Background: Insulin-like growth factor (IGF)–binding protein (IGFBP) isoforms have been implicated in the pathogenesis of human neoplasms including glioma. In view of this, we evaluated the expression of IGFBP isoforms (IGFBP-2, -3, and -5) during malignant progression of astrocytoma and their prognostic significance in glioblastoma.

Methods: The expression of IGFBP isoforms was analyzed in diffusely infiltrating astrocytomas by real-time quantitative PCR (n = 203) and immunohistochemistry (n = 256). Statistical methods were used to assess their grade-specific expression pattern and mRNA-protein intercorrelation. Survival analyses were done on a uniformly treated, prospective cohort of adult patients with newly diagnosed glioblastoma (n = 136) by using Cox regression models.

Results: The mean transcript levels of IGFBP-2 and -3 were significantly higher in glioblastomas (GBM) relative to anaplastic astrocytoma (AA), diffuse astrocytoma (DA), and controls whereas IGFBP-5 mRNA was higher in GBM relative to AA and controls (P < 0.05). By immunohistochemistry, the mean labeling index of all isoforms was significantly higher in GBM compared with AA, DA, and control (P < 0.05). A strong positive correlation was observed between their respective mRNA and protein expressions (P < 0.01). Multivariate analysis revealed IGFBP-3 expression (hazard ratio, 1.021; P = 0.030) and patient age (hazard ratio, 1.027; P = 0.007) to be associated with shorter survival in glioblastoma.

Conclusions: This study shows the associations of IGFBP-2, -3, and -5 expression with increasing grades of malignancy in astrocytomas. IGFBP-3 is identified as a novel prognostic glioblastoma biomarker. The strong correlation between their mRNA and protein expression patterns suggests their role in the pathogenesis of these tumors.

Impact: IGFBP isoforms have emerged as biomarkers with diagnostic and prognostic utility in astrocytomas. Cancer Epidemiol Biomarkers Prev; 19(6); 1399–408. ©2010 AACR.

This article is featured in Highlights of This Issue, p. 1387

Diffusely infiltrating astrocytomas are the most common intrinsic brain tumors in adults accounting for more than 60% of primary central nervous system neoplasms (1). They are further divided into three grades according to the current WHO classification scheme which includes diffuse astrocytoma (DA/grade II), anaplastic astrocytoma (AA/grade III), and glioblastoma (GBM/grade IV; ref. 2). Among them, GBM is the most malignant and the majority of affected patients exhibit rapid disease progression despite aggressive surgery, radiation, and chemotherapy with median survival times remaining in the 12- to 15-month range (3). However, the prognosis of individual patients is variable and ∼10% are known to survive for more than 2 years (4). At present, the established survival predictors in patients with glioblastoma are: O6-methylguanine-DNA methyltransferase promoter methylation status, age, Karnofsky performance score, extent of resection, and mini mental state examination (5, 6). However, the recent molecular and genetic profiling studies have provided a better understanding of prognostic and predictive factors of GBM along with underlying mechanisms of resistance to standard therapies (3).

Recently, the insulin-like growth factor (IGF) signaling pathway has been implicated in the progression of glioma (7, 8). The IGF signaling axis involves the complex coordinated actions of two ligands (IGF-I and IGF-II), three cell surface receptors (IGF-IR, IGF-IIR, and IR), IGF-binding proteins IGFBPs, and the proteases that affect the binding proteins. These interactions are thought to influence IGF bioavailability (7, 9). The mitogenic effects of IGFs are mainly mediated through interactions with IGF-IR followed by the activation of its downstream effector molecules.

Earlier studies have shown that human gliomas overexpress IGFs and IGFRs when compared with the normal brain (10, 11). Among the IGFBPs, the expression of isoforms-2, -3, and -5 have been studied in gliomas. IGFBP-2 overexpression has been reported in GBM (12-15), was found to increase the invasive potential of GBM cells (16), and has been associated with poor patient prognosis in GBM cases (15-17). A recent study reports increased plasma levels of IGFBP-2 to be associated with adverse prognosis in patients with high-grade glioma (18). Among other IGFBPs, increased expression of IGFBP-5 compared with IGFBP-3 has been reported in ascending grades of astrocytoma (19). In our previous cDNA microarray study, we have shown a differential expression of IGFBP-2 and -5 in astrocytoma tissues (20). In view of this and with emerging data on the role of some IGFBP isoforms in gliomas, the aim of this study was to evaluate the association of three IGFBP isoforms (IGFBP-2, -3, and -5) with the malignant progression of astrocytoma and assess their value as predictors of survival in patients with glioblastoma. To accomplish this, we have analyzed the expression of IGFBP-2, -3, and -5 both at the mRNA and protein level on tumor tissues of diffusely infiltrating astrocytoma, including DA, AA, and GBM. A subset of GBM tissues were from a large cohort of patients with newly diagnosed glioblastoma subjected to uniform standard therapeutic protocol and followed up prospectively. In this subset, the protein expressions of IGFBP-2, -3, and -5 in the tumor tissue were correlated with patient survival.

Patient and tissue samples

Samples of diffusely infiltrating astrocytoma of different grades (DA, AA, and GBM) were obtained from patients who underwent surgery at the National Institute of Mental Health and Neurosciences and Sri Sathya Sai Institute of Higher Medical Sciences, Bangalore, India. Control samples comprised a portion of anterior temporal cortex resected from patients who underwent surgery for intractable epilepsy. Tissues were freshly received from the neurosurgical operating rooms, bisected, and one half was placed in RNAlater (Ambion, Inc.), stored at −70°C and used for RNA isolation. The other half was fixed in 10% buffered neutral formalin, processed for paraffin sections, and was used for histopathology and immunohistochemistry. The mRNA expression of IGFBP-2, -3, and -5 was analyzed by real-time quantitative PCR (qPCR) on 223 samples (16 DA, 50 AA, 137 GBM, and 20 control brain tissues). The expression of these molecules at the protein level was analyzed on tumor tissue sections by immunohistochemistry on 268 samples (15 DA, 16 AA, 225 GBM, and 12 control brain tissues). Among these, 142 samples were analyzed in common by both qPCR and immunohistochemistry (3 DA, 10 AA, 125 GBM, and 4 control brain tissues). Among the GBM samples, 136 were from a set of patients with newly diagnosed glioblastoma, subjected to uniform treatment and followed up prospectively. In this cohort, both qPCR and immunohistochemistry was done on 102 samples and the rest were analyzed by immunohistochemistry only.

Prospective study of patients with newly diagnosed glioblastoma

The study has been scrutinized and approved by the ethics committee of two clinical centers (National Institute of Mental Health and Neurosciences and the Sri Sathya Sai Institute of Higher Medical Sciences) and patient consent was obtained prior to initiation of the study. In this prospective study, patients with newly diagnosed glioblastoma (n = 136) who underwent surgery in these two clinical centers between July 2006 to February 2009 were studied. Inclusion criteria for the study were (a) adult patients between the ages of 18 and 65 years with supratentorial lobar tumor, (b) patients who underwent macroscopic total/near total resection of the tumor as assessed by postoperative magnetic resonance imaging scan, and (c) patients with a postoperative Karnofsky's performance score of ≥70. The demographic data and clinical features of these patients were collected. All patients were treated uniformly with radiotherapy and chemotherapy. Radiotherapy was administered at a total dose of 59.4 Gy, given in 33 fractions along with concomitant chemotherapy with temozolomide, administered at a dose of 100 mg/d, which was continued daily for 45 days. Subsequently, five cycles of cyclical chemotherapy with temozolomide at a dose of 150 mg/m2 body surface area for 5 days every 28 days was administered.

The patients were followed up clinically including magnetic resonance imaging. The clinical status in the form of neurologic deficits and Karnofsky's performance score were documented regularly. Overall survival was defined as the duration between surgery and death of the patient due to disease. Outcome analysis was done in September 2009. All patients had completed the adjuvant therapy protocol and were either on follow-up or had expired at the time of evaluation. The maximum follow-up period was 34 months. The median survival of patients of the whole group was 17 months. The 2-year survival rate of this cohort was 32%.

Real-time qPCR assay

Total RNA was isolated from tissues using TRI reagent (Sigma-Aldrich) as per the manufacturer's protocol. The concentration of RNA was estimated by measuring the absorbance at 260 nm and integrity was verified on a denaturing 1% MOPS-formaldehyde agarose gel followed by ethidium bromide staining. The relative quantification of the mRNA transcripts and expression levels of IGFBP-2, -3, and -5 genes was carried out using a two-step strategy. In the first step, cDNA was generated from total RNA derived from different tissue samples using the cDNA synthesis kit (Applied Biosystems). Subsequently, real-time qPCR was carried out in the ABI PRISM 7900 sequence detection system (Applied Biosystems) with the cDNA as template (equivalent to 10 ng of RNA) and gene-specific primer sets for each of the IGFBP genes using the DyNAmo HS SYBR Green qPCR kit (Finnzymes, Finland). Data were analyzed according to the relative quantification model proposed by Pfaffl (21). All measurements were made in triplicate, and for each real-time qPCR primer set, the reaction efficiency estimates were derived from standard curves that were generated using serial dilutions of the pooled cDNA set used for the study. For normalization, the mean expression levels of four genes, ribosomal protein L35a (RP-L35a), 1-acylglycerol-3-phosphate O-acyltransferase 1 (AGPAT1), ATP synthase, H+ transporting, mitochondrial F0 complex, subunit C1 (subunit 9; ATP5G1), and glycyl-tRNA synthetase (GARS) were used as internal controls because their expression levels were found to be unaltered in our previous microarray data (20, 22). Control brain tissue samples (n = 20) from patients with epilepsy were individually used as reference. To assess grade-specific upregulation of IGFBP-2, -3, and -5 in diffusely infiltrating astrocytomas, qPCR was carried out on 223 samples that included 16 DA, 50 AA, and 137 GBM along with 20 control samples. The primer sequences for all the genes are shown in Table 1.

Table 1.

Sequences of the primers used for qPCR

Sequence no.GeneSequence (5′ → 3′)Primer lengthGenBank accession no.*Amplicon length
IGFBP2_F GACAATGGCGATGACCACTCA 21 NM_000597.2 (571-691)  
IGFBP2_R GCTCCTTCATACCCGACTTGA 21 120 
IGFBP3_F AGAGCACAGATACCCAGAACT 21 BC064987 (694-798)  
IGFBP3_R TGAGGAACTTCAGGTGATTCAGT 23 104 
IGFBP5_F CGGGGTTTGCCTCAACGAA 19 NM_000599 (1064-1180)  
IGFBP5_R TCTTGGGGGAGTAGGTCTCCT 21 116 
RPL35a_F ACGCCCGAGATGAAACAG 18 BC017093 (209-428)  
RPL35a_R GGGTACAGCATCACTCGGA 19 219 
AGPAT1_F TGTCCCCATAGTCATGTCCTC 21 NM_006411.2 (933-1105)  
AGPAT1_R GGAAAACAGTGAGCATGGAGT 21 172 
ATP5G1_F CCAGACGGGAGTTCCAGAC 19 NM_005175.2 (283-440)  
ATP5G1_R GACGGGTTCCTGGCATAGC 19 157 
GARS_F TCCTCTGTTTGAAGGGCAAG 20 NM_002047.1 (2525-2710)  
GARS_R AAATCAGGCAGCCACTGAAG 20 185 
Sequence no.GeneSequence (5′ → 3′)Primer lengthGenBank accession no.*Amplicon length
IGFBP2_F GACAATGGCGATGACCACTCA 21 NM_000597.2 (571-691)  
IGFBP2_R GCTCCTTCATACCCGACTTGA 21 120 
IGFBP3_F AGAGCACAGATACCCAGAACT 21 BC064987 (694-798)  
IGFBP3_R TGAGGAACTTCAGGTGATTCAGT 23 104 
IGFBP5_F CGGGGTTTGCCTCAACGAA 19 NM_000599 (1064-1180)  
IGFBP5_R TCTTGGGGGAGTAGGTCTCCT 21 116 
RPL35a_F ACGCCCGAGATGAAACAG 18 BC017093 (209-428)  
RPL35a_R GGGTACAGCATCACTCGGA 19 219 
AGPAT1_F TGTCCCCATAGTCATGTCCTC 21 NM_006411.2 (933-1105)  
AGPAT1_R GGAAAACAGTGAGCATGGAGT 21 172 
ATP5G1_F CCAGACGGGAGTTCCAGAC 19 NM_005175.2 (283-440)  
ATP5G1_R GACGGGTTCCTGGCATAGC 19 157 
GARS_F TCCTCTGTTTGAAGGGCAAG 20 NM_002047.1 (2525-2710)  
GARS_R AAATCAGGCAGCCACTGAAG 20 185 

*Numbers in parentheses denote the position of the amplicon in the GenBank sequence.

Intraexonic.

Interexonic.

Histopathology and immunohistochemistry

Histologic sections of normal brain and tumor tissues were examined by light microscopy using H&E stain. Tumor sections of diffusely infiltrating astrocytoma were graded using the WHO grading scheme (1, 2). Paraffin sections (4 μm) from the tumor tissue and control samples were collected on silane-coated slides and immunohistochemistry for the protein expression of IGFBP-2, -3, and -5 were carried out on 268 samples that included 12 control brain tissues, 15 DA, 16 AA, and 225 GBM tumors. The polyclonal IGFBP antibodies were from Santa Cruz Biotechnology. The details of the primary antibodies are IGFBP-2 (C-18: sc-6001 diluted to 1:100), IGFBP-3 (H-98: sc-9028, diluted to 1:50), and IGFBP-5 (H-100: sc-13093 diluted to 1:25). Antigen retrieval was done by heat treatment of the deparaffinized sections in a microwave oven for 25 to 35 minutes at 700 W in citrate buffer (10 mmol/L; pH 6.0). After the initial processing steps, sections were incubated overnight with primary antibody at 4°C. This was followed by incubation with the linked streptavidin-biotinylated secondary antibody (Universal LSAB, DAKO, Denmark) for IGFBP-2 and with supersensitive non-biotin horseradish peroxidase detection system (QD440-XAK, Biogenex) for the other antibodies. 3, 3′-Diaminobenzidine (Sigma-Aldrich) was used as the chromogenic substrate. GBM tumors that showed markedly increased mRNA levels of IGFBP-2, -3, and -5, respectively, by qPCR experiments served as positive controls. A negative control slide in which the primary antibody was excluded was incorporated with each batch of staining. A visual semiquantitative grading scale was applied to assess the intensity of the immunoreactivity as follows: zero (0) if the staining was absent, 1+ if it was weak, and 2+ if it was strong. Only 2+ staining intensity was considered for analysis. The immunopositivity of IGFBP-2, -3, and -5 was assessed in more than 1,000 cells from each tumor specimen. The IGFBP labeling index (LI) was expressed as a percentage of cells that showed 2+ positive staining among the total number of cells that were counted.

Statistical analysis

Grade-specific expression of IGFBP-2, -3, and -5 (mRNA and protein) in astrocytoma tissues

All continuous variables were tested for normal distribution and they were found to be non-normal. To determine the grade-specific upregulation, a nonparametric test, Kruskal-Wallis one-way ANOVA on ranks, followed by Dunn's post hoc test was done to adjust the α level to minimize the type I error rate. The results were expressed in the form of median and mean ± SD. P < 0.05 was considered to be statistically significant. Spearman rank-correlation coefficient was used to assess the pairwise correlation between mRNA and protein expression of IGFBP-2, -3, and -5 in all grades of astrocytoma and control tissues (n = 142; DA = 3, AA = 10, GBM = 125, control = 4).

Correlation of IGFBP-2, -3, and -5 protein expressions with survival in patients with newly diagnosed glioblastoma

Because the extent of surgical resection was uniform for all patients in this cohort (macroscopic total/near total resection), and postoperative Karnofsky's performance scores were uniformly ≥70, the only clinical variable included for analyses was patient age. The significance of continuous variables (patient age, IGFBP-2, -3, and -5 LIs) were assessed using univariate Cox regression models. Variables that were associated with survival in the univariate analyses (P < 0.10) were considered for potential inclusion into the multivariate model. Forward stepwise selection was used to determine the final Cox regression model. Results were reported using the P values and the estimated hazard ratio (HR) with their 95% confidence intervals. P < 0.05 was considered significant and all exact two-sided P values were reported.

Grade-specific expression pattern of IGFBP-2, -3, and -5: a distinct discrimination between GBM and other grades of astrocytoma

qPCR analysis of IGFBP-2, -3, and -5

The expression patterns of IGFBP-2, -3, and -5 across 223 samples that included 16 DA, 50 AA, 137 GBM tumors, and 20 control brain tissues using the SYBR Green–based qPCR assay was analyzed as described in Materials and Methods. We have chosen to use mean expression levels of four genes, i.e., RPL35a, AGPAT1, ATP5G1, and GARS for normalization of the gene expression data in our studies. These are not the standard normalization genes reported in the literature for glioma studies. We tested the expression levels of the standard genes such as glyceraldehyde-3-phosphate dehydrogenase and β-actin. These genes were found to have a regulated pattern between normals and tumors. Hence, we analyzed the microarray data and identified several unregulated genes that were consistent across all the tumors. These four genes were the most consistently unregulated genes between tumors compared with normal tissues. The microarray data were validated in large number of tumors by qPCR in comparison with control brain tissues (22). Because there are no standard reference normalization gene(s) reported for glioma, we propose that these genes could be used as routine normalization genes in glioma studies.

Significant differential mRNA expressions of IGFBP-2, -3, and -5 were observed in diffusely infiltrating astrocytomas (Fig. 1). Individual group differences in log 2 ratios were statistically significant as shown in the Table 2. The median (mean ± SD) log 2 ratio of IGFBP-2 in GBM was 4.815 (4.602 ± 1.724), whereas in AA, DA, and controls, it was 1.100 (1.210 ± 1.606), 0.789 (0.816 ± 1.195), and −0.116 (0.0000 ± 0.0.935), respectively. The comparison of median log 2 ratios between groups by Kruskal-Wallis one-way ANOVA followed by Dunn's post hoc test showed that the difference between GBM and control, GBM and DA, and GBM and AA were significant (P < 0.05).

Figure 1.

mRNA expression pattern of IGFBP-2, -3, and -5 in various grades of astrocytoma. Scatter plots showing the differentially regulated IGFBP isoform genes in diffusely infiltrating astrocytoma. Log 2-transformed gene expression ratios obtained from qPCR analyses are plotted for IGFBP-2, IGFBP-3, and IGFBP-5 on Y-axis against control, DA, AA, and GBM samples on the X-axis. A significant difference in the median log 2 ratios of IGFBP isoforms was observed between GBM, AA, DA, and control by Kruskal-Wallis one-way ANOVA followed by Dunn's post hoc test (P < 0.05). Bars, group medians.

Figure 1.

mRNA expression pattern of IGFBP-2, -3, and -5 in various grades of astrocytoma. Scatter plots showing the differentially regulated IGFBP isoform genes in diffusely infiltrating astrocytoma. Log 2-transformed gene expression ratios obtained from qPCR analyses are plotted for IGFBP-2, IGFBP-3, and IGFBP-5 on Y-axis against control, DA, AA, and GBM samples on the X-axis. A significant difference in the median log 2 ratios of IGFBP isoforms was observed between GBM, AA, DA, and control by Kruskal-Wallis one-way ANOVA followed by Dunn's post hoc test (P < 0.05). Bars, group medians.

Close modal
Table 2.

Individual group differences in the LI (immunohistochemistry) and log 2 ratios (qPCR) of IGFBP-2, -3, and -5

VariablesNormalDAAAGBMKruskal-Wallis test
χ2; PPost hoc P*
IGFBP-2 LI 0.0000 ± 0.0000 (0.0000) 0.0000 ± 0.0000 (0.0000) 2.500 ± 8.756 (0.0000) 30.070 ± 17.150 (30.000) 81.92; <0.0001 <0.05 
<0.05 
<0.05§ 
IGFBP-2 log 2 ratio 0.0000 ± 0.935 (−0.116) 0.816 ± 1.195 (0.789) 1.210 ± 1.606 (1.100) 4.602 ± 1.724 (4.815) 125.30; <0.0001 <0.05 
<0.05 
<0.05§ 
IGFBP-3 LI 0.0000 ± 0.0000 (0.0000) 6.333 ± 10.93 (0.0000) 12.500 ± 13.780 (10.000) 34.240 ± 17.020 (35.000) 73.53; <0.0001 <0.05 
<0.05 
<0.05§ 
IGFBP-3 log 2 ratio 0.0000 ± 1.017 (0.123) −0.624 ± 2.105 (−0.386) 0.866 ± 2.343 (0.819) 3.542 ± 2.123 (3.462) 89.00; <0.0001 <0.05 
<0.05 
<0.05§ 
IGFBP-5 LI 0.0000 ± 0.0000 (0.0000) 6.000 ± 10.390 (0.0000) 7.500 ± 10.000 (0.0000) 24.820 ± 13.680 (25.000) 64.09; <0.0001 <0.05 
<0.05 
<0.05§ 
IGFBP-5 log 2 ratio 0.0000 ± 1.038 (0.128) 1.039 ± 1.634 (1.621) 0.425 ± 1.818 (0.814) 1.844 ± 2.077 (1.909) 36.05; <0.0001 <0.05 
>0.05 
<0.05§ 
VariablesNormalDAAAGBMKruskal-Wallis test
χ2; PPost hoc P*
IGFBP-2 LI 0.0000 ± 0.0000 (0.0000) 0.0000 ± 0.0000 (0.0000) 2.500 ± 8.756 (0.0000) 30.070 ± 17.150 (30.000) 81.92; <0.0001 <0.05 
<0.05 
<0.05§ 
IGFBP-2 log 2 ratio 0.0000 ± 0.935 (−0.116) 0.816 ± 1.195 (0.789) 1.210 ± 1.606 (1.100) 4.602 ± 1.724 (4.815) 125.30; <0.0001 <0.05 
<0.05 
<0.05§ 
IGFBP-3 LI 0.0000 ± 0.0000 (0.0000) 6.333 ± 10.93 (0.0000) 12.500 ± 13.780 (10.000) 34.240 ± 17.020 (35.000) 73.53; <0.0001 <0.05 
<0.05 
<0.05§ 
IGFBP-3 log 2 ratio 0.0000 ± 1.017 (0.123) −0.624 ± 2.105 (−0.386) 0.866 ± 2.343 (0.819) 3.542 ± 2.123 (3.462) 89.00; <0.0001 <0.05 
<0.05 
<0.05§ 
IGFBP-5 LI 0.0000 ± 0.0000 (0.0000) 6.000 ± 10.390 (0.0000) 7.500 ± 10.000 (0.0000) 24.820 ± 13.680 (25.000) 64.09; <0.0001 <0.05 
<0.05 
<0.05§ 
IGFBP-5 log 2 ratio 0.0000 ± 1.038 (0.128) 1.039 ± 1.634 (1.621) 0.425 ± 1.818 (0.814) 1.844 ± 2.077 (1.909) 36.05; <0.0001 <0.05 
>0.05 
<0.05§ 

NOTE: Values in parentheses are the median.

*Dunn's post hoc P value.

Normal vs. GBM.

DA vs. GBM.

§AA vs. GBM.

For IGFBP-3, the median (mean ± SD) log 2 ratio in GBM was 3.462 (3.542 ± 2.123), whereas in AA, DA, and controls, it was 0.819 (0.866 ± 2.343), −0.386 (−0.624 ± 2.105), and 0.123 (0.0000 ± 0.1.017), respectively. The comparison of median log 2 ratios between groups by Kruskal-Wallis one-way ANOVA followed by Dunn's post hoc test showed that the differences between GBM and control, GBM and DA, and GBM and AA were significant (P < 0.05).

For IGFBP-5, the median (mean ± SD) log 2 ratio in GBM was 1.909 (1.844 ± 2.077) compared with AA (0.814; 0.425 ± 1.818), DA (1.621; 1.039 ± 1.634), and control (0.128; 0.0000 ± 1.038). The comparison of median log 2 ratios between groups by Kruskal-Wallis one-way ANOVA followed by Dunn's post hoc test showed that the difference between GBM and control and GBM and AA were significant (P < 0.05). However, the difference between GBM and DA was not statistically significant (P > 0.05).

Immunohistochemical analysis of IGFBP-2, -3, and -5

The protein expression patterns of IGFBP-2, -3, and -5 were analyzed by immunohistochemistry. Representative micrographs of immunohistochemical results in different grades of astrocytoma (DA, AA, and GBM) are shown in Fig. 2. The control brain sections stained negatively for these IGFBPs (data not shown). Among astrocytomas, the expressions of IGFBP-2, -3, and -5 were mostly confined to the cytoplasm of neoplastic astrocytic cells. The pattern of staining was intracytoplasmic, homogenous, and diffuse for IGFBP-2, -3, and -5 (Fig. 2: A-C1, D-F, and G-I, respectively). Interestingly, in GBM tumors, the cells lining pseudopalisading necrosis strongly expressed IGFBP-2 (Fig. 2-C2). Heterogeneity of staining was noted within individual tumor samples, which showed variations in intensity.

Figure 2.

Immunohistochemical staining pattern of IGFBP-2, -3, and -5. Representative micrographs showing staining of IGFBP-2 (A, B, C1, and C2), IGFBP-3 (D, E, and F), and IGFBP-5 (G, H, and I). DA staining (A, D, and G), AA staining (B, E, and H), and GBM staining (C1, C2, F, and I). C2 represents pseudopalisading necrosis region in a GBM section. All original magnifications are ×400 except in C2, which is ×100.

Figure 2.

Immunohistochemical staining pattern of IGFBP-2, -3, and -5. Representative micrographs showing staining of IGFBP-2 (A, B, C1, and C2), IGFBP-3 (D, E, and F), and IGFBP-5 (G, H, and I). DA staining (A, D, and G), AA staining (B, E, and H), and GBM staining (C1, C2, F, and I). C2 represents pseudopalisading necrosis region in a GBM section. All original magnifications are ×400 except in C2, which is ×100.

Close modal

A significant differential protein expression of IGFBP-2, -3, and -5 was observed in diffusely infiltrating astrocytomas (Fig. 3). Individual group differences in LIs were statistically significant as shown in Table 2. The median (mean ± SD) LI for IGFBP-2 in GBM was 30.000 (30.070 ± 17.150), whereas in AA, DA, and controls, it was 0.0000 (2.500 ± 8.756), 0.0000 (0.0000 ± 0.0000), and 0.0000 (0.0000 ± 0.0000), respectively. The comparison between groups by Kruskal-Wallis one-way ANOVA followed by Dunn's post hoc test showed that the differences between GBM and control, GBM and DA, and GBM and AA were significant (P < 0.05).

Figure 3.

Protein expression pattern of IGFBP-2, -3, and -5 in various grades of astrocytoma. A, B, and C, scatter plots showing differential expression of IGFBP isoforms in diffusely infiltrating astrocytoma. The LIs of different samples obtained from immunohistochemistry analyses are plotted for IGFBP-2, IGFBP-3, and IGFBP-5 (Y-axis) against control, DA, AA, and GBM samples (X-axis). A significant difference in the median LIs of IGFBP isoforms was observed between GBM, AA, DA, and control by Kruskal-Wallis one-way ANOVA followed by Dunn's post hoc test (P < 0.05). Bars, group medians.

Figure 3.

Protein expression pattern of IGFBP-2, -3, and -5 in various grades of astrocytoma. A, B, and C, scatter plots showing differential expression of IGFBP isoforms in diffusely infiltrating astrocytoma. The LIs of different samples obtained from immunohistochemistry analyses are plotted for IGFBP-2, IGFBP-3, and IGFBP-5 (Y-axis) against control, DA, AA, and GBM samples (X-axis). A significant difference in the median LIs of IGFBP isoforms was observed between GBM, AA, DA, and control by Kruskal-Wallis one-way ANOVA followed by Dunn's post hoc test (P < 0.05). Bars, group medians.

Close modal

For IGFBP-3, the median (mean ± SD) LI in GBM was 35.000 (34.240 ± 17.020), in comparison with AA (10.000; 12.500 ± 13.780), DA (0.0000; 6.330 ± 10.930), and control (0.0000; 0.0000 ± 0.0000). The differences between GBM and control, GBM and DA, and GBM and AA were statistically significant (P < 0.05).

For IGFBP-5, the median (mean ± SD) LI in GBM was 25.000 (24.820 ± 13.680) in comparison with AA (0.0000; 7.5008.000 ± 10.000), DA (0.0000; 6.000 ± 10.390), and normal (0.0000; 0.0000 ± 0.0000). Dunn's post hoc test for multiple comparisons showed that the differences between GBM and control, GBM and DA, and GBM and AA were significant (P < 0.05).

Overexpression of IGFBP-2, -3, and -5 mRNA as well as the protein was seen in a grade-dependent manner in diffusely infiltrating astrocytomas. A strong positive correlation was observed between the mRNA and protein expressions of IGFBP-2 (ρ = 0.498; P < 0.001), IGFBP-3 (ρ = 0.215; P < 0.010), and IGFBP-5 (ρ = 0.500; P < 0.001; Fig. 4).

Figure 4.

A scatter plot showing a positive correlation between the mRNA and protein expressions of IGFBP isoforms: Spearman's rank correlation coefficient for (A) IGFBP-2, ρ = 0.498; P < 0.001; (B) IGFBP-3, ρ = 0.215; P < 0.010; and (C) IGFBP-5, ρ = 0.500; P < 0.001. The linear regression line that best fits the data is shown.

Figure 4.

A scatter plot showing a positive correlation between the mRNA and protein expressions of IGFBP isoforms: Spearman's rank correlation coefficient for (A) IGFBP-2, ρ = 0.498; P < 0.001; (B) IGFBP-3, ρ = 0.215; P < 0.010; and (C) IGFBP-5, ρ = 0.500; P < 0.001. The linear regression line that best fits the data is shown.

Close modal

Effect of age, IGFBP-2, -3, and -5 protein expression on survival in patients with newly diagnosed glioblastoma

Univariate Cox regression models showed statistical significance for the continuous variables age, IGFBP-2, -3, and -5 LIs (Table 3A). Patient age (HR, 1.026; P = 0.008), IGFBP-2 LI (HR, 1.016; P = 0.028), IGFBP-3 LI (HR, 1.023; P = 0.016), and IGFBP-5 LI (HR, 1.018; P = 0.052) were all associated with worse clinical outcome.

Table 3.

Results of Cox regression analyses

VariableHR (95% CI)P
(A) Univariate Cox regression analysis 
Age (years) 1.026 (1.007-1.046) 0.008 
IGFBP-2 LI 1.016 (1.002-1.030) 0.028 
IGFBP-3 LI 1.023 (1.004-1.043) 0.016 
IGFBP-5 LI 1.018 (1.000-1.036) 0.052 
 
(B) Multivariate Cox regression analysis 
Age (years) 1.027 (1.007-1.047) 0.007 
IGFBP-3 LI 1.021 (1.002-1.041) 0.030 
VariableHR (95% CI)P
(A) Univariate Cox regression analysis 
Age (years) 1.026 (1.007-1.046) 0.008 
IGFBP-2 LI 1.016 (1.002-1.030) 0.028 
IGFBP-3 LI 1.023 (1.004-1.043) 0.016 
IGFBP-5 LI 1.018 (1.000-1.036) 0.052 
 
(B) Multivariate Cox regression analysis 
Age (years) 1.027 (1.007-1.047) 0.007 
IGFBP-3 LI 1.021 (1.002-1.041) 0.030 

The variables age, IGFBP-2, -3, and -5 LIs were associated with survival in the univariate analysis at the predetermined 0.1 significance level, and were therefore considered for inclusion in the multivariate Cox model. By enter method, it was observed that patient age was the most significant independent predictor of poor outcome (HR, 1.026; P = 0.010). Subsequently, using forward stepwise (Wald) selection of variables that determines the final multivariate Cox model, it was noted that patient age (HR, 1.027; P = 0.007) and IGFBP-3 LI (HR, 1.021; P = 0.030) predicted worse clinical outcome (Table 3B).

This family of IGFBPs comprises six members (IGFBP-1 to IGFBP-6) which bind and regulate the functions of IGFs. They modulate the bioavailability of IGFs, and in turn, regulate tumor growth and invasion (7, 23). Among the various isoforms, the expression and role of IGFBP-2 in gliomas has been extensively studied (12-18, 24), and there are few reports on the expression pattern of IGFBP-3 and -5 in gliomas (19, 25). The present study shows a grade-specific differential overexpression of IGFBP-2, -3, and -5 in diffusely infiltrating astrocytoma tissues along with a linear correlation between their mRNA and protein expressions. More importantly, their prognostic significance is established on a large cohort of patients with newly diagnosed glioblastoma who received standard therapy including, surgery, radiation and temozolamide as per the accepted international norms (26), and followed up.

Grade-specific expression pattern of IGFBP-2, -3, and -5

Analysis of the expression pattern of IGFBP-2, -3, and -5 in the present study clearly establishes a statistically significant grade-specific overexpression of all the three isoforms in diffusely infiltrating astrocytomas. It is noteworthy that the protein expression of these isoforms, as determined by immunohistochemistry, showed significant concordance with their corresponding mRNA expression as observed by qPCR results. This suggests a possible role for these isoforms in the pathogenesis and malignant progression of astrocytomas.

Gene expression profiling and immunohistochemistry experiments have shown IGFBP-2 to be highly expressed in GBM tumors when compared with low-grade glioma and normal brain, thus correlating the expression of this biomarker with the aggressive growth of this tumor (12-16, 24). In accordance with this, Marucci et al. have shown that gene and protein expression of IGFBP-2 correlates with the overexpression of CDC-20 and suggest the use of these two markers in combination for the diagnosis of GBM in small tissue biopsies (27). IGFBP-2 has also been shown to enhance matrix metalloproteinase-2 gene expression resulting in an increased invasion of GBM tumor cells (16). In another report, IGFBP-2 and platelet-derived growth factor-β together have been shown to be responsible for the progression of diffuse gliomas into a malignant phenotype (28). The present study shows a statistically significant overexpression of IGFBP-2 transcript as well as protein in GBM tumors when compared with DA, AA, and control brain. We have noted IGFBP-2 immunoreactivity within the cytoplasm of tumor cells, in agreement with the previous studies (12, 14, 27). In GBM tumors that had pseudopalisading necrosis, a strong IGFBP-2 staining was observed in the perinecrotic cells. In an earlier study, IGFBP-2 was shown to be colocalized with vascular endothelial growth factor receptor by in situ hybridization in perinecrotic cells surrounding pseudopalisading necrosis (29), indicating a possible involvement of IGFBP-2 in tumor angiogenesis. It is well established that hypoxic conditions result in necrosis within the core of the tumor. The observation that IGFBP-2 staining is seen in the perinecrotic area suggests the induction of this gene by hypoxia. Indeed, induction of IGFBP-2 has been shown in hypoxic conditions including hypoxic injury to the rat brain (30), giving strength to the argument that IGFBP-2 is expressed in GBMs under hypoxic conditions. Other experimental studies have shown that IGFBP-2 overexpression increases the invasion of glioma cells (16). Taken together, IGFBP-2 seems to be an important hypoxia-induced gene which is essential for tumor, angiogenesis, and invasion.

IGFBP-3 has been shown to have dual roles in tumor behavior. Both proapoptotic/antiproliferative and growth promotion actions have been proposed (31, 32). IGFBP-3 has been identified as a hypoxia-induced gene, in a study on a malignant glioma cell line (U-251) by cDNA microarray analysis (25). In this study, the mRNA and protein expression of IGFBP-3 showed a differential overexpression in diffusely infiltrating astrocytomas with ascending grades of malignancy, with the highest expression in GBM tissues. This suggests a plausible role for IGFBP-3 in tumor aggressiveness and malignant progression of astrocytomas. Our results differ from a previous observation in which the protein expression of IGFBP-5, but not IGFBP-3, was shown to increase with glioma anaplastic progression (19). However, in a recent report on the expression profiling of genes in GBM tumors, IGFBP-3 was classified as an upregulated gene (33). A strong cytoplasmic expression of IGFBP-3 was observed within the neoplastic astrocytes, which is similar to the reported cytoplasmic localization in human breast cancers (34). The present study is the first of its kind to show a strong cytoplasmic overexpression of IGFBP-3 in a grade-specific manner in astrocytomas.

We observed that the relative mRNA expression of IGFBP-5 was significantly higher in GBM tissues in comparison to control and AA tissues, albeit statistically not significant relative to DA tissues. However, its protein expression showed significant overexpression with ascending grades of malignancy in astrocytomas. This is in agreement with the previous report of IGFBP-5 expression in gliomas, which has shown its overexpression with increasing grades of malignancy (19). In an earlier study involving the application of a protein lysate array, IGFBP-5 was found to be overexpressed in GBM and was noted to be closely clustered with IGFBP-2, indicating that both proteins might contribute to glioma invasion and/or other common functions (35).

IGFBP3 is a strong independent predictor of survival in patients with newly diagnosed glioblastoma

In the present study, we have correlated the protein expression of IGFBP-2, -3, and -5 with survival on a large prospective and uniformly treated cohort of adult patients with newly diagnosed glioblastoma. The expression of these isoforms was found to be significantly associated with shorter survival by univariate Cox regression analysis. Subsequently, multivariate Cox proportional hazards model showed IGFBP-3 expression (HR, 1.021; P = 0.030) and patient age (HR, 1.027; P = 0.007) to be associated with worse clinical outcome in patients with glioblastomas.

The present study is the first of its kind to show that IGFBP-3 protein overexpression is significantly associated with shorter survival in human glioblastomas. The prognostic significance of IGFBP-3 in glioblastoma has not been previously explored. Interestingly, IGFBP-3 overexpression has been shown to be associated with poor prognosis in patients with breast cancer (36). Further strengthening this observation, there is mounting evidence through in vitro studies that the potential mitogenicity of IGFBP-3 is mediated through its interactions with epidermal growth factor receptor and RAS-p44/42 mitogen-activated protein kinase signaling in breast epithelial cells (37).

The prognostic value of IGFBP-2 has been proposed in earlier studies (15, 17, 18). It has been reported that IGFBP-2, along with IQGAP-1, is an important prognostic biomarker in high-grade astrocytic neoplasms (17). Importantly, plasma levels of IGFBP-2 have been correlated with recurrence and progression-free survival in high-grade gliomas (18). By using a larger prospective cohort of patients who received uniform therapy, this study establishes IGFBP-2 as an important biomarker in predicting poor prognosis in patients with newly diagnosed glioblastoma, although not an independent biomarker for predicting survival. Therefore, the value of IGFBP-2 as a predictor of survival could be more significant in conjunction with other molecular markers as shown in previous reports (17).

The expression level of IGFBP-5 is known to contribute to the development of breast cancer and has been proved to be a prognostic factor in breast cancers (38). Wang et al. have shown IGFBP-2 and IGFBP-5 to be useful markers to predict lymph node metastasis in patients with small (T1) invasive breast carcinomas (39). In the present study, IGFBP-5 showed a lower predictive value than the other two isoforms. However, its prognostic potential could improve when considered in combination with other molecular markers.

In conclusion, the results of the present study indicate that the differential overexpression of IGFBP-2, -3, and -5 with increasing grades of malignancy and the strong interdependence between their mRNA and protein expression patterns suggest their possible role in pathogenesis and malignant progression in diffusely infiltrating astrocytomas. Such a high correlation warrants the analysis of gene copy number for these IGFBP isoforms in astrocytomas. Most importantly, IGFBP-3 protein overexpression in tumor tissues emerges as a strong independent predictor of shorter survival in patients with newly diagnosed glioblastoma. Further studies are required to elucidate their functional role to exploit these isoforms as candidates for new therapeutic targets.

No potential conflicts of interest were disclosed.

We thank K. Chandrashekar, K. Prem, A.R. Ananthalakshmi, and K. Manjunath for the help with collection of tumor samples, patient coordination, and technical assistance.

Grant Support: CSIR, under the NMITLI program. K. Somasundaram is a Wellcome Trust International Senior Research Fellow. Infrastructural support by funding from ICMR (Center for Advanced Studies in Molecular Medicine), to the Department of Microbiology and Cell Biology; Department of Biotechnology, Department of Science and Technology (Fund for Improvement of S&T Infrastructure in Higher Educational Institutions) and University Grants Commission to Departments of Microbiology and Cell Biology and Molecular Reproduction, Development and Genetics, Indian Institute of Science is acknowledged.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1
Kleihues
P
,
Louis
DN
,
Scheithauer
BW
, et al
. 
The WHO classification of tumors of the nervous system
.
J Neuropathol Exp Neurol
2002
;
61
:
215
25
.
2
Kleihues
P
,
Louis
DN
,
Wiestler
OD
,
Burger
PC
,
Schiethauer
BW
. 
WHO grading of tumors of central nervous system
. In:
Louis
DN
,
Ohgaki
H
,
Wiestler
OD
,
Cavenee
WK
, editors.
WHO classification of tumors of the central nervous system
. 4th ed.
Lyon (France)
:
IARC Press
; 
2007
, p.
10
11
.
3
Palanichamy
K
,
Erkkinen
M
,
Chakravarti
A
. 
Predictive and prognostic markers in human glioblastomas
.
Curr Treat Options Oncol
2006
;
7
:
490
504
.
4
Ho
DM
,
Hsu
CY
,
Ting
LT
,
Chiang
H
. 
MIB-1 and DNA topoisomerase IIα could be helpful for predicting long-term survival of patients with glioblastoma
.
Am J Clin Pathol
2003
;
119
:
715
22
.
5
Gorlia
T
,
van den Bent
MJ
,
Hegi
ME
, et al
. 
Nomograms for predicting survival of patients with newly diagnosed glioblastoma: prognostic factor analysis of EORTC and NCIC trial 26981-22981/CE.3
.
Lancet Oncol
2008
;
9
:
29
38
.
6
Adamson
C
,
Kanu
OO
,
Mehta
AI
, et al
. 
Glioblastoma multiforme: a review of where we have been and where we are going
.
Expert Opin Investig Drugs
2009
;
18
:
1061
83
.
7
Wang
H
,
Fuller
GN
,
Zhang
W
. 
Insulin-like growth factors and insulin-like growth factors binding proteins in CNS tumors
. In:
Zhang
W
,
Fuller
GN
, editors.
Genomic and molecular neuro-oncology
.
Sudbury
:
Jones and Bartlett Publishers
; 
2004
, p.
119
30
.
8
Trojan
J
,
Cloix
JF
,
Ardourel
MY
,
Chatel
M
,
Anthony
DD
. 
Insulin-like growth factor type I biology and targeting in malignant gliomas
.
Neuroscience
2007
;
145
:
795
811
.
9
Riedemann
J
,
Macaulay
VM
. 
IGF1R signalling and its inhibition
.
Endocr Relat Cancer
2006
;
13
:
S33
43
.
10
Hirano
H
,
Lopes
MB
,
Laws
ER
 Jr.
, et al
. 
Insulin-like growth factor-1 content and pattern of expression correlates with histopathologic grade in diffusely infiltrating astrocytomas
.
Neuro-oncol
1999
;
1
:
109
19
.
11
Zumkeller
W
,
Westphal
M
. 
The IGF/IGFBP system in CNS malignancy
.
Mol Pathol
2001
;
54
:
227
9
.
12
Fuller
GN
,
Rhee
CH
,
Hess
KR
, et al
. 
Reactivation of insulin-like growth factor binding protein 2 expression in GBM multiforme: a revelation by parallel gene expression profiling
.
Cancer Res
1999
;
59
:
4228
32
.
13
Wang
H
,
Wang
H
,
Zhang
W
,
Fuller
GN
. 
Tissue micro arrays: applications in neuropathology research, diagnosis and education
.
Brain Pathol
2002
;
12
:
95
107
.
14
Elmlinger
MW
,
Deininger
MH
,
Schuett
BS
, et al
. 
In vivo expression of insulin-like growth factor-binding protein-2 in human glioma increases with the tumor grade
.
Endocrinology
2001
;
142
:
1652
8
.
15
Sallinen
SL
,
Sallinen
PK
,
Haapasalo
HK
, et al
. 
Identification of differentially expressed genes in human gliomas by DNA microarray and tissue chip techniques
.
Cancer Res
2000
;
60
:
6617
22
.
16
Wang
H
,
Wang
H
,
Shen
W
, et al
. 
Insulin-like growth factor binding protein 2 enhances GBM invasion by activating invasion-enhancing genes
.
Cancer Res
2003
;
63
:
4315
21
.
17
McDonald
KL
,
O'Sullivan
MG
,
Parkinson
JF
, et al
. 
IQGAP1 and IGFBP-2: valuable biomarkers for determining prognosis in glioma patients
.
J Neuropathol Exp Neurol
2007
;
66
:
405
17
.
18
Lin
Y
,
Jiang
T
,
Zhou
K
, et al
. 
Plasma IGFBP-2 levels predict clinical outcomes of patients with high-grade gliomas
.
Neuro-oncol
2009
;
11
:
468
76
.
19
Wang
H
,
Wang
H
,
Zhang
W
,
Fuller
GN
. 
Over expression of IGFBP5, but not IGFBP3, correlates with the histologic grade of human diffuse glioma: a tissue microarray and immunohistochemical study
.
Technol Cancer Res Treat
2006
;
5
:
195
200
.
20
Reddy
SP
,
Britto
R
,
Vinnakota
K
, et al
. 
Novel GBM markers with diagnostic and prognostic value identified through transcriptome analysis
.
Clin Cancer Res
2008
;
14
:
2978
87
.
21
Pfaffl
MW
. 
A new mathematical model for relative quantification in real-time RT-PCR
.
Nucleic Acids Res
2001
;
29
:
e45
.
22
Somasundaram
K
,
Reddy
SP
,
Vinnakota
K
, et al
. 
Upregulation of ASCL1 and inhibition of Notch signaling pathway characterize progressive astrocytoma
.
Oncogene
2005
;
24
:
7073
83
.
23
Vincent
AM
,
Feldman
EL
. 
Control of cell survival by IGF signaling pathways
.
Growth Horm IGF Res
2002
;
12
:
193
7
.
24
Song
SW
,
Fuller
GN
,
Khan
A
, et al
. 
IIp45, an insulin-like growth factor binding protein 2 (IGFBP-2) binding protein, antagonizes IGFBP-2 stimulation of glioma cell invasion
.
Proc Natl Acad Sci U S A
2003
;
100
:
13970
5
.
25
Ragel
BT
,
Couldwell
WT
,
Gillespie
DL
,
Jensen
RL
. 
Identification of hypoxia-induced genes in a malignant glioma cell line (U-251) by cDNA microarray analysis
.
Neurosurg Rev
2007
;
30
:
181
7
.
26
Stupp
R
,
Mason
WP
,
van den Bent
MJ
, et al
. 
Radiotherapy plus concomitant and adjuvant temozolomide for glioblastoma
.
N Engl J Med
2005
;
352
:
987
96
.
27
Marucci
G
,
Morandi
L
,
Magrini
E
, et al
. 
Gene expression profiling in glioblastoma and immunohistochemical evaluation of IGFBP-2 and CDC20
.
Virchows Arch
2008
;
453
:
599
609
.
28
Dunlap
SM
,
Celestino
J
,
Wang
H
, et al
. 
Insulin-like growth factor binding protein 2 promotes glioma development and progression
.
Proc Natl Acad Sci U S A
2007
;
104
:
11736
41
.
29
Godard
S
,
Getz
G
,
Delorenzi
M
, et al
. 
Classification of human astrocytic gliomas on the basis of gene expression: a correlated group of genes with angiogenic activity emerges as a strong predictor of subtypes 1,2
.
Cancer Res
2003
;
63
:
6613
25
.
30
Beilharz
EJ
,
Russo
VC
,
Butler
G
, et al
. 
Co-ordinated and cellular specific induction of the components of the IGF/IGFBP axis in the rat brain following hypoxic-ischemic injury
.
Brain Res Mol Brain Res
1998
;
59
:
119
34
.
31
Firth
SM
,
Baxter
RC
. 
Cellular actions of the insulin-like growth factor binding proteins
.
Endocr Rev
2002
;
23
:
824
54
.
32
Baxter
RC
. 
Signalling pathways involved in antiproliferative effects of IGFBP-3: a review
.
Mol Pathol
2001
;
54
:
145
8
.
33
Ruano
Y
,
Mollejo
M
,
Camacho
FI
, et al
. 
Identification of survival-related genes of the phosphatidylinositol 30-kinase signaling pathway in GBM multiforme
.
Cancer
2008
;
112
:
1575
84
.
34
Vestey
SB
,
Perks
CM
,
Sen
C
,
Calder
CJ
,
Holly
JM
,
Winters
ZE
. 
Immunohistochemical expression of insulin-like growth factor binding protein-3 in invasive breast cancers and ductal carcinoma in situ: implications for clinicopathology and patient outcome
.
Breast Cancer Res
2005
;
7
:
R119
29
.
35
Jiang
R
,
Mircean
C
,
Shmulevich
I
, et al
. 
Pathway alterations during glioma progression revealed by reverse phase protein lysate arrays
.
Proteomics
2006
;
6
:
2964
71
.
36
Renehan
AG
,
Zwahlen
M
,
Minder
C
,
O'Dwyer
ST
,
Shalet
SM
,
Egger
M
. 
Insulin like growth factor (IGF)-I, IGF binding protein-3, and cancer risk: systematic review and meta-regression analysis
.
Lancet
2004
;
363
:
1346
53
.
37
Martin
JL
,
Weenink
SM
,
Baxter
RC
. 
Insulin-like growth factor-binding protein-3 potentiates epidermal growth factor action in MCF-10A mammary epithelial cells. Involvement of p44/42 and p38 mitogen-activated protein kinases
.
J Biol Chem
2003
;
278
:
2969
76
.
38
Akkiprik
M
,
Hu
L
,
Sahin
A
,
Hao
X
,
Zhang
W
. 
The subcellular localization of IGFBP5 affects its cell growth and migration functions in breast cancer
.
BMC Cancer
2009
;
9
:
103
.
39
Wang
H
,
Arun
BK
,
Wang
H
, et al
. 
IGFBP2 and IGFBP5 over expression correlates with the lymph node metastasis in T1 breast carcinomas
.
Breast J
2008
;
14
:
261
7
.