This article is mentioned in:

Introduction

Although tumor size is not a factor in deciding tumor stage according to TNM classification of the UICC (Union for International Cancer Control) or the Japanese classification of colorectal carcinoma, tumor size is generally considered to be an indicator of proliferation potency and thereby malignancy. There are some colorectal cancer (CRC) cases with small tumors and metastasis, and, conversely, tumors of large size without metastasis. Such small tumors with the capability to metastasize are estimated to have a tendency of vertical invasion and vascular invasion, and these tumors often have a high malignancy grade that induces lymph node or distant metastasis. Therefore, we hypothesized that such small but advanced cancer might have distinctive characteristics which are involved in cancer metastasis, particularly for the lymph nodes.

Several researchers have studied metastasis-regulating factors (1–3). Gene expression arrays using in vivo models (2) and clinical samples (1,2) and proteomics analysis using clinical samples (3) have all been carried out, and certain genes and pathways have been identified as biomarkers for CRC metastasis. There is some discord between past reports, however, therefore the underlying factors that cause tumor metastasis remain to be fully understood.

Lymph node metastasis is a significant factor that has an impact on disease prognosis in CRC cases. While patients without metastasis can mostly be cured by resection of the primary tumor and thus have a 5-year survival rate exceeding 80%, patients with lymph node metastasis often experience a relapse and therefore have a 5-year survival rate of <50% (4,5). We have, thus, focused on identifying biomarkers for CRC lymph node metastasis, which will be helpful in determining treatment strategy and may provide further insight into tumor biology.

In the present study, we investigated significant factors for cancer metastasis, particularly lymph node metastasis, by using gene expression microarray analysis of small tumors with metastases and large tumors without metastases.

Materials and methods

Patients and sample collection

We used a total of 320 CRC samples, in which 148 were used as pure cancer tissues separated by laser microdissection (71 cases for set 1 and all 148 cases for set 2) and 172 were used in bulk (set 3). All samples were obtained during surgery. All patients underwent resection of the primary tumor at Kyushu University Hospital at Beppu and affiliated hospitals between 1992 and 2007. Written informed consent was obtained from all patients. All patients had a clear histological diagnosis of CRC and were closely followed up every 3 months. The follow-up period in set 1 ranged from 0.1 to 12.3 years, with a mean of 3.8 years; follow up in set 2 ranged from 0.1 months to 3.2 years with a mean of 2.1 years. Resected cancer tissues were immediately cut and stored in RNAlater (Ambion) or embedded in Tissue-Tek OCT (optimum cutting temperature) medium (Sakura, Tokyo, Japan), frozen in liquid nitrogen, and kept at −80ºC until RNA extraction. Frozen tissue specimens were homogenized in guanidinium thiocyanate, and total RNA was obtained by ultracentrifugation through a cesium chloride cushion. cDNA for reverse-transcription PCR was synthesized from 8.0 μg of total RNA with M-MLV Reverse Transcriptase (Invitrogen, Carlsbad, CA, USA). Clinicopathological factors and clinical stage were classified using the TNM system of classification. All sample data, including age, gender, histology, tumor depth, lymph node metastasis, lymphatic invasion, vascular invasion, liver metastasis and postoperative liver recurrence, were obtained from the clinical and pathological records.

Laser microdissection

Tissue samples were microdissected using the LMD6000 Laser microdissection system (Leica Laser Microdissection System; Leica Microsystems, Wetzlar, Germany) as previously described (6). For laser microdissection, five micron frozen sections were fixed in 70% ethanol for 30 sec, stained with hematoxylin and eosin, and dehydrated for 5 sec each in 70, 95 and 100% ethanol. Sections were air-dried, then microdissected with the LMD system. Target cells were excised, at least 100 cells per section, and bound to the transfer film. Then, total RNA was extracted.

Gene expression microarray

We used the commercially available Human Whole Genome Oligo DNA Microarray kit (Agilent Technologies, Santa Clara, CA, USA). A list of genes on this cDNA microarray is available at http://www.chem.agilent.com/scripts/generic.asp?lpage=5175&indcol=Y&prodcol=Y&prodcol=N&indcol=Y&prodcol=N. Cyanine (Cy)-labeled cRNA was prepared using T7 linear amplification as described in the Agilent Low RNA Input Fluorescent Linear Amplification kit manual (Agilent Technologies). Labeled cRNA was fragmented and hybridized to an oligonucleotide microarray (Whole Human Genome 4×44K Agilent G4112F). Fluorescence intensities were determined with an Agilent DNA Microarray Scanner and were analyzed using G2567AA Feature Extraction software version A.7.5.1 (Agilent Technologies), which used the LOWESS (locally weighted linear regression curve fit) normalization method (7). This microarray study followed MIAME guidelines issued by the Microarray Gene Expression Data group (8).

Gene Ontology analysis

All 1662 genes which were significantly differentially expressed between the two groups were further analyzed using the Gene Ontology database (http://www.geneontology.org/).

Pathway analysis

Genesets and pathways mentioned in the Molecular Signatures Database (http://www.broadinstitute.org/gsea/msigdb/index.jsp), Kyoto Encyclopedia of Genes and Genomes (http://www.genome.jp/kegg/), SABioscience pathway central (http://www.sabiosciences.com/pathwaycentral.php) and Reactome (http://www.reactome.org/ReactomeGWT/entrypoint.html) were corrected and analyzed with the EEM (Extraction of Expression Modules) method (9).

Quantitative real-time reverse transcription-PCR

For quantitative real-time reverse transcription (qRT)-PCR, high mobility group A1 (HMGA1) (NM_145903.2, NM_002131.3, NM_145902.2, NM_145905.2, NM_145899.2, NM_145901.2) primer sequences were 5′-GAAAAGGACGGCACTGAGAA-3′ and 5′-CTCTTAGGTGTTGGCACTTCG-3′. To normalize for RNA concentration, glyceraldehyde-3-phosphate dehydrogenase (GAPDH) served as an internal control. The sequences of the GAPDH primers were: sense, 5′-TTGGTATCG TGGAAGGACTCA-3′ and antisense, 5′-TGTCATCATATT TGGCAGGTT-3′. The amplification protocol included an initial denaturation step at 95ºC for 10 min, followed by 45 cycles of 95ºC for 10 sec and 60ºC for 30 sec. qRT-PCR was performed in a LightCycler 480 instrument (Roche Applied Science, Basel, Switzerland) using the LightCycler 480 Probes Master kit (Roche Applied Science). All concentrations were calculated relative to the concentration of cDNA using Human Universal Reference Total RNA (Clontech, Palo Alto, CA, USA). The concentration of HMGA1 was then divided by the concentration of the endogenous reference (GAPDH) to obtain normalized expression values.

Statistical analysis

For gene expression array analysis, the differences between groups were estimated using the Student’s t-test after expression signals were calculated by log2-transformation of the normalized data. Differentially expressed genes were detected by using the P-value, fold-change value and q-value. All differences were considered statistically significant at the level of P<0.05 or false discovery rate (FDR) P<0.05. Data from RT-PCR analyses were analyzed using JMP 5 software (JMP, Cary, NC, USA). Overall survival rates were calculated actuarially according to the Kaplan-Meier method and were measured from the day of surgery. Differences between groups were estimated using the Chi-square test, Student’s t-test, repeated-measures ANOVA and log-rank test. A probability level of 0.05 was selected for statistical significance.

Results

A total of 1662 genes are differently expressed between the two groups

We first selected 71 cases (set 1) and subdivided them into two groups. Group 1 consisted of 6 colorectal cancers which were <2 cm in size and had metastasis; group 2 consisted of 65 cases with tumors >2 cm in size that lacked metastasis (Table I). A significant difference in expression level (FDR <0.05) between the two groups was found in 62 genes and P<0.05 was observed in 1662 genes (data not shown). These included some well described genes, such as the angiogenesis factor hypoxia inducible factor 1 (HIF1A); cell cycle regulators such as CDC34 and CD20; snail homolog 1 (SNAI1), which is involved in the epithelial to mesenchymal transition (EMT); the NF-κB pathway gene nuclear factor of κ light polypeptide gene enhancer in B-cells inhibitor (NFKBI); the oncogenic pathway gene RAS protein activator like 1 (RASAL1); and the colon cancer stem cell relating gene leucine-rich repeat containing G protein-coupled receptor 5 (LGR5).

Table I Clinicopathological factors in colorectal cancer cases of group 1 and group 2.

Table I

Clinicopathological factors in colorectal cancer cases of group 1 and group 2.

	Group 1 (N=6)	Group 2 (N=65)
Factors	n, %	n, %
Age (years)	57.5±10.8	67.0±11.0
Gender
Male	3 (50.0)	38 (58.0)
Female	3 (50.0)	27 (42.0)
Tumor size (mm)	19±2	49±19
Histological grade
Well	1 (16.7)	44 (67.7)
Moderate	5 (83.3)	21 (32.3)
Depth
M		1 (1.5)
SM		5 (7.7)
MP	1 (16.7)	14 (21.5)
SS, SE	5 (83.3)	41 (63.1)
SI		4 (6.2)
Lymph node metastasis
Present	6 (100.0)	0 (0.0)
Absent	0 (0.0)	65 (100.0)
Lymphatic invasion
Present	6 (100.0)	36 (55.4)
Absent	0 (0.0)	29 (44.6)
Venous invasion
Present	5 (83.3)	31 (47.7)
Absent	1 (16.7)	34 (52.0)
Liver metastasis
Present	1 (16.7)	0 (0.0)
Absent	5 (83.3)	65 (100.0)
Peritoneal dissemination
Present	0 (0.0)	0 (0.0)
Absent	6 (100.0)	65 (100.0)
Distant metastasis
Present	0 (0.0)	0 (0.0)
Absent	6 (100.0)	65 (100.0)

[i] Well, well differentiated tubular adenocarcinoma; Moderate, moderately differentiated tubular adenocarcinoma; M, mucosa; SM, submucosa; MP, muscularis propria; SE, serosa exposed; SS, subserosa; SI, serosa infiltrating.

Anti-apoptotic activity induces metastasis formation

We performed Gene Ontology analysis to annotate the 1662 genes which were significantly differently regulated in group 1. Results are shown in Table II. The Gene Ontologies ‘NF-κB binding’, ‘negative regulation of MAP kinase activity’, ‘regulation of epidermal growth factor receptor signaling pathway’ and ‘apoptotic process’ were significantly enriched in this set of genes.

Table II Significantly different Gene Ontology (GO) in 1662 significant genes.

Table II

Significantly different Gene Ontology (GO) in 1662 significant genes.

Category	GO ID	Name	P-value
Molecular function	GO:0005515	Protein binding	1.58E-08
	GO:0019901	Protein kinase binding	0.00892753
	GO:0019899	Enzyme binding	0.00489524
	GO:0005096	GTPase activator activity	0.00288377
	GO:0004843	Ubiquitin-specific protease activity	0.00878492
	GO:0003899	DNA-directed RNA polymerase activity	0.00575996
	GO:0051059	NF-κB binding	0.00691413
Biological process	GO:0006915	Apoptotic process	0.00842709
	GO:0006468	Protein phosphorylation	0.00921
	GO:0009615	Response to virus	0.00453601
	GO:0006974	Response to DNA damage stimulus	0.00822752
	GO:0006368	Transcription elongation from RNA polymerase II promoter	0.00160609
	GO:0043547	Positive regulation of GTPase activity	0.00107171
	GO:0042384	Cilium assembly	0.00586548
	GO:0050434	Positive regulation of viral transcription	0.00386628
	GO:0021987	Cerebral cortex development	0.00424518
	GO:0043407	Negative regulation of MAP kinase activity	0.00318089
	GO:0006446	Regulation of translational initiation	0.00468904
	GO:0006884	Cell volume homeostasis	0.00686783
	GO:0042058	Regulation of epidermal growth factor receptor signaling pathway	0.00861317
Cellular component	GO:0005829	Cytosol	0.00019253
	GO:0005730	Nucleolus	0.00011503
	GO:0005654	Nucleoplasm	0.00130461
	GO:0005794	Golgi apparatus	0.00767627
	GO:0005856	Cytoskeleton	0.00269772
	GO:0043234	Protein complex	0.00182768
	GO:0000242	Pericentriolar material	0.00386903

We performed pathway analysis to identify the pathways which actually determine the character of these tumors. A summary result of the analysis is shown in Table III. No significant difference of activation level was found in pathways associated with oncogenes such as RAS and cell-cycle genes. However, we identified significant inactivation of genes involved in the WNT signaling pathway and several apoptosis-related pathways in group 1. Inactivation of WNT signaling seemed to be associated with the smaller tumor size of group 1. The inactivation of apoptosis in group 1 tumors may contribute to their ability to form metastasis sites.

Table III Pathways differently activated between two groups.

Table III

Pathways differently activated between two groups.

	Average score
Gene set	Group 1	Group 2	P-value
HOLLMAN_APOPTOSIS_VIA_CD40_UP	−0.849748231	0.472445257	0.001910051
LAU_APOPTOSIS_CDKN2A_UP	−0.833935125	1.308750297	0.004621955
KEGG_PHOSPHATIDYLINOSITOL_SIGNALING_SYSTEM	−0.662010266	−0.386950146	0.007638363
KEGG_WNT_SIGNALING_ PATHWAY	−0.735968923	−0.413898949	0.010503868
TIAN_TNF_SIGNALING_VIA_NFKB	−0.584241501	0.67357094	0.026420951
BREDEMEYER_RAG_ SIGNALING_VIA_ATM_NOT_ VIA_NFKB_UP	−0.777042962	0.118487532	0.032926144
WONG_EMBRYONIC_STEM_CELL_CORE	0.402726415	−0.179854864	0.039863172
Apoptosis	−0.55085111	−0.263926672	0.046064516

We also carried out gene expression array for 77 more cases for further analysis. Using the microarray results for all 148 CRC cases, we investigated the correlation of the above 1662 genes with prognosis and the following clinicopathological factors: overall survival, disease-free survival, tumor size, histological grade, serosal invasion, lymph node metastasis, lymphatic invasion, venous invasion, peritoneal dissemination, liver metastasis and distant metastasis. Twenty-three genes, whose count of significant factors was over three, were extracted (Table IV). We focused on HMGA1 as it is reported to have a critical role in both neoplastic transformation and the inactivation of p53’s apoptotic function (10). HMGA1 had an FDR <0.05 in the analysis described above for the 1662 identified genes (data not shown).

Table IV Twenty-three genes significantly associated with more than three clinicopathological factors.

Table IV

Twenty-three genes significantly associated with more than three clinicopathological factors.

Gene	No. of significant factors	Overall survival	Disease- free survival	Tumor size (cut-off 2 cm)	Histological grade	Serosal invasion	Lymph node metastasis	Lymphatic invasion	Venous invasion	Peritoneal dissemination	Liver metastasis	Distant metastasis
ARHGDIG	6	*				*	*		*		*	*
IQSEC1	5	*	*						*	*	*
LRRC8E	5	*	*		*			*			*
IL17RD	4	*	*			*			*
RHBDF1	3						*	*	*
DBN1	4	*				*					*	*
ST3GAL4	4	*	*				*				*
ZNF296	4				*		*	*		*
UTP23	4	*	*			*					*
NUBPL	3							*	*		*
PTPN21	3	*						*			*
HIST1H2AL	3				*		*				*
TMEM17	3	*						*			*
TGFBR2	3		*								*	*
GLB1L2	3	*					*				*
DENND4C	3	*				*					*
C11orf51	3	*					*				*
MEGF8	3		*								*	*
PVT1	3						*	*	*
RECQL4	3					*	*		*
HMGA1	3	*			*		*
BCL11B	3	*			*						*
KIAA1429	3					*			*		*

* Significant factor (P-value <0.05).

HMGA1 mRNA expression is a robust indicator of lymph node metastasis and patient prognosis

HMGA1 mRNA expression in the bulk 172 tumor tissues and corresponding normal tissues (case set 3) was examined by qRT-PCR to validate the clinical significance of HMGA1 expression in CRC cases. HMGA1 mRNA levels in cancerous tissues were significantly higher than those in non-cancerous tissues (P<0.0001; Fig. 1A). The significant difference was maintained in the analysis of small sized tumors (<2 cm, P<0.0001; Fig. 1B).

Figure 1 HMGA1 mRNA expression level is a predictive indicator of prognosis and lymph node metastases in CRC cases. (A) Analysis of HMGA1 mRNA expression in colorectal tumor tissues (tumor) and corresponding normal mucosa (normal) by real-time RT-PCR for 90 CRC cases. HMGA1 mRNA was significantly upregulated in tumor tissues compared with normal mucosa (P<0.0001). (B) Analysis of HMGA1 mRNA expression in colorectal tumor tissues (tumor) and corresponding normal mucosa (normal) by real-time RT-PCR for 16 CRC cases with smaller tumor size (<20 mm). HMGA1 mRNA was significantly upregulated in tumor tissues compared with normal mucosa (P<0.0001), even in cases with smaller tumor size. (C) Kaplan-Meier analysis of overall survival for 172 patients with CRC according to HMGA1 mRNA expression. (D) Analysis of HMGA1 mRNA expression in colorectal tumor tissues for 10 submucosal invasive cases. HMGA1 mRNA was significantly higher in tumor tissues with lymph node metastasis than in cases without lymph node metastasis (P=0.0413). Lymph node metastasis (+), lymph node metastasis positive; lymph node metastasis (−), lymph node metastasis negative.

Next, we divided the 172 patients with CRC into a high HMGA1 expression group (n=86) and a low HMGA1 expression group (n=86), classified as having expression levels higher or lower than the median value, respectively. Clinicopathological factors were compared between the high and low HMGA1 mRNA expression groups (Table V). The high HMGA1 expression group showed higher risk for lymph node metastasis. Univariate analysis of lymph node metastasis revealed that the relative level of HMGA1 expression was a lymph node metastasis risk factor similar to serosal invasion, lymphatic invasion, venous invasion and liver metastasis (Table VI). Variables with a value of P<0.05 were selected for multivariate analysis. Multivariate analysis showed that HMGA1 expression was an independent lymph node metastasis risk factor in patients with CRC (relative risk, 3.46; P=0.001; Table VI). With regard to overall survival, patients with high HMGA1 expression had a significantly poorer prognosis than those with low HMGA1 expression (P=0.0046; Fig. 1C). Furthermore, we performed subgroup analysis for submucosal invasive cancer, which can be removed completely by local resection if lymph node metastases do not exist. Lymph node metastasis was observed in 3 out of 10 submucosal invasive cancer cases, and 3 cases with lymph node metastasis had significantly higher HMGA1 mRNA expression than the other 7 cases without lymph node metastasis (P=0.0413; Fig. 1D).

Table V HMGA1 mRNA expression and clinicopathological factors in 172 cases of colorectal cancer.

Table V

HMGA1 mRNA expression and clinicopathological factors in 172 cases of colorectal cancer.

	Low expression (N=86)	High expression (N=86)
Factors	n, %	n, %	P-value
Age (years)
<65	25 (29.07)	35 (40.70)	0.109
≥66	61 (70.93)	51 (59.30)
Gender
Male	52 (60.47)	51 (59.30)	0.8761
Female	34 (39.53)	35 (40.70)
Histological grade
Well/moderate	80 (93.02)	80 (93.02)	1
Other	6 (6.98)	6 (6.98)
Tumor size (mm)
≤20	4 (4.65)	11 (12.79)	0.0505
>20	78 (90.70)	70 (81.40)
Serosal invasion
Absent	27 (31.40)	23 (26.74)	0.5016
Present	59 (68.60)	63 (73.26)
Lymph node metastasis
N0	57 (66.28)	34 (39.53)	0.0003a
N1–2	28 (32.56)	52 (60.47)
Lymphatic invasion
Absent	56 (65.12)	44 (51.16)	0.076
Present	30 (34.88)	41 (47.67)
Venous invasion
Absent	70 (81.40)	64 (74.42)	0.3321
Present	16 (18.60)	21 (24.42)
Liver metastasis
Absent	79 (91.86)	75 (87.21)	0.3172
Present	7 (8.14)	11 (12.79)
Peritoneal dissemination
Absent	84 (97.67)	80 (93.02)	0.1389
Present	2 (2.33)	6 (6.98)
Distant metastasis
Absent	84 (97.67)	84 (97.67)	1
Present	2 (2.33)	2 (2.33)
UICC stage
0, I, II	54 (62.79)	34 (39.53)	0.0022a
III, IV	32 (37.21)	52 (60.47)

a Statistically significant; well, well differentiated tubular adenocarcinoma; moderate, moderately differentiated tubular adenocarcinoma; UICC, Union for International Cancer Control.

Table VI Univariate and multivariate analysis for lymph node metastasis (logistic regression model).

Table VI

Univariate and multivariate analysis for lymph node metastasis (logistic regression model).

	Univariate analysis			Multivariate analysis
Factors	RR	95% CI	P-value	RR	95% CI	P-value
Age (years) (<65/≥66)	0.39	0.204–0.744	0.0046	-	-	-
Gender (male/female)	0.92	0.498–1.706	0.7991	-	-	-
Serosal invasion (absent/present)	5.25	2.481–12.007	<0.0001a	2.80	1.177–7.040	0.023a
Lymphatic invasion (absent/present)	5.48	2.852–10.843	<0.0001a	2.83	0.297–6.253	0.009a
Venous invasion (absent/present)	5.83	2.575–14.578	<0.0001a	2.68	0.993–7.697	0.057
Liver metastasis (absent/present)	0.55	3.024–71.924	0.0017a	7.00	1.503–52.834	0.026a
HMGA1 mRNA expression (low/high)	3.11	1.679–5.882	0.0004a	3.46	1.676–7.832	0.001a

{ label (or @symbol) needed for fn[@id='tfn4-or-30-03-1488'] } RR, relative risk; CI, confidence interval;

a Statistically significant.

Discussion

In the present study, we focused on colorectal cancer (CRC) with metastasis in spite of small size (<2 cm). These cancer cells seem to have a greater ability for invasion and migration. Therefore, we comprehensively analyzed the gene expression profile of these cancer tissues to identify the genes or pathways which regulate the cancer metastasis.

Biomarkers for CRC lymph node metastasis have been reported as follows; Grade et al(1) carried out gene expression array analysis on 73 colon cancer tissues and comparative genomic hybridization (CGH) for 32 tumors. They identified 68 genes that were significantly differentially expressed between lymph node-negative and lymph node-positive tumors, the functional annotation of which revealed a preponderance of genes that play a role in cellular immune response and surveillance. Using an in vivo orthotopic CRC model and clinical samples, Hao et al(2) discovered that a five-gene signature [v-yes-1 Yamaguchi sarcoma viral related oncogene homolog (LYN), syndecan binding protein (SDCBP), mitogen-activated protein kinase kinase kinase kinase 4 (MAP4K4), dickkopf 1 homolog (DKK1), and midline 1 (MID1)] was closely correlated with lymph node metastasis in CRC. Using proteomics analysis, Meding et al(3) revealed that expression levels of FXYD domain containing ion transport regulator 3 (FXYD3), S100 calcium binding protein A11 (S100A11), and glutathione S-transferase mu 3 (GSTM3) are novel markers for regional lymph node metastasis in colon cancer. In addition, specific signatures associated with tumor stage and lymph node metastases were described (11,12).

In the present study, the FXYD5 and S100A genes were significantly regulated in group 1; however, we did not find most of the previously described markers. In the present study, we identified lymph node metastasis-related genes, particularly in small tumors (<20 mm), since we considered that they would be the most informative for clinical applications. This might explain the discrepancy between our results and past reports.

As for the characteristics of the identified genes, we found that anti-apoptotic activity played a key role in cancer metastasis. In general, deregulated cell proliferation together with suppressed apoptosis constitute the minimal common platform upon which tumorigenesis is based (13). The initial population of malignant cells avoids the apoptotic pathway and then continues to rampantly proliferate. Anti-apoptotic activity must, therefore, be required for cancer cells to form metastases. Notably, genes regulating the epithelial-to-mesenchymal transition related pathways, which have been considered to promote cancer cell invasion and migration (14,15), and those regulating cancer stem cell related pathways, which have the potential to initiate and sustain tumor growth and metastasis (16,17), did not affect metastatic ability in our study.

We showed that HMGA1 expression has a significant correlation with lymph node metastasis in CRC cases. The high-mobility group (HMG) proteins are low-molecular weight nuclear factors with non-histone chromosomal accessory functions (18). The A subgroup of HMG interacts with the minor groove of numerous AT-rich promoters and enhancers (19) and plays key roles in chromatin architecture and gene transcription control (19,20). Under physiological conditions, HMGA protein expression is high during embryogenesis (21,22) and decreases to low to undetectable levels in adult tissues. High HMGA expression in adult life is associated only with pathological conditions such as human carcinomas of the thyroid (23,24), colon (25–27), prostate (28), pancreas (29), cervix (30), ovary (31) and breast (32). Moreover, large scale gene expression studies show that high expression portends a poor prognosis in some tumors (33,34). HMGA1 is also enriched in embryonic stem cells and high grade (poorly differentiated) cancer, including breast, bladder, and brain cancer (34) and is associated with tumor invasion (10,35) and poorer clinical staging (36). HMGA1 overexpression induces inactivation of p53’s apoptotic function to escape apoptosis in neoplastic transformation (37,38) and drives stem cell properties in colon cancer cells (39). In our results, HMGA1 expression level was an indicator of poor prognosis in CRC cases and an independent risk factor for lymph node metastasis. We consider that these results are due to the anti-apoptotic function of HMGA1.

Intramucosal CRC generally does not metastasize to lymph nodes and is thus a good candidate for endoscopic local resection (40). By contrast, lymph node metastasis occurs in approximately 6–12% patients with submucosal invasive CRC, which requires surgical resection, including lymph node dissection, for curative treatment (41–44). Despite the low possibility of lymph node metastasis in submucosal invasive CRC, surgical resection and removal of regional lymph nodes are considered the standard treatment for this disease (45). It is noteworthy that tumors with lymph node metastasis had significantly higher HMGA1 expression levels in subgroup analysis for submucosal invasive CRC cases in our results. Perhaps submucosal invasive CRC cases without metastases, which might be cured by endoscopic local resection, could be extracted by HMGA1 expression level. Further studies using a larger number of cases are required.

In conclusion, our data strongly support the clinical significance of HMGA1 expression as a predictive indicator of lymph node metastasis in CRC cases, even in submucosal invasive CRC tumors.

Acknowledgements

The present study was supported in part by the following grants and foundations: CREST, Japan Science and Technology Agency (JST), and the Funding Program for Next Generation World-Leading Researchers (LS094). We would like to thank T. Shimooka, M. Kasagi and T. Kawano for their technical assistance.

References