TBK1 reprograms metabolism in breast cancer: an integrated omics approach

Abstract

Metabolic rewiring is required for cancer cells to survive in harsh microenvironments and is considered to be a hallmark of cancer. Specific metabolic adaptations are required for a tumor to become invasive and metastatic. Cell division and metabolism are inherently interconnected, and several cell cycle modulators directly regulate metabolism. Here we report that TBK1, which is a non-canonical IKK kinase with known roles in cell cycle regulation and TLR signaling, affects cellular metabolism in cancer cells. While TBK1 is reported to be over-expressed in several cancers and its enhanced protein level correlates with poor prognosis, the underlying molecular mechanism involved in the tumor-promoting role of TBK1 is not fully understood. In this study, we show a novel role of TBK1 in regulating cancer cell metabolism, using combined metabolomics, transcriptomics, and pharmacological approaches. We find that TBK1 mediates regulation of nucleotide and energy metabolism through aldo-keto reductase B10 (AKRB10) and thymidine phosphorylase (TYMP) genes, suggesting that this TBK1-mediated metabolic rewiring contributes to its oncogenic function. In addition, we find that TBK1 inhibitors can act synergistically with AKRB10 and TYMP inhibitors to reduce cell viability. These findings raise the possibility that combining these inhibitors might be beneficial in combating cancers that show elevated levels of TBK1.

The mass spectrometry metabolomics data have been deposited at the MetaboLights - Study Editor with the data set identifier MTBLS11377 (www.ebi.ac.uk/metabolights/MTBLS11377).

Introduction

Metabolic rewiring is a hallmark of cancer which helps tumor cells to make maximum use of limited resources for survival. Cancer cells undergo metabolic reprogramming to ensure survival and proliferation in the acidic and hypoxic tumor microenvironment ¹. In addition, oncogene-driven metabolic cooperation between cancer and stromal cells and their ability to utilize diverse sources of energy contribute to their survival and proliferation in harsh tumor microenvironments ^2,3. Also, fatality from cancer is determined by metastasis rather than the growth at the primary tumor site; however, there are several biological challenges that cancer cells must overcome for metastasizing. These include escape from primary tumors, survival in circulation, colonization in distant organs, and growth of secondary tumors at the target sites. Specific metabolic adaptations allow the cancer cells to overcome these bottlenecks and establish metastasis. While the intravasation of cells is promoted by an acidic microenvironment, cancer cells need to counter oxidative stress by producing glutathione and NADPH for survival in circulation. Anabolic metabolism is activated for biomass accumulation to support the rapid cell proliferation required for the formation of secondary tumors ^4,5.

There is a bi-directional cross-talk between cellular metabolism and cell cycle machinery. The cells are critically dependent on the availability of the requisite metabolites to enter and complete the cell cycle, while the cell cycle regulates metabolism to ensure survival and proliferation. One of the earlier studies suggested such a correlation when a time-dependent cyclical fluctuation in levels of nucleotide, lipid, and protein metabolites as a function of cell cycle progression was observed ⁶.

TANK-binding kinase 1 (TBK1) is a non-canonical IκB kinase (IKK) with a wide range of functions reported. It is known to regulate innate immunity ^7,8, inflammation ⁹, autophagy ^{10,11,12,13,14,15} and cell death signalling ¹⁶. Recent reports, including two from our group, also showed that TBK1 is involved in regulating centrosome homeostasis and mitosis ^17,18,19. It was also reported to directly phosphorylate mitotic kinase Polo-like kinase 1 (PlK1) and its upstream regulator Protein kinase B (PKB) ¹⁷. We have previously shown that TBK1 regulates microtubule dynamics and centrosomal localization of CEP170 and NuMA both of which are essential to mitosis ¹⁸. More recently we showed that it regulates spindle assembly checkpoint (SAC), through which the cell ensures that genetic fidelity is maintained during mitosis by curbing chromosome missegregation ¹⁹. Evidence is also emerging indicating that TBK1 has metabolic roles ^9,20,21. It appears to play a role in cellular metabolism by regulating mammalian targets of rapamycin (mTOR) ²². Dysregulated TBK1 function has been reported to result in several disease states including cancer ^23–25. However, the underlying mechanism is still poorly understood. As TBK1 has roles in both mitosis and cellular metabolism and reports suggest that aberrant TBK1 functioning may result in cancer, we hypothesize that TBK1 is yet another component of the cell cycle with a role in regulating cellular metabolism. In addition, its metabolic role supports its oncogenic function. In the present study, we used an integrated transcriptomics and metabolomics approach to show that TBK1 silencing results in aberrant nucleotide and energy metabolism. Aldo keto reductase B10 (AKR1B10) and thymidine phosphorylase (TYMP) were found to be downregulated upon TBK1 silencing. Both of these enzymes have been reported to have tumor supporting roles. We find that chemical inhibition of these enzymes in combination with TBK1 has a significant combinatorial effect on cancer cell survival. In summary, we have combined metabolomics, transcriptomics and pharmacological approaches to propose TBK1 as a therapeutic target in cancer, in combination with AKR1B10 and TYMP for increased efficacy.

Materials and Methods

Cell Culture, Plasmids and Reagents

Mesenchymal, triple-negative breast cancer cell lines MDAMB-231 and Hs578T were obtained from the American Type Culture Collection (ATCC, Manassass, VA) and grown in DMEM containing 10% fetal bovine serum. Cells were treated with BX795 (#s1274, Selleckchem) to inhibit TBK1, diclofenac (#1188800, Sigma Aldrich), or flufenamic acid (#F9005, Sigma Aldrich) to inhibit AKR1B10 and tipiracil (#SML1552, Sigma Aldrich) to inhibit TYMP. Control siRNA (sc-37007) was purchased from Santa Cruz Biotechnology and TBK1 siRNA (# 4457298, ID: s761) was purchased from Ambion. siRNAs were transfected using Oligofectamine (Invitrogen Corporation) according to the manufacturer’s protocols

Gene Expression analysis using GEPIA and UALCAN database

Analysis of TBK1, TYMP, and AKR1B10 expression at mRNA and protein levels in breast tumors as compared to normal tissues was done using GEPIA (http://gepia.cancer-pku.cn/) and UALCAN databases (https://ualcan.path.uab.edu/index.html) respectively ²⁶. Gene Expression Profiling Interactive Analysis (GEPIA) is an interactive open-source web server based on gene expression data from TCGA and GTEx databases ²⁷. UALCAN is a comprehensive resource that provides access to cancer databases like TCGA (The Cancer Genome Atlas) and CAPTAC (Clinical Proteomic Tumor Analysis Consortium).

Survival analysis using cBioPortal

Survival analysis of patients with high expression of TBK1 as compared to control subjects was done using cBioPortal (https://www.cbioportal.org/) database. cBioPortal is an online analysis tool that hosts data from TCGA, TARGET as well as lab individual lab publications and can be used for the analysis of cancer genomics and clinical data ²⁸.

Lentiviral sgRNA production and infection

TBK1 guide RNA (gRNA) was cloned into the pLentiCRISPRV2 backbone vector using the BsmBI (NEB) restriction enzyme. Lentiviral vectors expressing sgRNAs specific for control sequences (5’ GACGGAGGCTAAGCGTCGCAA 3’) or for TBK1 (Forward Oligo: 5’ CACCGCAGTGATCCAGTAGCTGCA 3’, Reverse Oligo: 5’ CACCGCAGTGATCCAGTAGCTGCA 3’) were transfected into 293FT cells along with the packaging plasmids using Fugene HD (Roche). Culture supernatants containing lentivirus were collected 48 and 72 hours after transfection. The virus was pooled and stored at −80 °C. Cells (seeded at the density of 120,000 cells per well of 6 well plates) were infected using the supernatant containing the virus in polybrene (8ug/mL) containing media. The plate was then centrifuged at 800g for 1h at RT to enhance the efficacy of infection. After two rounds of infection, cells were allowed to grow for another 48 hours before puromycin addition (1–2ug/mL).

Metabolite extraction and identification

TBK1 was knocked down in the MDA-MB-231 cell line using the CRISPR-Cas9 technique. sgControl and sgTBK1 cells were harvested and metabolites in the cell were extracted with chilled 80% methanol. The Metabolomics Quality Control (QC) kit was spiked in the samples as Internal Standards. This kit contains 14 stable isotope-labeled metabolite standards, was obtained from Cambridge Isotope Labs; it includes the following labeled compounds: L-Alanine (13C3, 99%), L-Leucine (13C6, 99%), L-Phenylalanine (13C6, 99%), L-Tryptophan (13C11, 99%), L-Tyrosine (13C6, 99%), Caffeine (13C3, 99%), D-Glucose (13C6, 99%), Benzoate (13C6, 99%), Citrate (13C3, 99%), Octanoate (13C8, 99%), Propionate (13C3, 99%), Stearic acid (13C18, 98%), Succinic acid (13C4, 99%) and D-Sucrose (13C6, 98%). After incubation and centrifugation, the supernatant was removed and the protein pellets were left behind for the Bradford assay to measure protein concentration. The metabolites in the supernatant were concentrated via drying and reconstitution. Ultra-high performance liquid chromatography-mass spectrometry (UHPLC-MS) was performed using a Vanquish LC (Thermo, San Jose, CA) interfaced with a Q Exactive HF mass spectrometer (Thermo, San Jose, CA). LC method is modified from a previous publication ²⁹. Briefly, chromatographic separation was performed on a SeQuant ZIC-pHILIC LC column (Millipore Sigma, Burlington, MA) with a guard column connected to it. The mobile phase was 10 mM ammonium carbonate and 0.05% ammonium hydroxide in water (A), and 100% acetonitrile (B). The gradient program included the following steps: start at 80% B, a linear gradient from 80 to 20% B over 13 minutes, stay at 20% B for 2 minutes, return to 80% B in 0.1 minutes, and re-equilibration for 4.9 minutes for a total run time of 20 minutes. The flow rate was set to 0.250 mL/min. The total running time was 20 min. The column temperature was set to 30 °C, and the injection volume was 2 μL. The full MS was performed in positive and negative modes separately between the mass scan range of 60 to 900 m/z. The metabolites are identified by MS1. The IDs are obtained by searching the in-house library which includes both accurate mass and Retention Time for ~600 metabolites. Data analysis was carried out on MZmine 2 to identify the metabolites against the internal library.

RNA sequencing

MDA-MB-231 cell line was transfected with either scrambled or TBK1 siRNA. RNA was isolated using a Qiagen RNasy miniprep kit (#74104) as per the instructions in the manufacturer’s protocol and RNA pu48rity was checked using nanodrop. Reference genome and gene model annotation files were downloaded from the genome website browser (NCBI/UCSC/Ensembl) directly. Indexes of the reference genome were built using STAR and paired-end clean reads were aligned to the reference genome using STAR (v2.5). HTSeq v0.6.1 was used to count the read numbers mapped of each gene. Then FPKM of each gene was calculated based on the length of the gene and the read count mapped to this gene. FPKM, Reads Per Kilobase of exon model per Million mapped reads, considers the effect of sequencing depth and gene length for the reads count at the same time.

Cell Viability Assay

Cell viability was assessed using MTT (Thiazolyl Blue Tetrazolium Bromide) staining. 2500 cells were seeded in 96 well plates in triplicates. TBK1, AKR1B10, and TYMP inhibitors were added after 24 hours. 96 hours post-treatment, cells were stained by incubating with 5mg/mL MTT in PBS for 1 hour at 37 ͦ C. The formazan crystals were dissolved by incubating with 100μl DMSO on a shaker. Absorbance was measured on a plate reader at 590nm. For the IC₅₀ calculation of BX795, a concentration range of 40μM to 0.158μM was used, while keeping the dose of AKR1B10 inhibitors (diclofenac, flufenamic acid) and TYMP inhibitor tipiracil constant at 100μM. A ten-point curve was drawn using GraphPad Prism software to calculate IC₅₀ values. For IC₅₀ calculation of diclofenac, flufenamic acid, and tipiracil a concentration range of 500μM to 1.953μM was used while keeping the dose of TBK1 inhibitor BX795 constant at 100μM. Error bars represent the standard deviation (SD). All figures depict experiments that have been replicated a minimum of two times.

Synergy analysis

The drug synergy of TBK1 inhibitor (BX795) with AKR1B10 inhibitors (diclofenac, flufenamic acid) and TYMP inhibitor tipiracil was analyzed using CompuSyn 1.0. CompuSyn is a freely available software that uses the median-effect principle of the mass-action law and combination index theorem for quantitation of drug synergy ^30,31. The normalized isobologram and Fraction affected (Fa) vs Combination Index (CI) plots generated by CompuSyn were then used to support synergy.

Real-time PCR

RNA was isolated using a Qiagen RNasy miniprep kit as per the instructions in the manufacturer’s protocol. First-strand cDNA was synthesized using a Bio-Rad-iScript cDNA synthesis kit. Levels of TBK1, AKR1B10, and TYMP mRNA were analyzed using quantitative real-time PCR (CFX96 software). Data was analyzed using the Ct method. GAPDH levels were used for normalization. all samples were run in technical (assay) triplicate. Error bars represent the SD. All figures depict experiments that have been replicated a minimum of two times. Primers were designed using Primer blast as described before³². Details of primers are as follows:

AKR1B10	Forward primer	5’-CCCAAAGATGATAAAGGTAATGCC-3’
	Reverse primer	5’-TCAGTCCAGGTTTGTTCAAGAGC-3’
TYMP	Forward primer	5’-CTGGTCGACGTGGGTCAGAG-3’
	Reverse primer	5’-TCGCGGCAAAGGAGCTTTAT-3’
TBK1	Forward primer	5’-GGATCACTGCCATTTAGACCC-3’
	Reverse primer	5’-CAGGCATGTCTCCACTCCAG-3’
GAPDH	Forward primer	5’-GGTGGTCTCCTCTGACTTCAACA-3’
	Reverse primer	5’-GTTGCTGTAGCCAAATTCGTTGT-3’

Data processing and analysis

The Metabolomics dataset was normalized against total cellular protein concentration. Differentially expressed metabolites with statistical significance were identified based on fold change ≥ 1.5 or ≤ 0.67 with p-value < 0.05 (t-test). Afterward, the metabolite pathway network was constructed using GAM webservice ³³ based on the KEGG REACTION database, using the RPAIR database. Multivariate statistical analysis using SIMCA 14.0.2 was also performed to identify the most relevant metabolites responsible for class discrimination. First, principal component analysis (PCA) was used to determine outliers and classify the groups in an unbiased manner. Next, partial least square discriminant analysis (PLS-DA) was performed for variable extraction. The robustness and predictive ability of the model were evaluated using leave-one-out cross-validation. The most important variables were extracted using variance importance in projection (VIP).

The Transcriptomics dataset was subjected to one scaling normalized factor using the edgeR program package where the read counts for each sequenced library were adjusted. Differential expression analysis of two conditions was performed using the edgeR R package (3.16.5). The p values were adjusted using the Benjamini & Hochberg method. The corrected p-value of 0.05 and log2(fold change) of ±1 were set as the threshold for significantly differential expression. Gene ontology (GO) analysis was then performed with the differentially expressed genes using the PANTHER classification system. Pathway analysis against the PANTHER metabolic and cell signaling database was performed using the web-based Enrichr tool ³⁴. An Integrated metabolomics and transcriptomics analysis was performed using Metaboanalyst 6.0 ³⁵ according to the KEGG database.

Univariate statistical analysis for the MTT and Real-time PCR dataset are reported as the means ± standard deviation. Comparisons were carried out using unpaired Student’s two‐tailed t-test or one-way ANOVA. All statistical tests were carried out using PRISM version 5.03 (GraphPad Software, La Jolla, CA, USA), statistics software. P < 0.05 was accepted as significant. *p<0.05, **p<0.01 and ***p<0.001.

Results

TBK1 negatively impacts the survival of breast cancer patients

First, we wanted to explore pre-existing datasets for TBK1 transcript and protein levels in breast cancer patients. For this, mRNA expression levels were analyzed by GEPIA database (Figure 1A), and UALCAN database was used to analyze TBK1 protein levels (Figure 1B) using the Clinical Proteomic Tumor Analysis Consortium (CPTAC) and the International Cancer Proteogenome Consortium (ICPC) datasets. Results indicate highly significant over-expression of TBK1 in breast invasive carcinoma as compared to normal tissue at both protein and mRNA levels. Next, a survival analysis was performed to study the effect of TBK1 expression on the overall survival of breast cancer patients using cBioPortal database (Figure 1C). Patients with high TBK1 expression were found to have poor overall survival. TCGA showed with high confidence, a median overall survival of 152 months for patients with low TBK1 expression, while only 90 months for patients with high TBK1.

Figure 1:

TBK1 impairs the survival of breast cancer patients. (A-B) TBK1 Expression levels in breast invasive carcinoma at (A) transcript level and (B) protein level were plotted using GEPIA and UALCAN databases respectively. (C) cBioPortal database was used to construct overall survival curves based on TBK1 mRNA levels in breast cancer patients.

TBK-1 influenced metabolite alterations

The above analysis shows that high levels of TBK1 play a role in poorer prognosis in breast cancer patients raising the possibility that this occurs through the modulation of cellular functions. TBK1 has been shown to alter proliferative signaling and facilitate mitotic progression, both of which might promote oncogenesis if unregulated. In addition, TBK1 has been shown to play an important role in the regulation of cellular metabolism ^8,9,20; therefore, we next examined whether TBK1-mediated dysregulation of cancer cell metabolism plays a potentiating role in breast cancer. To determine the effect of TBK-1 on cellular metabolism, LC-MS/MS-based metabolomics analysis was performed on mesenchymal, triple-negative (lacking the estrogen, progesterone, and Her2 receptor expression) MDA-MB-231 breast cancer cells with TBK-1 knockdown compared with control. A total of 104 and 158 metabolites were identified in positive and negative modes, respectively in both cell types (Supplementary Table 1). Taking together the metabolites of both modes, volcano plot filtering revealed 83 molecules to be differentially expressed (fold change ≥ 1.5 and p ≤ 0.05). While 20 metabolites were found to be downregulated, 63 were upregulated in TBK-1 knockdown cells compared to the control cells (Figure 2A). It is worth noting that some of the metabolites detected in both positive and negative modes showed similar trends in their levels, thus validating the robustness of our findings. To name a few, cytidine 5’-diphosphate, adenosine 2’,3’-cyclic monophosphate and adenosine 5’-diphosphate showed 1.84 fold (p = 0.038), 1.52 fold (p = 0.0025) and 1.81 fold (p = 0.036) respectively in positive mode and 1.83 fold (p = 0.039), 1.55 fold (p = 0.0055) and 1.77 fold (p = 0.037) respectively in negative mode (Supplementary Table 1). Pathway analysis with the differentially expressed metabolites revealed several pathways involved in purine, pyrimidine, amino acid, fatty acid, lipid, glucose and galactose metabolism were dysregulated in TBK-1 knockdown cells (Figure 2B). To identify the most important metabolites dysregulated in response to TBK-1 silencing, multivariate statistical analysis was performed. Unsupervised PCA analysis revealed separation of the two groups of cells indicating that the metabolites were able to distinguish them in an unbiased manner (Figure 2C). Class separation further improved when supervised classification, PLS-DA was used (Figure 2D). Since supervised classification systems have an overfitting tendency several model validation parameters were assessed. R2 (goodness of fit) and Q2 (goodness of prediction) values were found to be 0.997 and 0.994, respectively. The values close to 1.0 indicated the robustness of the model. Finally, permutation test statistics showed that the R2 and Q2 values of 100 different permutated models were lower than that of the original model (Figure 2E). Once the robustness of the PLS-DA model was validated, the important metabolites responsible for class separation were extracted from the VIP. A total of 71 metabolites were found to have VIP >1.25 (Supplementary Table 1). Interestingly, all these metabolites were also found to have p ≤ 0.05 in the t-test (Supplementary Table 1). Some of these metabolites included orotate, dihydroorotate, thymine, 5,6-dihydrouracil, uracil, and deoxycytidine which were involved in pyrimidine metabolism and fructose 1,6 bisphosphate and galactitol involved in glucose and galactose metabolism.

Figure 2:

TBK-1 influenced metabolite alterations. (A) Volcano plot analysis of metabolomics dataset between sgTBK-1 and sgC groups. Red dots indicate significantly dysregulated metabolites with fold change > 1.5 and t-test p-value < 0.05. (B) Metabolic pathway network analysis with the significantly dysregulated metabolites. The nodes and edges represent metabolites and reactions, respectively. Red nodes indicate upregulated metabolites whereas green ones indicate downregulated. The size of the nodes is a function of the p value. The larger the size lower the p-value. Multivariate statistical analysis with the metabolites showing (C) PCA score scatter plot, (D) PLS-DA score scatter plot, and (E) results of permutation test statistics.

TBK-1 induced expression of metabolic pathway genes

To assess if the observed changes in metabolic pathways were due to TBK1-mediated regulation of gene expression, RNA sequencing was performed on the MDA-MB-231 cell line after knocking down TBK1 using the siRNA technique. RNA sequencing analysis led to the identification of more than 40,000 RNA transcripts which included, protein-coding mRNAs, pseudogenes, lincRNAs, etc (Supplementary Table 2). Volcano plot filtering was used on the protein-coding mRNAs to identify the statistically significant differentially expressed genes between the two groups (fold change ≥2 and p ≤ 0.05) (Figure 3A). Our results indicate 159 genes to be differentially expressed with statistical significance. Among them, 43 genes were upregulated and 116 downregulated between the TBK-1 knockdown and control groups. To investigate the possible biological function of these differentially expressed genes, Gene Ontology (GO) analysis was performed. In the molecular function category, 33.8 % of genes were involved in binding activity, 11.3 % in catalytic activity, 7.9 % were molecular function regulators and 6.6 % had transporter activity. Cellular (49.7 %) and metabolic (31.1 %) processes were the most highly represented under the biological process category followed by biological regulation (29.8%), response to stimulus (25.2 %), signaling (16.6 %), localization (13.2 %), multicellular organismal process (11.3 %), immune system process (11.3 %), etc. The majority of the mRNAs in the cellular component category were cellular anatomical entities (60.9 %) and intracellular (34.4 %) (Figure 3C). Furthermore, the top pathways obtained from pathway analysis using Panther 16.0 included inflammation mediated by chemokine and cytokine signaling, integrin signaling, epidermal growth factor receptor (EGF) receptor signaling, nicotinic acetylcholine receptor signaling, cholecystokinin receptors (CCKR) signaling, etc. (Figure 3B).

Figure 3:

TBK-1 influenced transcriptome alterations. (A) Volcano plot analysis of transcriptomics dataset between siTBK-1 and siC groups. Red dots indicate significantly dysregulated genes with fold change > 2.0 and FDR-adjusted p-value < 0.05. (B) Panther pathway analysis with the significantly dysregulated genes. (C) Gene ontology (GO) analysis of the significantly dysregulated genes. The genes were classified into three major domains- molecular function, biological process, and cellular component.

Integrated metabolomics and transcriptomics analysis

Integrated KEGG pathway analysis with significantly dysregulated metabolites and genes obtained from metabolomics and transcriptomics datasets, respectively revealed the involvement of several metabolic pathways including pyrimidine metabolism, galactose metabolism, biosynthesis of unsaturated fatty acids, thiamine metabolism, taurine and hypotaurine metabolism, fatty acid biosynthesis, β-alanine metabolism, and pantothenate and CoA biosynthesis. Pyrimidine and galactose metabolism pathways are further explored and discussed below.

Pyrimidine metabolism

Pyrimidine biosynthesis involves two pathways: salvage pathway and de novo pathway. In the salvage pathway, the preformed free bases are recovered and joined to the ribose sugar. In contrast, de novo synthesis of nucleotides begins from scratch where the pyrimidine bases are synthesized from simpler molecules. First, dihydroorotate is synthesized from glutamine, bicarbonate, and aspartate which is then converted to orotate. Subsequently, ribose 5-phosphate is attached to orotate and converted to uridine monophosphate (UMP) in two catalytic reactions. UMP acts as a precursor for the common pyrimidine di- and tri-nucleotides required for nucleic acid synthesis. Our results indicate low levels of orotate and dihydroorotate and high levels of UMP, uridine diphosphate (UDP), and cytidine diphosphate (CDP) in TBK-1 silenced cells compared to the control cells (Figure 2A, Figure 4A and Table 1). This implies an imbalance in the de novo synthesis pathway highlighting the role of TBK-1 in pyrimidine metabolism. Furthermore, our results also show the downregulation of thymidine phosphorylase (tymp) in TBK-1 knocked-down cells (Figure 3A, Figure 4A and Supplementary Table 2). This was confirmed by performing quantitative PCR after TBK1 knockdown (Figure 4D). mRNA and protein expression of TYMP were also analyzed for breast cancer patients and normal controls using GEPIA and UALCAN databases respectively. TYMP was found to be over-expressed at both mRNA (Figure 4B) and protein level (Figure 4C) with high confidence in breast cancer patients as compared to normal subjects. TYMP plays an important role in the pyrimidine salvage pathway. It catalyzes the reversible conversion of thymidine to thymine and 2-deoxy-α-D-ribose-1-phosphate, and the phosphorolysis of deoxyuridine to uracil and 2-deoxy-α-D-ribose-1-phosphate. Our results showing lower levels of tymp accompanied by higher levels of thymine and uracil indicate the role of TBK-1 in controlling the reverse reaction catalyzed by TYMP (Figure 4).

Figure 4:

Pyrimidine metabolism alterations due to TBK-1 silencing. (A) KEGG pathway with round nodes representing metabolites within the core regulatory network. Genes are represented by square nodes. Red and green indicate upregulation and downregulation, respectively of the differentially expressed genes/metabolites. Grey indicates genes/metabolites not detected in the experiment. TYMP expression levels in breast invasive carcinoma at (B) transcript level and (C) protein level were plotted using GEPIA and UALCAN databases respectively. (D) Validation of the TYMP gene expression using Real-time PCR. Plot is an average of three independent experiments. Error bars represent mean ± S.D.

Table 1:

Details of metabolites significantly expressed in the pyrimidine and galactose metabolism pathways.

Metabolites	RT (min)	m/z	Compound	HMDB	KEGG	Mode	Fold Change	P value
Uridine 5’-Diphosphate	6.00	402.994339	C9H14N2O12P2	HMDB0000295	C00015	NEG	2.04	0.026
Uridine-5’-monophosphate	5.15	323.0281576	C9H13N2O9P	HMDB0000288	C00105	NEG	2.10	0.041
Cytidine 5’-diphosphate	6.25	402.010376	C9H15N3O11P2	HMDB0001546	C00112	NEG	1.83	0.039
	6.23	308.0904032				POS	1.84	0.038
Deoxycytidine	2.68	226.0830739	C9H13N3O4	HMDB0000014	C00881	NEG	0.74	0.029
	2.70	146.1172714				POS	0.71	0.019
Uracil	2.36	111.0198046	C4H4N2O2	HMDB0000300	C00106	NEG	3.59	0.003
	2.36	104.0703653				POS	3.33	0.003
5,6-dihydrouracil	2.07	113.0354659	C4H6N2O2	HMDB0000076	C00429	NEG	2.28	0.004
Thymine	2.03	125.0354125	C5H6N2O2	HMDB0000262	C00178	NEG	3.35	0.004
Dihydroorotate	3.29	157.025177	C5H6N2O4	HMDB0003349	C00337	NEG	0.31	0.0003
Orotate	2.87	155.0096105	C5H4N2O4	HMDB0000226	C00295	NEG	0.37	0.0002
D-glucose 1-phosphate	5.59	259.0220133	C6H13O9P	HMDB0001586	C00103	NEG	1.64	0.028
Galactitol	4.47	181.0715764	C6H14O6	HMDB0000107	C01697	NEG	0.57	0.011

Galactose metabolism:

Galactose is mainly metabolized via the Leloir pathway ^{36 37}. In this pathway, α-D-galactose is first phosphorylated into α-D-galactose-1-phosphate. It then reacts with UDP-glucose to form α-D-glucose-1-phosphate and UDP-galactose. The glucose-1-phosphate produced leads to the formation of α-D-glucose-6-phosphate which enters the glycolysis pathway for energy. Apart from the Leloir pathway, three accessory pathways exist for galactose metabolism ³⁸. One of them involves the conversion of excess galactose into galactitol by an aldose reductase which is encoded by the gene aldo-keto reductase family 1 member B10 (akrb10). Galactitol, however, cannot be further efficiently metabolized and is toxic to the cells. Excess galactitol may result in depletion of NADPH leading to lowered glutathione reductase activity and accumulation of hydrogen peroxide or other free radicals ³⁹. High concentration of this metabolite is reported in several galactose metabolism disorders ⁴⁰. Our results indicate higher levels of α-D-glucose-1-phosphate along with lower expression of akrb10 and lower levels of galactitol in TBK-1 silenced cells compared to the control cells (Figure 2A, 3A, 5A and Table 1). Downregulation of AKR1B10 on TBK1 silencing was also confirmed by performing quantitative PCR (Figure 5D). This could indicate the role of TBK-1 in regulating galactose and thereby, energy metabolism in breast cancer cells. Comparative analysis of AKR1B10 at mRNA and protein levels between breast cancer patients and normal subjects was done using UALCAN database. However, average AKR1B10 expression was not observed to be significantly different between the two groups (Figure 5B, 5C). Interestingly, on closer look, several patient samples were found to have higher expression both at the mRNA and protein levels as is evident from the outliers in Figure 5B and 5C, respectively. In addition, several independent studies have reported high expression of AKR1B10 both in breast cancer tissue and serum ^41–43.

Figure 5:

Galactose metabolism alterations due to TBK1 silencing. (A) KEGG pathway with round nodes representing metabolites within the core regulatory network. Genes are represented by square nodes. Red and green indicate upregulation and downregulation, respectively of the differentially expressed genes/metabolites. Grey indicates genes/metabolites not detected in the experiment. AKR1B10 expression levels in breast invasive carcinoma at (B) transcript level and (C) protein level were plotted using GEPIA and UALCAN databases respectively. (D) Validation of the AKR1B10 gene expression using Real-time PCR. Plot is an average of three independent experiments. Error bars represent mean ± S.D.

TBK1 mediates cell survival in an AKR1B10 and TYMP-dependent manner

TBK1, AKR1B10, and TYMP all have been reported to have oncogenic functions, and we found TBK1 silencing to result in downregulation of both TYMP (Figure 4D) and AKR1B10 (Figure 5D). Therefore, we next examined whether TBK1 cooperated with AKR1B10 and TYMP to regulate cell viability. An MTT assay was performed for breast cancer cell lines- MDA-MB-231 and Hs578T (Table 2) after chemical inhibition of TBK1 either alone or in combination with AKR1B10 or TYMP. We found that the TBK1 inhibitor cooperated with both AKR1B10 inhibitors diclofenac and flufenamic acid as well as with TYMP inhibitor tipiracil to have a cumulative effect on the viability. We observed that IC₅₀ for the TBK1 inhibitor was significantly reduced when a constant dose of either of the two AKR1B10 inhibitors or TYMP inhibitor was used along with the TBK1 inhibitor (Table 2). In addition, IC₅₀ of AKR1B10 and TYMP inhibitors was also markedly reduced when used in combination with TBK1 inhibitors in MDA-MB-231 and Hs578T cell lines (Table 2). Next, we did a synergy analysis for TBK1 inhibitor, BX795 with AKR1B10 and TYMP inhibitors by feeding the viability data into CompuSyn software. The software gave fraction affected (Fa) versus combination index (CI) and isobologram plots as readouts. A CI<1 indicates a synergy between the drugs being tested. CI was found to be less than 1 for all the tested concentrations of diclofenac and flufenamic acid with BX795 except for the highest two concentrations of BX795 (20μM and 40μM) in both MDA-MB-231 (Figure 6A, 6C respectively) and HS578 (Figure 6B, 6D respectively) cell lines. Both these concentrations are above physiological levels and therefore not relevant. Isobologram is a plot of the dose reduction index (DRI) of the drugs under study. DRI is the measure of the number of folds the dose of each drug can be reduced in a synergistic combination for a given effect, as compared to the dose of the drug when used alone. The isobologram plots also indicated similar findings, all concentrations of diclofenac (Figure 7A-B) and flufenamic acid (Figure 7C-D) were found to be in synergy with all except the two highest concentrations of BX795 in both the cell lines. CompuSyn however did not find a synergy between BX795 and TYMP inhibitor tipiracil (Supplementary Figure 1). This might be because a typical sigmoidal IC₅₀ curve could not be obtained for tipiracil alone as well as in combination with BX795. However, the plots indicate a marked drop in cell viability in the presence of a combination treatment with tipiracil and BX795 in both cell lines (Supplementary Figure 2). We believe that these results provide a strong indication that targeting TBK1 along with AKR1B10 might be a more efficient way to combat breast cancer, as compared to pursuing them in isolation. Also, combination therapy for TBK1 and TYMP might be potentially useful but needs further exploration.

Table 2:

TBK1 regulates cell viability in an AKR1B10 and TYMP-dependent manner. IC₅₀ was calculated for TBK1 inhibitor BX795, AKR1B10 inhibitors- diclofenac, flufenamic acid, and TYMP inhibitor tipiracil alone and in combination in MDA-MB-231 and Hs578T cell lines.

DRUGS	MDA-MB-231	Hs578T
	IC50 (μM)
BX795	2.8	3.8
BX795+ 100μM DICLOFENAC	0.82	0.95
BX795+ 100μM FLUFENAMIC ACID	1.6	2.6
BX795+ 100μM TIPIRACIL	2.07	2.23
DICLOFENAC	225.2	232
DICLOFENAC + 2.5μM BX795	0.03	13.76
FLUFENAMIC ACID	352	356.3
FLUFENAMIC ACID + 2.5μM BX795	3.5	1.33
TIPIRACIL	>500	528
TIPIRACIL + 2.5μM BX795	<1.953	<1.953

Figure 6:

Fa-CI plots for TBK1 inhibitor BX795 and AKR1B10 inhibitors diclofenac/ flufenamic acid. CompuSyn software was utilized to generate Fa-CI plots from viability data of BX795 and diclofenac combination in (A) MDA-MD-231 and (B) Hs578T cell lines. Fa-CI plots for BX795 and flufenamic acid combination in (C) MDA-MD-231 and (D) Hs578T cell lines

Figure 7:

Isobologram plots for TBK1 inhibitor BX795 and AKR1B10 inhibitors diclofenac/ flufenamic acid. CompuSyn software was utilized to generate isobologram plots from viability data of BX795 and diclofenac combination in (A) MDA-MD-231 and (B) Hs578T cell lines. Fa-CI plots for BX795 and flufenamic acid combination in (C) MDA-MD-231 and (D) Hs578T cell lines

Discussion

Lately, emerging technical developments have allowed holistic investigation of cancer metabolism using high throughput omics data including, transcriptomics, proteomics, metabolomics, and lipidomics ^44,45,46,47. These studies have shown potential for a detailed understanding of metabolic perturbations in cancer. Metabolomics is the global study of small molecules or metabolites in a cell, tissue, or biological fluid ⁴⁸. Metabolites are the end products of several biological processes and may be indicative of upstream genetic alterations in disease. Metabolomics therefore provides a snapshot of these metabolic processes or perturbations at a given time point. While informative, metabolomics alone cannot fully account for the underlying metabolic phenotype. When complemented with other omics approaches, such as transcriptomics, it provides a comprehensive understanding of the metabolic phenotypes along with their underlying genetic alterations. Transcriptomics is the survey of the entire genome for global gene expression profiling in a biological sample. An integrated metabolomics and transcriptomics approach has provided a detailed understanding of metabolic rewiring in cancer ^49,50,51 and has also led to the identification of new therapeutic targets ⁵². Therefore, in this study, we have used an integrated metabolomics and transcriptomics approach to study the role TBK1 plays in rewiring cancer metabolism.

TBK1 is over-expressed in invasive breast carcinoma at both transcript and protein levels. Importantly, this over-expression has been correlated with poor overall patient survival for breast cancer (Figure 1). Recently, some reports showed that TBK1 plays an important role in regulating metabolism and energy homeostasis ^9,20,21. TBK1-mediated inhibition of AMPK1 and mTORC1 are major factors in this deregulated cellular metabolism ²². These pro-survival pathways are then instrumental in cancer drug resistance and metastasis ⁵³. It has been reported that TBK1 dysregulates metabolism in an mTORC1 dependent manner in animal models of breast cancer⁵⁴. It was also recently reported that TBK1 phosphorylates substrates involved in energy metabolic signalling in lung cancer cell lines ⁵⁵. However, the contribution of metabolic regulation in TBK1-mediated oncogenesis is not fully known. The role of TBK1 in regulating metabolism is further substantiated by our finding that nucleotide, energy, amino acid, and lipid metabolism pathways are altered on TBK1 knockdown (Figure 2). On an integrated analysis of metabolomics and RNAseq datasets, we found pyrimidine and galactose metabolism to be the pathways and thymidine phosphorylase (TYMP) (Figure 4) and Aldo-keto reductase1 B10 (AKR1B10) (Figure 5) to be the metabolic genes which are significantly altered on TBK1 knockdown. Thymidine phosphorylase (TYMP) catalyzes the reversible phosphorylation of thymidine and other pyrimidine-2′-deoxyribonucleosides. It plays an important role in the pyrimidine salvage pathway. It is also known as platelet-derived endothelial cell growth factor (PD-ECGF) which is an angiogenic factor. It is found to be elevated in colorectal, pancreatic, breast, lung, and liver cancers ^{56,57,58,59,60} and is associated with poor outcome ⁶¹. It is pro-angiogenesis, pro-metastasis, and invasion and anti-apoptotic in cancer cells ^62,63. TYMP is also reported to play a role in metastasis ^61,64,5,6.

AKR1B10 is involved in galactose metabolism and has been reported to be a metastasis enhancer in breast cancer ⁴¹. It was found to be over-expressed on the tumorigenic transformation of mammary epithelial cells. Its expression in breast cancer was found to be positively correlated with metastasis and tumor size and was inversely related to patient survival. Silencing of AKR1B10 in breast cancer cell lines was found to inhibit cell proliferation ⁴². AKR1B10 has been reported to play an oncogenic role in squamous cell carcinoma ⁶⁵, hepatocellular carcinoma ⁶⁶, and smoking-related non-small cell lung carcinoma ⁶⁷.

TYMP, AKR1B10, and TBK1 have been separately reported to play a role in tumor proliferation. We observed a reduction in both TYMP and AKR1B10 expression levels on TBK1 knockdown (Figures 4–5). Also, combined chemical inhibition of TBK1 along with TYMP or AKR1B10 resulted in a significant decrease in cell viability of breast cancer cell lines as compared to when they were individually inhibited (Table 2). TBK1 and AKR1B10 inhibitors were found to act synergistically (Figures 6–7). In contrast, synergy analysis results remained inconclusive and require further exploration (Supplementary figures 1-2). These findings suggest that TYMP and AKR1B10 are involved in TBK1-mediated oncogenesis. This indicates that TYMP and AKR1B10 may be involved in the metabolic rewiring that TBK1 carries out in the cell. Our study provides novel information about the molecular mechanisms involved in TBK1-mediated oncogenesis. Based on our findings, we postulate that elevated levels and activity of TBK1 enhance tumor growth, epithelial-to-mesenchymal transition (EMT), and metastasis via dysregulation of cellular metabolism. TYMP and AKR1B10 along with TBK1 therefore make for potentially attractive targets for cancer therapy.

Future Scope

The molecular mechanics and signalling pathways involved in the roles of TYMP and AKR1B10 in TBK1- mediated metabolic re-wiring of cancer metabolism need further elucidation.

Supplementary Material

Figures 1 and 2

Supplementary Figure 1: Synergy analysis for TBK1 inhibitor BX795 and TYMP inhibitor tipiracil. Fa-CI plots for the combination in (A) MDA-MB-231 and (B) Hs578T cell lines. Isobologram plots for the combination in (C) MDA-MB-231 and (D) Hs578T cell lines.

Supplementary Figure 2: Viability curves for tipiracil alone and in combination with a constant dose of 2.5μM TBK1 inhibitor BX795 in (A, B) MDA-MB-231 and (C-D) Hs578T cell lines

Table 1

Supplementary Table 1: Metabolomics dataset

Table 2

Supplementary table 2: Transcriptomics dataset