Identification of Nkx2–3 and TGFB1I1 expression levels as potential biomarkers to predict the effects of FOLFOX4 chemotherapy

Abstract

This study was designed to detect global gene expressions of primary advanced colorectal cancer (ACC) patients who have undergone FOLFOX4 chemotherapy and screen valuable biomarkers to predict the effects of chemotherapy. Samples from primary ACC patients who have undergone FOLFOX4 chemotherapy were collected. Their chemotherapy effects were evaluated and divided into chemotherapy sensitive group (experimental group) and non-sensitive group (control group). Cancerous tissue gene expression profiles were detected by chip technology. Two groups with differentially expressed genes were screened by cluster analysis and significance analysis of microarrays (SAM). Valuable biomarkers were screened by bioinformatics analysis. Immunohistochemical analysis was performed to characterize the pattern of Nkx2–3 and TGFB1I1 expression. Nkx2–3 and TGFB1I1 signal log ratio were used Receiver Operating Characteristic (ROC) analyses to calculate its own predicting accuracy. Thirty cases were divided into experimental group (13 cases) and control group (17 cases). There was evident difference in the tumor cell biology states of the two groups; that is, 25 ESTs (21 genes) were upregulated and 5 ESTs (5 genes) were downregulated. Nkx2–3 protein was observed on the nucleus of the cancer cells and TGFB1I1 protein was observed on the nucleus and cytoplasm of the cancer cells in experimental group. Their prediction accuracies were 85.3% and 76.7% respectively. Nkx2–3 and TGFB1I1 expressions in control group are very low, but highly expressed in the experimental group; Nkx2–3 and TGFB1I1 may be classified as valuable biomarkers, as these can predict the effects of primary ACC patients who will undergo FOLFOX4.

Keywords: :

• colorectal cancer
• FOLFOX4
• chemotherapy
• gene expression
• biomarker
• Nkx2-3
• TGFB1I1

Introduction

Colorectal cancer (CRC) is known as the third highest incidence rate of cancer and is the fourth leading cause of death due to cancer worldwide.¹ The World Health Organization (WHO) estimates that 945,000 new cases are diagnosed yearly, resulting in 492,000 deaths.¹ Early CRC patients may be cured by surgical operation, but 30–40% of them are confirmed only while in the advanced metastatic stage and hence cannot be cured by surgery alone.² In addition, more than half of the patients who are assumed initially as curable suffer from recurrence and eventually die of the disease.³ In the past 40 y, fluorouracil (FU)-based chemotherapy has been the main treatment for CRC metastasis. Recently, significant progress with oxaliplatin and other new cytotoxic drugs and monoclonal antibody targeted therapies have been introduced into chemotherapy.^4-7 Some of the most commonly combined chemotherapy, such as FOLFOX4 (fluorouracil, leucovorin, and oxaliplatin), can obtain nearly 50% objective remission rates.⁸ However, about half of the patients on this new combination chemotherapy have remained non-sensitive; the majority eventually appeared as drug-resistant. From among different CRC patients who have received the same treatment of chemotherapy regimens and drug doses, there have been notable individual differences in curative effects resulting in tolerability.^9,10 Hence, there is an urgent need to identify some valuable biomarkers based on their expressions in order to predict the effects of chemotherapy, as well as to carry out tailored therapies. At present, many have attempted to explore various chemotherapy reaction predictors.^11-21 However, two recent systematic review analyses have shown that most prediction effects are somewhat controversial,^22,23 and conducting research on the basic fields of chemotherapy predictors remain a difficult task.

Gene chip technology is important in functional gene detection. It has many advantages, such as high throughput, high efficiency, and low consumption in the analysis of biological information processing and large-scale access to relevant information; it can also screen gene-related cancer occurrence and development, as well as its transfer mechanism and chemotherapy drugs reaction.^24-26 It contributes to the research of tumor to the genome level in a systemic and holistic manner.

Herein, by using microarray analysis of gene expression profile, we compared the global gene expression patterns of primary ACC patients who have undergone FOLFOX4 chemotherapy from the experimental group with the control group. The objective of this study was designed to detect gene expressions and screen valuable biomarkers according to their expressions in order to predict the effects of chemotherapy, as well as to implement tailored therapies.

Results

Patient evaluation of the study sample

The study start enrolled 34 patients; 3 patients could not tolerate chemotherapy, and hence failed to fulfill the FOLFOX4 chemotherapy and exited the research midway. One of these patients was unable to finish R0 resection of the primary tumor after chemotherapy and was also eliminated from the study. Finally, 30 cases completed the required treatment; all were included in the analysis. Among these patients (Table 1), none was observed with complete response (CR); 13 cases were with partial response (PR) and were included into the experimental group; and 6 cases were with stable disease (SD) while 11 cases were with progressive disease (PD), and these were included into the control group.

Table 1. Baseline Patient Characteristics (intent-to-treat population)

Characteristic	control group		experimental group
	No.	%	No.	%
No. of patients	17		13
Male	12	70.6	9	69.2
Female	5	29.4	4	30.8
Age (years)	57.1 ± 11.6		52.8 ± 13.0
BMI (kg/m2)	22.2 ± 2.4		22.1 ± 2.3
ECOG performance status
0	11	64.7	7	53.8
1	6	35.3	6	46.2
Primary tumor site
Colorectal	2	11.8	2	13.3
Colon	8	47.1	6	47.6
Rectal	7	41.2	5	40.0
No. of metastatic sites
0	0		0
1	7	41.2	5	40.0
2	6	35.3	5	40.0
3	2	11.8	2	13.3
≥ 4	2	11.8	1	8.3
Differentiation
WD	2	11.8	0	0
MD	9	52.9	12	91.7
PDi	6	35.3	1	8.3
Evaluation
CR	0	0	0	0
PR	0	0	13	100.0
SD	6	35.3	0	0
PD	11	64.7	0	0

Microarray data and bioinformatics analysis

After SAM analysis of microarray data for experimental vs. control group, there were differentially expressed genes or ESTs, upregulated genes or ESTs (9,359), and downregulated genes or ESTs (1,1104). Thirty ESTs were certified as significantly differentially expressed between the two groups, with EST |Score(d)| ≥ 2, with Fold Change ≥ 2 or Fold Change ≤ 0.5. Meanwhile, 25 ESTs (21 genes) were upregulated whereas 5 ESTs (5 genes) were downregulated (Tables 2 and 3). The list of 26 differentially expressed genes was used for unsupervised hierarchical clustering, and the results were analyzed and visualized by a TreeView program (Fig. 1).

Table 2. Upregulated gene (n = 21) for experimental group vs. control group

Probe Set ID	UniGene ID	Gene Symbol	Location	Description
227662_at 225720_at	Hs.655519	SYNPO2	chr4q26	synaptopodin 2
201497_x_at 213764_s_at	Hs.512842	MFAP5	chr12p13.1-p12.3	microfibrillar associated protein 5
37512_at 205700_at	Hs.524513	HSD17B6	chr12q13	hydroxysteroid (17-β) dehydrogenase 6 homolog (mouse)
211737_x_at 209465_x_at	Hs.371249	PTN	chr7q33-q34	pleiotrophin
204042_at	Hs.635221	WASF3	chr13q12	WAS protein family, member 3
217897_at	Hs.713034	FXYD6	chr11q23.3	FXYD domain containing ion transport regulator 6
204069_at	Hs.526754	MEIS1	chr2p14-p13	Meis homeobox 1
209651_at	Hs.513530	TGFB1I1	chr16p11.2	transforming growth factor β 1 induced transcript 1
210302_s_at	Hs.584852	MAB21L2	chr4q31	mab-21-like 2 (C. elegans)
203766_s_at	Hs.519075	LMOD1	chr1q32	leiomodin 1 (smooth muscle)
223468_s_at	Hs.271277	RGMA	chr15q26.1	RGM domain family, member A
202274_at	Hs.516105	ACTG2	chr2p13.1	actin, gamma 2, smooth muscle, enteric
207480_s_at	Hs.510989	MEIS2	chr15q14	Meis homeobox 2
229339_at	Hs.462257	—	chr17p11-p12	-
204797_s_at	Hs.501898	MRVI1	chr11p15	murine retrovirus integration site 1 homolog
225915_at	Hs.12451	EML1	chr14q32	echinoderm microtubule associated protein like 1
209708_at	Hs.87159	CAB39L	chr13q14.1-q14.3	calcium binding protein 39-like
1553808_a_at	Hs.6909	MOXD1	chr6q23.1-q23.3	monooxygenase, DBH-like1
213765_at	Hs.243272	NKX1–3	chr10q24.2	NK2 transcription factor related, locus 3 (Drosophila)
228640_at	Hs.460109	MYH11	chr16p13.11	myosin, heavy chain 11, smooth muscle
210302_s_at	Hs.479439	PCDH7	chr4p15	protocadherin 7

Table 3. Downregulated gene (n = 5) for experimental group vs. control group

Probe Set ID	UniGene ID	Gene Symbol	Location	Description
227462_at	Hs.482910	ERAP2	chr5q15	Endoplasmic reticulum aminopeptidase 2
205470_s_at	Hs.57771	KLK11	chr19q13.3-q13.4	kallikrein-related peptidase 11
229667_s_at	Hs.514292	HOXB8	chr17q21.3	homeobox B8
202888_s_at	Hs.1239	ANPEP	chr15q25-q26	alanyl (membrane) aminopeptidase
223509_at	Hs.522746	CLDN2	chrXq22.3-q23	claudin 2

Figure 1. Clustering display of microarray data and functional classifications of differentially expressed genes for experimental group vs. control group. Comparisons of 26 differentially expressed genes (30 ESTs) between the two groups were performed using the SAM software. Upon hierarchical clustering, they have been visualized with TreeView tools. Gene symbols are labeled on the right. Expression levels are represented by a color tag; red represents the highest while green is for the lowest levels of expression.

Nkx2–3 and TGFB1I1 prediction accuracies analysis

AUCs of Nkx2–3 and TGFB1I1 were 0.824 and 0.742(Fig. 2), cutoff values were 2.392 and 3.290, sensitivity were 1.000 and 0.769, specificities were 0.706 and 0.765, Youden indexes were 0.706 and 0.534, accuracies were 85.3% and 76.7%, p values were 0.003 and 0.025 respectively.

Figure 2. ROC analysis of Nkx2–3 and TGFB1I1. (A) AUC of Nkx2–3; (B) AUC of TGFB1I1.

Immunohistochemistry

Immunohistochemical analysis revealed that Nkx2–3 protein was observed on the nucleus of the cancer cells in 13 cancer tissues in experimental group, lower or no expression in 17 cancer tissues in control group. There was a statistically significant difference in expression levels between experimental group and control group (p < 0.01, Figure 3A). The result indicated that higher expression of Nkx2–3 was associated with the sensitivity of FOLFOX4 chemotherapy. TGFB1I1 protein was observed on the nucleus and cytoplasm of the cancer cells in 13 cancer tissues in experimental group, lower or no expression in 17 cancer tissues in control group. There was a statistically significant difference in expression levels between experimental group and control group (p < 0.01, Figure 3B). The result also indicated that higher expression of TGFB1I1 was associated with the sensitivity of FOLFOX4 chemotherapy.

Figure 3. Immunohistochemistry analysis of Nkx2–3 and TGFB1I1. (A) Nkx2–3 high expression in cancer tissues in experimental group (200 size); (B) Nkx2–3 no expression in cancer tissues in control group (200 size); (C) TGFB1I1 high expression in cancer tissues in experimental group (200 size); (D) TGFB1I1 no expression in cancer tissues in control group (200 size).

Discussion

It has been widely agreed that personalized treatment is the ideal approach to tumors. A necessary premise of personalized treatment is to screen its valuable biomarkers. FOLFOX4 chemotherapy has been used extensively in ACC patients, but the lack of prediction biomarks of the regime therapeutic effect prohibits ACC patients to achieve ideal personalized treatments. This was the first time that primary ACC patients who have undergone FOLFOX4 chemotherapy had their global gene expression analyzed. The 26 genes identified as differentially expressed almost involved various biological function. Some of these interesting genes may be related to CRC development, the body’s immunes, and drug chemotherapy.

There were two tumor suppressor genes (Nkx2–3 and TGFB1I1), who were the strongest prediction abilities in the upregulated genes.

Nkx2–3 was grouped primarily within the following two functional categories: (1) immune and inflammatory response and (2) cell proliferation, growth and oncogenesis. Ingenuity pathway analysis indicated that the affected top pathway included genes directly involved in Wnt signaling. Nkx2–3 could contribute to the pathogenesis of Inflammatory Bowel Disease (IBD)-associated CRC and sporadic CRC by regulating the Wnt signaling pathway.²⁸ Nkx2–3 might be a new tumor suppressor gene related to sporadic CRC.²⁹ However there is little research in this field now.

TGFB1I1 encodes a coactivator of the androgen receptor, a transcription factor which is activated by androgen and has a key role in male sexual differentiation.^30-32 The encoded protein is thought to regulate androgen receptor activity and adjust the epithelial cells of the growth process.^33-36 Multiple transcript variants encoding different isoforms have been found for this gene. TGFB1I1 different subtypes have different functions in platelet gathered, adhesion and movement and muscle cells generating process,^37-40 these effects perform mainly through integrin signaling pathways,⁴¹ serotonin, protein kinase C and TGFB1I1-associated redistribution of the platelet serotonin transporter.⁴² TGFB1I1 tyrosine phosphorylation functions to regulate signaling associated with lamellipodia formation, a process fundamental to cell motility.⁴³ TGFB1I1 was related to prostate cancer cell differentiation, and decrease of its expression can cause the tumor.^44,45 Long-term stimulation by tumor necrosis factor-α could increase TGFB1I1 expression in prostate cancer cells. The different expression of coactivators may contribute to the different response of prostate cancer to androgenic stimulation or endocrine therapy.^46,47 The encoded protein may have a role to play in the treatment of prostate cancer.⁴⁸ In addition, in colorectal cancer TGFB1I1 can reduce colorectal tumor cell growth rate and induce tumor cell apoptosis.^49,50 However there is little research on its mechanism of the action now.

Results from the microarray data and immunohistochemistry were not certified by qRT-PCR, and hence might be somewhat flawed. However, the study was based on the reliability of the Affymetrix chip technology combined with comprehensive bioinformatics analysis immunohistochemistry. Therefore, we believe the results are consistent. Clearly, more work is needed to provide molecular biology index for CRC treatment and prognosis based on differentially expressed genes, as well as in establishing a forecasting model for FOLFOX4 chemotherapy.

In summary, Nkx2–3 and TGFB1I1 expression in the control group are very low, but they are highly expressed in the experimental group. Nkx2–3 and TGFB1I1 may be classified as valuable biomarkers, as these can predict the effects of primary ACC patients who will undergo FOLFOX4. Future studies are required to address relationships between their signaling pathways and drug response.

Materials and methods

Eligibility criteria and patient evaluation

From 2008 to 2010, we gathered samples donated by primary ACC patients who had undergone FOLFOX4 chemotherapy. Thereafter, we carried on the primary tumor R0 resection. Clinical data was collected by the Department of Colorectal and Anal Surgery of the Affiliated Union Hospital of Fujian Medical University; this information was then organized as a database. The inclusion criteria were as follows: (1) histologically confirmed metastasis in CRC; (2) ages 18 to 80 y; (3) an Eastern Cooperative Oncology Group (ECOG) performance status of ≤ 1; and (4) patients have not received radiation, chemotherapy or palliative surgery before study registration. The exclusion criteria were as follows: (1) those who have nerves (around sensory neuropathy with dysfunction) and mental illnesses; (2) those with severe renal insufficiency (creatinine clearance is less than 30 mL/min); (3) pregnant and lactating women; and (4) those with another disease and hence cannot tolerate FOLFOX4 chemotherapy. The elimination criteria were as follows: (1) those with incomplete chemotherapy according to FOLFOX4 schedule standards or if they have terminated it halfway and (2) those with unfinished R0 resection of the primary tumor after chemotherapy. A tissue and patient's data usage protocol was approved by the Ethics Committee of the Affiliated Union Hospital of Fujian Medical University. Informed written consent was obtained from each patient. All patients underwent standard FOLFOX4 regimen chemotherapy in four cycles after signing the chemotherapy agreement; then, they were evaluated in accordance with the response evaluation criteria in solid tumors (RECIST).²⁷ Pretreatment evaluations included a complete medical history and physical examination. A baseline radiographic tumor evaluation was required within 4 weeks before study registration. The interim medical history, physical examination, and the laboratory studies listed herein were repeated before the start of each cycle of therapy. Tumor assessment was performed after the fourth cycle of therapy. Patients underwent R0 resection of the primary tumor after chemotherapy within 28 d.

Sample collection

Each sample was collected immediately following resection. Then, each sample was divided in half, one was fixed in formalin and embedded in paraffin, the other floated in ice-cold PBS and finally stored in liquid nitrogen until future total RNA extraction.

RNA extraction

Tissue homogenization and RNA extraction, together with subsequent microarray, was performed at the premises of CapitalBio Corporation. Tissue homogenization and isolation of total RNA were performed according to manufacturer’s instructions using the Macherey-Nagel NucleoSpin RNA II kit (Macherey-Nagel); this RNA isolation kit has significantly reduced the contamination of both genome DNA and proteins. The RNA, extracted with ribosomal 28S and 18S RNA with a ratio of intensities of 1.0–1.5:1, was used in subsequent microarray assay.

Microarray analysis

In comparing the differentially expressed genes of the two groups, an aliquot (2 μg) of total placental RNA was used to synthesize double-stranded cDNA, which was subsequently transcribed into biotin-tagged cRNA using the MessageAmp II aRNA Amplification Kit (Ambion). The cRNA was then fragmented to produce strands of 35–200 bases in length in accordance with protocols (Affymetrix). The fragmented cRNA was hybridized to the Affymetrix GeneChip Human Genome U133 Plus 2.0 Array containing 47,000 transcripts. Microarray hybridization was performed at 45°C with rotation lasting 16 h using the Affymetrix GeneChip Hybridization Oven 640. The arrays were washed and stained (streptavidin-phycoerythrin) by Affymetrix GeneChip Fluidics Station 450 and then scanned on the Affymetrix GeneChip Scanner 3000 in order to analyze the hybridization data. The obtained scanned images were assessed first by visual inspection and then analyzed by the Affymetrix GeneChip Operating Software (GCOS 1.4). To normalize the different arrays, the dChip software was used in global scaling procedures. In the comparative analysis, a two-class unpaired method obtained from the significance analysis of microarrays (SAM; version 3.02, Stanford University) software was used. Significant differentially expressed genes in both the experimental and control groups were compared.

Bioinformatics analysis

Signal log ratio and the gene symbol of differentially expressed genes were inputted in the MAS3.0 software for both gene ontology and pathway analyses.

Nkx2–3 and TGFB1I1 prediction accuracies analysis

Nkx2–3 and TGFB1I1 signal probe values transformed into signal log ratio using Biweight algorithm and ROC analysis to calculate Area Under The Curve (AUC), cutoff value, sensitivity, specificity, Youden index, judged prediction ability and calculated its own accuracy respectively.

Immunohistochemistry

Immunohistochemical studies of Nkx2–3 and TGFB1I1 were performed on surgical specimens of colorectal cancer, and tissue arrays using the avidin–biotin-peroxidase methods (DakoCytomation) on formalin-fixed, paraffin-embedded tissues. All sections were counterstained with hematoxylin. The primary mouse polyclonal antibody against Nkx2–3 (Santa Cruz Biotechnology, Inc.) and TGFB1I1 (Santa Cruz Biotechnology, Inc.) were used at dilutions of 1:100 and 1:200 respectively.

Abbreviations:
BSA	=	body surface area
CR	=	complete response
MD	=	moderately differentiated
PD	=	progressive disease
PDi	=	poorly differentiated
PR	=	partial response
SD	=	stable disease
TNM	=	tumor, nodes and metastases
WD	=	well-differentiated

Acknowledgments

The authors are grateful to the insightful work provided the leaders of Affiliated Union Hospital of Fujian Medical University, all of the fellows in Department of Colorectal and Anal Surgery and students that contributed with some of the studies discussed in this article. The source: National Natural Science Foundation of China. Project Approval number: 30872479; Number of grants: 300,000 RMB. Gainer: Pan Chi.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.