenase, EC 1.5.1.6) converts 10-formyltetrahydrofolate (10-formyl-THF) to tetrahydrofolate (THF) and CO 2 in an NADP + -dependent reaction [24]. The ALDH1L1 protein is the product of a natural fusion of three unrelated genes and consequently consists of three distinct domains: formyl dehydrogenase, 10-formyl-THF hydrolase, and ALDH [24]. Like other ALDH isotypes, ALDH1L1 performs the aldehyde dehydrogenase reaction using NADP + or NAD + , although the K m for NAD + is three orders of magnitude higher [25]. However, in the cytosol, the NADP + /NADPH ratio [26] is as much as three orders of magnitude lower than the NAD + /NADH ratio [27], because NADPH is abundantly supplied for anabolism, whereas NADH is rapidly oxidized to NAD + for catabolism. Exact measurements of NADH or NADPH production by ALDH1L1 have not been performed. NSCLC cells harboring an ALDH1L1 knockdown produce about 10% less NADH than wild-type cells, but no change in the NADPH level was observed [10]. Treatment Cancers 2020, 12, 1382 9 of 12 with the pan-ALDH inhibitor gossypol also decreases the NADH level by about 60% but does not affect the level of NADPH [10]. We observed no reduction of tumor growth in Aldh1l1 −/− ; Kras LA2 mice, but treatment of these mice with phenformin decreased tumor growth by~70% ( Figure 5), consistent with the 70% reduction in NADH level when ALDH1L1-knockdown NSCLC cells are treated with phenformin [10]. We also confirmed that combined treatment with gossypol and phenformin synergistically decreased the lung tumor area in Kras LA2 mice ( Figure 6). As a result of the catalytic reaction by ALDH1L1, NADH is yielded as a by-product from the conversion of 10-formyltetrahydrofolate to carbamate, which turns into ATP through oxidative phosphorylation [28]. We demonstrated that knock down of ALDH1L1 using siRNA or ALDH inhibition using gossypol induced a significant reduction of ATP production in NSCLC [10]. Anti-cancer effect of gossypol alone, however, showed about 20% reduction of tumor growth, while combination treatment of gossypol and phenformin showed about 80% reduction of NSCLC tumor growth [10]. These observations are consistent with a previous report, showing that severe depletion of ATP to levels less than 25% of control triggers cell death [29].
In summary, targeting ALDH1L1 alone did not have an anti-cancer effect in the Kras LA2 lung cancer model, whereas simultaneous inhibition of ALDH1L1 and oxidative phosphorylation significantly decreased tumor formation.

Cell Culture
Cell lines were obtained from the National Cancer Institute (NCI; MTA no. 2702-09). Growth medium was complete RPMI-1640, supplemented with 10% fetal bovine serum. Cells were maintained at 37 • C in a humidified incubator with 5% CO 2 .

Immunoblotting
Whole cell lysates were prepared using RIPA buffer (Cat. R0278, Sigma-Aldrich), with protease inhibitor cocktail (P8340, Sigma-Aldrich). Protein concentrations were determined using the Bradford protein assay (Cat. 23227, Thermo Fisher Scientific, Waltham, MA, USA). Proteins were resolved by SDS-PAGE and transferred to PVDF membranes. Membranes were blocked in 5% BSA for 1 h at room temperature (RT), and then incubated overnight at 4 • C with the indicated antibodies. Membranes were washed for 1 h at RT in TBS-T, followed by incubation with a horseradish peroxidase-conjugated secondary antibody for 1 h at RT. Finally, membranes were washed for 1 h at RT in TBS-T. Protein band images were visualized using ECL reagent (Ab frontier, Seoul, Korea) on a FUSION solo (VILBER, Collégien, France).
Firefly and Renilla luciferase activities were measured using the Dual-Glo Luciferase Assay system E2940 (Promega, Madison, WI, USA). Briefly, after treatment with KRAS siRNA for 24 h, 75 µL Dual-Glo luciferase reagent was added to each well, and the plates were incubated for 10 min at room temperature. After measurement of firefly luminescence, Dual-Glo Stop & Glo reagent was added to the plate. After incubation at room temperature for 10 min, Renilla luminescence was measured, and the ratio of firefly to Renilla luminescence was calculated.

Spontaneous Lung Cancer Model
To assess the therapeutic effects of gossypol and phenformin, we employed the Kras LA2 murine lung cancer model, which contains a targeted, latent Kras G12D allele that is activated by recombination [18]. Mice were obtained from the NCI mouse repository (strain number: 01BM3, common strain name: Kras LA2 , strain nomenclature: B6.129S-Kras tm3Tyj /Nci). Kras LA2 mice were backcrossed to C57BL/6 for at least six generations. After the weaning period (6 weeks after birth) the Kras LA2 mice (four per group) were treated with a combination of gossypol (40 mg/kg) and phenformin (100 mg/kg) for 5 days a week, until they were sacrificed at 8, 12, and 16 weeks of age. For determination of tumor incidence and grade, whole lungs were manually inflated with 10% neutral-buffered formalin, placed in fixative for 1 day, embedded in paraffin, and sectioned. H&E staining was performed by standard procedures. Lung tumor areas were determined using ImageJ. Tumor burden was expressed as the total tumor area divided by normal lung area. This study was reviewed and approved by the Institutional Animal

Automated Immunohistochemistry
Immunohistochemistry assays were performed on a VENTANA Discovery XT automated staining instrument (Ventana Medical Systems, Tucson, AZ, USA). Slides were de-paraffinized using EZ Prep solution (Ventana Medical Systems) for 30 min at 75 • C. Epitope retrieval with CC1 solution (Ventana Medical Systems) was performed for 64 min at 95 • C. Antibodies were first titered over a range of concentrations to provide the optimum ratio of specific staining to background staining. Once titers were set, antibodies were transferred with diluent to user-fillable dispensers for use on the automated stainer. Anti-ALDH1L1 (ab175198, 1:50) and cytokeratin 19 (ab52625, 1:1000) antibodies were acquired from Cell Signaling Technology (Danvers, MA, USA). Slides were developed using the OptiView DAB detection kit (Ventana Medical Systems). Briefly, samples were incubated with inhibitor for 8 min, linker for 8 min, multimer for 12 min, DAB/peroxide for 8 min, and copper for 4 min. The slides were then counterstained for 8 min with hematoxylin II (Ventana Medical Systems). Antibody titers were determined for each antibody using positive and negative control tissues, according to the manufacturer's instructions.

Conclusions
KRAS upregulates the expression of ALDH1L1 in NSCLC cells, which exhibited an increase of ALDH1L1 in Kras-driven lung cancer model. Although targeting ALDH1L1 alone did not have an anti-cancer effect in the Kras LA2 lung cancer model, simultaneous inhibition with gossypol and phenformin significantly decreased tumor formation. These results suggest that combined inhibition of ALDH activity and oxidative phosphorylation represents a promising therapeutic strategy for NSCLC.