Background
Endometrial cancer is the fourth most common cancer in women with 12,990 new diagnoses and 4120 deaths in 2016 in the United States [
1]. Over 710,200 women are living with endometrial cancer in the United States, and approximately 2.8% of women will be diagnosed with the disease at some point during their lifetime. As the most significant risk factor for endometrial cancer is obesity, a majority of the biomarkers used to detect and monitor endometrial cancer development are related to metabolic and endocrine alterations [
2]. Androgens, estrogens, prolactin, thyroid stimulating hormone, leptin, and adiponectin are a few of the biomarkers utilized to highlight risk of endometrial cancer development. While these biomarkers can be useful, they are oftentimes somewhat subjective as the levels of these hormones fluctuate naturally, are generally elevated with obesity, and are not necessarily unique to cancer development [
2,
3]. In order to find new biomarkers that may act as diagnostic biomarkers for endometrial cancer, we evaluated Jagged2 (JAG2), Aurora Kinase A (AURKA), Phosphoglycerate Kinase 1 (PGK1), and Hypoxanthine Guanine Phosphoribosyltransferase 1 (HPRT1) for their role in cellular proliferation and cancer development. We evaluated these genes because of their upregulation and diagnostic potential in other cancer types [
4‐
9].
JAG2 is a notch transmembrane ligand. Notch signaling is a conserved signaling pathway linked to the development of several cancers due to its role in cell fate, cellular proliferation regulation, and cell death [
10]. This is exemplified by the fact that Notch signaling regulates stem cell proliferation and differentiation [
11]. Within cancer, Notch signaling mediates hypoxia, invasion, and chemoresistance [
12], and JAG2 expression in primary tumors has been correlated with vascular development and angiogenesis [
13]. In addition, elevated levels of JAG2 result in significant chemoresistance, and when JAG2 is knocked down in mice, tumor cells become sensitive to chemotherapeutics (doxorubicin) [
8]. Notch signaling has been identified as an important pathway for carcinogenesis of the endometrium [
14]. Additionally, JAG2 has been shown to be a promising target in several cancer cell lines, as specific antibody–drug conjugates have resulted in tumor reduction [
15].
AURKA is a cell-cycle regulated kinase that functions in spindle formation and chromosome segregation during the M phase of the cell cycle. AURKA has been shown to be a downstream target of MAPK1, which is a major force in cellular proliferation in several cancer cells [
16]. The protein is also elevated in a variety of cancers and has a significant association with disease recurrence [
6,
7]. Because AURKA is upregulated in cancers, efforts have been made to target the protein to aid in tumor reduction. Upon AURKA suppression, cancer cells become sensitive to chemotherapeutics and overall tumor growth is suppressed in a variety of cancer cells (docetaxel and taxane) [
17,
18]. The role AURKA may play as a diagnostic biomarker in endometrial cancer has not been well studied, although it has shown promising results in other cancer types [
6,
7,
19‐
21].
PGK1 is involved in the glycolysis pathway and functions by transferring a phosphate group from 1,3-bisphosphoglycerate to ADP to form ATP [
22,
23]. As an enzyme involved in generating valuable energy for the cell, especially in hypoxic conditions, PGK1 has been correlated with cancer development and progression in a variety of tumor types [
9,
24,
25]. Its role in promoting tumor proliferation is linked to PGK1’s ability to promote tumor angiogenesis [
26,
27], DNA replication and repair [
28,
29], and cancer metastasis [
25,
30]. While the protein is elevated internally in several cancers, it is also actively secreted from tumor cells, where it cleaves plasminogen to create angiostatin [
31]. PGK1 has been shown to be upregulated in several cancer types, but has not been evaluated for upregulation in endometrial cancer [
25,
32].
HPRT1 is a nucleotide salvage enzyme involved in the cell cycle [
33,
34]. This enzyme is a transferase responsible for producing guanine and inosine nucleotides by transferring a phosphoribose from PRPP to guanine and inosine bases, respectively, during cellular maintenance [
35,
36]. As cells rapidly divide, the need for nucleotides increases, and subsequently HPRT1, has been shown to be elevated in several malignant settings [
4,
37]. As the enzyme shows upregulation in malignant tissue while maintaining stable levels in normal tissue, it has the potential to be used as a biomarker for cancer development in several cancer types.
We decided to evaluate these enzymes in endometrial cancer because they have all shown promising diagnostic potential in other tissue types as biomarkers for disease development and progression but have not been evaluated in endometrial cancer. As malignant endometrial biomarkers are less established, we hope to identify additional markers for malignancy to aid in the early diagnosis and possible treatment of endometrial cancer.
Materials and methods
Chemicals/reagents
DIVA Decloaker 10x, Background Sniper, Mach 4 HRP polymer, DAB Peroxidase, Hematoxylin, Hydrophobic pen, and Universal Negative antibodies were all obtained from Biocare Medical, Concord, CA. Anti-JAG2 (LifeSpan Biosciences, Inc. Seattle, USA), Anti-AURKA (Sigma-Aldrich, St. Louis, USA), and anti-PGK1 (Abcam, Cambridge, UK) were stored at − 20 °C. Anti-HPRT monoclonal antibody (Abcam, Cambridge, UK) was aliquoted and stored at − 20 °C. GAPDH polyclonal antibody (Cell signaling) was aliquoted and stored at − 20 °C. Tween20 (Fisher Reagents, Waltham MA) was stored at room temperature. Hydrogen Peroxide at 30% (Fisher Reagents, Waltham MA) was stored at 4 °C.
Tissue microarray samples
Tissue microarrays were obtained from Biomax and stained for GAPDH, HPRT, JAG2, AURKA, PGK1, and with an isotype control. Patients were all female and ranged in age from 21 to 63. Normal (n = 9), cancer adjacent (n = 9), and malignant tissue (n = 54) (grade 1–3) were included in the analysis (Table
1).
Table 1
Protein expression within patient tissue
HPRT | 68 | Nucleotide salvage | 157.206 | 186.176 | 223.207 |
PGK1 | 71 | Glycolytic enzyme | 107.273 | 154.437 | 171.748 |
AURKA | 72 | Cycle-regulated kinase | 209.994 | 236.147 | 244.352 |
Jag2 | 72 | Protein coding | 143.635 | 194.297 | 186.269 |
Immunohistochemistry
Protein levels were assessed using protocols described by Townsend et al. with slight modifications [
4]. Briefly, tissues were rehydrated, washed, and treated with DIVA Decloaker. Following a hydrogen peroxide wash, tissues were treated with a Background Sniper followed by a primary antibody (1:100 dilution). After a series of washes, the tissues were treated with DAB Peroxidase and hematoxylin and imaged using a standard light microscope.
Tissue quantification
ImageJ software was utilized to quantify staining intensity [
38]. An IHC toolbox plugin was selected with the “DAB (more brown)” option to remove staining that did not result from DAB. After this modification, the images were converted to a grayscale and a threshold was applied to eliminate areas of negative space that could potentially bias the results. Once a universal threshold was applied, the average gray intensity of the tissue was collected.
Tumor gene-expression analysis
We obtained RNA-sequencing and clinical outcomes data for Uterine Corpus Endometrial Carcinoma (UCEC) samples from The Cancer Genome Atlas (TCGA) [
39]. We used transcripts-per-million values, summarized at the gene level. These data were derived from tumor and normal samples.
Survival was calculated using a Cox proportional hazard model. In addition to gene expression (primary variable), covariates included gene expression and clinical factors such as age, race, and tumor purity. Kaplan–Meier curves were generated to compare survival of patients with the highest 20% of target gene expression against those with the lowest 20% of target gene expression. The statistical analyses and curve generations were calculated utilizing the TIMER program developed by Li et al. [
40].
Drug response analyses
We evaluated the effects of chemotherapy treatments on cell lines using two publicly available databases. First, we examined data from the Cancer Cell Line Encyclopedia (CCLE) [
41]. We obtained treatment-response data for 24 drugs that were available from the CCLE portal and used the area above the fitted dose–response curve (ActArea) as a metric of treatment response [
42]. We obtained transcript-level expression levels for CCLE [
43] and summed protein-coding transcript values to gene-level values using a custom Python script (
https://python.org). For each of four genes (HPRT1, AURKA, JAG2, and PGK1), we identified cell lines for which drug-response and gene-expression data were available and then ranked the cell lines according to expression of the respective genes. Next, we selected the lowest- and highest-expressing cell lines for each gene and used a Mann–Whitney U test to evaluate differences in ActArea values between these cell-line groups. To perform these calculations, we used the R statistical software (version 3.4.3) [
44].
Second, we evaluated data from the Library of Integrated Network-based Cellular Signatures, which contains gene-expression profiles for cell lines after drug perturbations. We wrote a Python (version 3.6.5) script to extract HPRT1 and AURKA expression values from the LINCS database for samples from 26 cell lines for which data were available. We used the Level 5 data, which were generated using the L1000 platform [
45], normalized using a z-score methodology within each plate, and averaged across replicates. Using the R (version 3.4.4) [
44] statistical software and the readr package (version 1.1.1) [
46], we parsed the metadata file to identify experiments where the cell lines had been treated with chemotherapeutic compounds (pert_type = “trt_cp”). The summarized data values indicate relative gene-expression levels for cells treated with a given compound relative to control-treated cells. To perform this filtering and data transformation, we used the dplyr (version 0.7.4) [
47] and reshape2 (version 1.4.3) packages [
48]. Before plotting the data, we grouped the values for each cell line by compound name. We identified the median value for each group and sorted the values from lowest to highest. Then we used the superheat package (version 1.0.0) to create heatmaps with data from the 7 cell lines with the most treatment data [
49]. The code and data we used for this analysis can be found at
https://bitbucket.org/alyssaparker99/lincs-heatmaps.
Statistical analysis
Staining intensities between tissue samples were analyzed using an ANOVA test with the multiple comparison method. Additionally, unpaired t tests were utilized in conjunction to confirm statistical significance. These statistical tests were performed in GraphPad Prism 7 software. Differences were considered significant when the p value was < 0.05. Asterisks were used in figures to indicate levels of significance with ns = p > 0.5, *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, and ****p ≤ 0.0001.
Discussion
We have determined that there is a significant elevation of JAG2, HPRT1, AURKA, and PGK1 expression in endometrial cancer. With elevated expression upon malignancy, these genes can be utilized as a companion diagnostic tool to both identify and characterize endometrial cancer. As cancer specific biomarkers, these genes may serve as useful markers when analyzing endometrial cancer development within patient tissue. Additionally, HPRT and PGK1 show additional promise as possible biomarkers for cancer grade as the levels of the proteins elevated in a stepwise manner with higher cancer grade. These biomarkers have already shown utility in other cancer types [
4‐
6,
8,
9,
16] and we have shown that their use may also extend to endometrial cancer.
While there are several epigenetic biomarkers for endometrial cancer (p52, KRAS, VEGF. PTEN, etc.) [
50,
51], there remains a need to find suitable protein biomarkers for not only endometrial diagnosis, but also as possible targets for future therapies. Future directions to this work include evaluating a larger cohort of patients to determine whether the expression of these biomarkers could have clinical application. Especially in the case of both HPRT1 and AURKA, it may be beneficial to develop therapies to reduce their expression, thereby determining whether these genes play a critical role in cancer survival and proliferation as they show significant impact on overall patient survival.
In addition, the conserved pathways that HPRT1 and AURKA have in terms of drugs that inhibit or induce their expression, may indicate a regulatory relationship between the inhibited pathway and the proteins that have not yet been identified. The merit of this hypothesis is demonstrated as AURKA has a reciprocal regulation with PLK1 in mitotic entry and within spindle assembly [
52]. This corresponds to the data we have observed as the drugs with the largest impact on AURKA elevation with the highest consistency are inhibitors of PLK1 and microtubule formation. Yet, the consistent relationship between drugs that inhibit HPRT1 expression are both inhibitors of Topo I and the MEK pathway. There has not been any investigation into the relationship between HPRT1 and these proteins/pathways and our initial data show that a possible link exists. With this in mind, this potential relationship merits further examination and could potentially elucidate novel gene interactions specific to cancer.
Authors’ contributions
MHT created the concepts and designs of the experiments, carried out the acquiring and analysis of data and drafted the manuscript. ZE performed the bioinformatic analysis of expression in addition to the analysis of different chemotherapy drugs. AF carried out the IHC staining. AP did the bioinformatic analysis on the response of cell lines to various drugs. SP aided in the bioinformatic statistical evaluation and conceptual design and in manuscript revisions. RR was involved in the manuscript revisions. KO participated in the study design and revision of the manuscript. All authors read and approved the final manuscript.