HIF‐1α and HSP90: Target molecules selected from a tumorigenic papillary thyroid carcinoma cell line

Abstract

It is important to properly identify aggressive tumors among differentiated thyroid cancers that are most often indolent. By comparison of a tumorigenic clone with an originally less tumorigenic papillary thyroid carcinoma (PTC) cell line, we looked for markers involved in the aggressive biology of thyroid cancer. Human PTC cell lines BHP10‐3 and its tumorigenic subclone BHP10‐3SC_mice were compared using microarray analysis. Upregulated genes in the tumorigenic clone were selected for RT‐PCR, immunoblot analysis and immunohistochemistry in human tissue. Hypoxia‐inducible factor (HIF)‐1α and its chaperone protein heat shock protein (HSP)90 showed significantly increased expression in BHP10‐3SC_mice and human PTC tissue. These two genes, HIF‐1α and HSP90, were further validated using siRNA gene knockdown, pharmacological inhibition using 17‐N‐allylamino‐17‐demethoxygeldanamycin (17‐AAG), an inhibitor of both HSP90 and HIF‐1α and in vivo orthotopic animal model. Invasiveness of BHP10‐3SC_mice was abrogated by blockade of HIF‐1αin vitro by both siRNA and 17‐AAG. The same finding was demonstrated in the orthotopic animal model. These findings support that HIF‐1α is important in tumorigenesis of PTC and that it may serve to be an important target for identification and treatment of aggressive tumors. (Cancer Sci 2012; 103: 464–471)

The incidence of thyroid cancer has increased worldwide during the past decade and it has become the most common endocrine malignancy.^{( 1 )} Previously, increased diagnostic accuracy was regarded as the reason for the increased incidence of thyroid carcinoma, mainly in papillary microcarcinoma.^{( 2 )} Recently, the incidence of thyroid carcinoma seems to have significantly increased for all tumor sizes, which cannot be explained by just more frequent detection of indolent diseases.^{( 3 )} According to the surveillance, epidemiology and end results (SEER) cancer registry, the incidence of thyroid cancer rose from 6.4 per 100 000 in 1988 to 14.9 per 100 000 in 2005. However, the mortality rate remained unchanged during this time period. This raises another issue in the management of thyroid cancer. The prevalence of latent thyroid cancer detected by autopsy is reported from 6.0% to 35.6%.^{( 4} , ^{5 )} A Japanese group observed asymptomatic papillary microcarcinoma patients without surgery and reported that only 22.6% of the microcarcinomas increased in size.^{( 6 )}

Despite excellent prognosis, some patients with papillary thyroid carcinoma (PTC) suffer from locally invasive and metastatic diseases that cannot be controlled by conventional therapy. Some tumors undergo anaplastic change. In a recent paper, PTC was divided into three different types and the authors emphasized the necessity of a different therapeutic strategy for each group.^{( 7 )} Hence, the understanding of tumorigenesis in thyroid cancer is important in predicting the prognosis, which will aid in avoiding overtreatment in most patients and also identifying aggressive tumors so that proper treatment can save their lives.

Studies searching for tumor‐initiating cells are one such way of finding the mechanism of thyroid cancer tumorigenesis. However, a low proliferation rate, inconsistent molecular aberration, and morphological and functional heterogeneity of diverse thyroid cancers are obstacles in tumorigenesis research.^{( 8 )} Another obstacle is that a thyroid cancer animal model is rare because most PTC cell lines cannot make a tumor in vivo. In a previous experiment for the establishment of an orthotopic animal model for PTC in nude mice, a clone (BHP10‐3SC_mice) that has more tumorigenic potential than the original cell line (BHP10‐3) could be established.^{( 9 )} Because the two cell lines share the same genetic background with different tumorigenesis potential, this can be utilized in identifying important genes that have a role in initiation, progression and metastasis of thyroid papillary carcinoma. On these grounds, we aimed to find candidate genes that are important in tumor initiation of PTC and characterize the discovered gene in vitro and in vivo.

Materials and Methods

Cell lines, culture and preparation.  Papillary thyroid carcinoma cell lines BHP10‐3, BHP10‐3SCmice, BCPAP, SNU790 and KTC1 and anaplastic thyroid carcinoma cell line KTC2 were used. The BHP10‐3 cell line was a kind gift from Dr Jerome Hershman (University of California at Los Angeles, Los Angeles, CA, USA). BHP10‐3SC_mice, a tumorigenic clone of the BHP10‐3 cell line, was developed and subcultured by Dr Gary L. Clayman (MD Anderson Cancer Center, Houston, TX, USA).^{( 9 )} SNU790 was purchased from the Korean Cell Line Bank (Seoul, Korea). BCPAP was acquired from the German Collection of Microorganism and Cell Cultures (Braunschweig, Germany). KTC1 and KTC2 were established in the Kawasaki Medical School in Japan and were kindly offered for the experiment.^{( 10 )} The cells were cultured in RPMI 1640 medium (Invitrogen, Carlsbad, CA, USA) supplemented with 10% fetal bovine serum, 2 mM l‐glutamine and antibiotics in a 37°C incubator with 5% CO₂.

Microarray analysis.  Three sets of Whole Human Genome 4 × 44k Oligo Microarray Chip (Agilent Technologies, Santa Clara, CA, USA) were used for microarray analysis, which contains 43 376 transcript sequences of the human genome. In brief, RNA was reverse transcribed and labeled with aminoallyl deoxyuridine triphosphate (dUTP), which was subsequently coupled with Cy3 and Cy5 dyes for BHP10‐3SC_mice and BHP10‐3 cell lines, respectively, to hybridize the “on” arrays, as described previously.^{( 11} , ^{12 )} Genes with at least a twofold increase in expression in BHP10‐3SC_mice cells over BHP10‐3 cells were identified for further analysis.

RT‐PCR.  Total RNA was isolated using a RNeasy Mini kit (Qiagen, Valencia, CA, USA) according to the manufacturer’s instructions. Single‐strand cDNA was synthesized with oligo‐dT primer for 50 min at 50°C using reagents provided in the ThermoScript RT‐PCR System (Invitrogen). RT‐PCR was performed using CYBR Green PCR Master Mix on an ABI PRISM 7700 sequence detection system (Applied Biosystems, Foster City, CA, USA). The primers used are shown in Table 1.

Table 1.

 Primer sequence for PCR

Primer	Sequence
TTC9 forward	GCCATCGAGATCGACTGTTAC
TTC9 reverse	ACTCGTTCATAGTTTACCAGCTCA
Lumican forward	TGGAGGTCAATCAACTTGAGAA
Lumican reverse	CAAACGCAAATGCTTGATCTT
TPR forward	AGTTGGGACCACCAGTTCAG
TPR reverse	GTCAACTGAAGGCCACGTC
Humanin1 forward	CGATTAAAGTCCTACGTGATCTGA
Humanin1 reverse	AGGGAGGAATTTGAAGTAGATAGAAA
Integrin β1 forward	CGATGCCATCATGCAAGT
Integrin β1 reverse	ACACCAGCAGCCGTGTAAC
HSP105 forward	CCACCAGAAAACCCAGACAC
HSP105 reverse	GGTTTGTTGAGCATCAGTTTGT
HIF‐1α forward	GGTTCACTTTTTCAAGCAGTAGG
HIF‐1α reverse	TGGTAATCCACTTTCATCCATTG
HSP90 forward	GTCCTGTGCGGTCACTTAGC
HSP90 reverse	AAAGGCGAACGTCTCAACC

HIF, hypoxia‐inducible factor; HSP, heat shock protein.

Immunoblot analysis.  The following antibodies were used: anti‐hypoxia‐inducible factor (HIF)‐1α, anti‐heat shock protein (HSP)105, anti‐HSP90, anti‐TPR, anti‐Integrin β1 (Cell Signaling Technology, Danvers, MA, USA), anti‐RACK1 (BD Biosciences, San Jose, CA, USA) and anti‐GAPDH (Sigma, St Louis, MO, USA).

Immunohistochemistry.  Nineteen patients undergoing total thyroidectomy for PTC were selected for immunohistochemistry (IHC). Normal thyroid tissue with the furthest distance from the tumor in the same thyroid specimen was used for comparison. The IHC staining was reviewed under high‐power field (×400) and graded from 0 to 4+ according to the strength of staining. The IHC staining was performed with the following antibodies: anti‐HIF‐1α, anti‐HSP105, anti‐HSP90, anti‐TPR and anti‐Integrin β1 (Cell Signaling Technology). The protocol was approved by the institutional review board (IRB B‐1012‐117‐302).

Gene knockdown using siRNA.  HIF‐1α siRNA (siRNA no. 1068432), HSP90A siRNA (siRNA nos. 1071294, 1071295, 1071296) and control siRNA were purchased from Bioneer (Daejeon, Korea). siRNA mediated gene knockdown was performed using DharmaFECT 4 according to the manufacturer’s instructions (Dharmacon, Chicago, IL, USA). The cells were harvested for experiment after 24 h of incubation with siRNA.

Cell proliferation assay.  Growth analysis for the cells was performed using a water soluble tetrazolium (WST) assay as follows. Cells were plated in triplicate at concentrations of 2 × 10³ cells per well into 96‐well plates. The IC₅₀ value, at which 50% of the cell growth is inhibited compared with the control, was calculated by nonlinear regression analysis using GraphPad Prism 4 software (GraphPad Software, San Diego, CA, USA).

Cell invasion assay.  The ability of invasion was evaluated in Boyden chambers fitted with 8‐μm pore membranes with Matrigel (BD Biocoat Matrigel Invasion Chambers, San Jose, CA, USA). Briefly, 1 × 10⁵ cells were loaded in serum‐free medium in the top wells of each Boyden chamber and normal medium containing fetal bovine serum was placed into the bottom well as a chemoattractant. After incubation for 22 h, any cells that had invaded the bottom side of the filter were fixed and stained (Harleco Hemacolor staining kit; EMD Chemicals, Gibbstown, NJ, USA). Three different fields on each fixed and stained filter were examined under a microscope at ×200 magnification and the cells in each field were counted. All assays were performed in duplicate and the results of at least three independent experiments were averaged. The results were compared using unpaired t‐tests.

Orthotopic thyroid cancer mouse model.  All animal studies involving mice were performed in accordance with the protocol approved by the Institutional Animal Care and Use Committee. The orthotopic thyroid cancer model was developed in 8‐ to 12‐week‐old female nude mice, as described previously.^{( 9 )} In brief, after a 1‐cm‐long midline incision on the anterior neck, cultured cells (1 × 10⁶) in PBS were injected through the exposed strap muscles into the thyroid gland.

Mice were randomly assigned to treatment and control groups. In the treatment group, mice were injected with 50 mg/kg of 17‐N‐allylamino‐17‐demethoxygeldanamycin (17‐AAG) intraperitoneally on days 7, 9, 12, 14 and 16 following tumor injection. Mice in the control group were injected with the same volume of control vehicle. On day 28, mice were killed and the thyroid gland was inspected grossly. After extraction of the gland, tumor dimensions were measured using calipers. The tumor was then dissected, measured and then placed in 10% formalin solution for fixation.

Statistical analysis.  Data are expressed as mean ± standard deviation. The Student’s t‐test was used for continuous variables and the Mann–Whitney U‐test was used for categorical data. The level of statistical significance was set at P < 0.05.

Results

Identification and selection of candidate gene.  Microarray analysis was used to determine differences between the more tumorigenic clone (BHP10‐3SC_mice) and the original BHP10‐3 cell line. Triplicates of determinations of this analysis revealed a significant upregulation of 513 genes and downregulation of 175 genes (at least a twofold difference) in the BHP10‐3SC_mice clone compared with the BHP10‐3 cell line (Table 2). Among the upregulated genes, the cell cycle and translation‐related genes were expressed highly in the BHP10‐3SC_mice clone. Several promising genes that might play important roles in tumorigenesis were selected for further verification. TTC9, lumican, TPR, humanin1, integrin‐β1, HSP105, HIF‐1α and HSP90 were selected for validation of upregulation using RT‐PCR in triplicates (Fig. 1A). RT‐PCR showed high expression of lumican, TPR, integrin‐β1, HSP105, HIF‐1α and HSP90 in the BHP10‐3SC_mice clone.

Table 2.

 Genes differentially expressed in the BHP10‐3SC_mice cell line compared with the original BHP10‐3 cells

Gene	GenBank identification	Function	Fold change	P‐value
Upregulated
TTC9	D 86980	Protein binding	6.359	0.0011
LUM	NM 002345	Collagen fibril organization	5.167	0.0449
TPR	NM 003292	Protein import to nucleus	4.874	0.0025
HN1	AY 029066		4.773	0.0016
ATP13A3	AJ 306929	Cation transport	4.513	0.0387
ITGB1	NM 002211	Cell adhesion, migration	4.438	0.0496
EIF3S10	NM 003750	Regulation of translation	4.300	0.0207
HSPH1	NM 006644	Response to unfolded protein	4.236	0.0158
C6orf111	NM 032870		4.131	0.0462
LOC441666	AK 131313		4.111	0.0214
MRNA	AK 026668		4.044	0.0352
ZBTB38	BC 072415	Regulation of transcription	4.003	0.0415
HIF1A	NM 181054	Regulation of transcription	3.684	0.0225
PRAMEF8	NM 001012276		3.617	0.0232
HSP90AA1	NM 005348	Mitochondrial transport	3.579	0.0159
Downregulated
C1QTNF4	NM 031909		0.130	0.0047
TBX1	NM 080647	Embryonic morphogenesis	0.271	0.0036
ODF3L1	NM 175881		0.294	0.0070
XPO7	NM 015024	Intracellular protein transport	0.302	0.0010
SNX8	NM 013321	Cell communication	0.307	0.0014
C11orf35	NM 173573		0.313	0.0020
HOXA6	NM 024014	Multicellular organism development	0.320	0.0025
RHPN2	NM 033103	Signal transduction	0.324	0.0034
SLFNL1	NM 144990		0.326	0.0047
ONECUT2	NM 004852	Chromatin remodeling	0.331	0.0023

Figure 1.

 Differential expression of candidate molecules in BHP10‐3SC_mice and BHP10‐3. (A) RT‐PCR using total RNA. (B) Immunoblot to compare the expression between nuclear and cytoplasmic protein and also among different passages. Cells in the 3rd, 29th and 60th passages cultured in normal culture medium were used. HIF, hypoxia‐inducible factor; HSP, heat shock protein.

Immunoblotting was performed for the five candidate genes for evaluation at the protein level (Fig. 1B). HSP105 showed no difference between the two clones. Expression of integrin β1 was slightly increased in the nucleus but not in the cytoplasm. Expression of HIF‐1α, HSP90 and TPR was increased both in the nucleus and the cytoplasm.

As the BHP10‐3SC_mice clone has a tendency to decrease in tumorigenic potential with increasing passage number (data not shown), we also compared the early and late passages. The cytoplasmic HIF‐1α showed a tendency to decrease in expression with increasing passage number and the cytoplasmic HSP90 looked like it increased with passages. Other molecules did not show any difference with passages.

Validation of candidate gene expression in human PTC tissue.  There was a significant difference in the expression of HIF‐1α and HSP90 between PTC and normal thyroid tissue (Fig. 2, Table 3). Integrin β1 was not detected in any of the specimens. HSP105 and TPR were strongly expressed in both PTC and normal thyroid tissue.

Figure 2.

 Immunohistochemical staining of HIF‐1α, HSP90, HSP105, TPR and integrin 1β in human papillary thyroid carcinoma (PTC) and normal thyroid tissue. Hypoxia‐inducible factor (HIF)‐1α and heat shock protein (HSP)90 was highly expressed in PTC tissue but not in normal thyroid tissue (×400 magnification).

Table 3.

 Comparison of immunohistochemical expression of candidate genes in PTC (n = 19) and NTT (n = 18); highly expressed specimens (staining score 3+ or higher) were compared between the two groups

	PTC (n = 19) (%)	NTT (n = 18) (%)	P‐value
TPR	19 (100	16 (88.9)	0.230
Integrin β1	0 (0)	0 (0)	–
HSP105	19 (100)	18 (100)	–
HIF‐1α	13 (68.4)	0 (0)	0.000
HSP90	11 (57.9)	0 (0)	0.000

HIF, hypoxia‐inducible factor; HSP, heat shock protein; NTT, normal thyroid tissue; PTC, papillary thyroid carcinoma.

SiRNA knockdown of HIF‐1α and HSP90.  With the results obtained from immunoblotting and immunohistochemistry, HIF‐1α and HSP90 were selected for further validation. The protein expression of HIF‐1α was blocked by both HIF‐1α siRNA and HSP90AA1 siRNA (Fig. 3A). The efficacy of HSP90 siRNA was different between siRNA and no. 1071294, and 1071296 inhibited the expression of HIF‐1α effectively. A combination of HIF‐1α and HSP90AA1 siRNA did not show a synergistic effect in suppressing HIF‐1α. In contrast, HSP90 expression was not affected by HSP90AA1 siRNA. Although the expression of HIF‐1α in cytoplasm was faint compared with the nucleus, suppression of HIF‐1α expression by siRNA could also be observed, consistent with the finding in the nucleus.

Figure 3.

 siRNA knockdown of hypoxia‐inducible factor (HIF)‐1α and heat shock protein (HSP)90. (A) Immunoblotting of HIF‐1α and HSP90 with HIF‐1α and three kinds of HSP90 siRNA (294, 295 and 296). (B) Water soluble tetrazolium (WST) assay for cell proliferation. (C) Comparison of invaded cell number with siRNA treatment (average cell numbers of three different fields under ×200 magnification). (D) Invasion assay results with siRNA (×200 magnification).

For the proliferation and invasion assays, HSP90AA1 siRNA no. 1071296 was used. siRNA inhibition of HIF‐1α and HSP90AA1 did not effect proliferation (Fig. 3B). In contrast, the invasion assay showed that HIF‐1α and HSP90AA1 siRNA inhibition decreased the number of invaded cells through the Matrigel significantly (Fig. 3C,D). The combined use of HIF‐1α and HSP90AA1 siRNA failed to show an additive effect in blocking the invasion of cells, also consistent with the finding of immunoblotting.

17‐AAG inhibition of HSP90 and HIF‐1α.  Treatment of 17‐AAG to the BHP10SC_mice cell line blocked the expression of HIF‐1α significantly in a dose‐dependent fashion (Fig. 4A). The inhibition lasted up to 24 h after the treatment. But the expression of HSP90 did not change with the 17‐AAG treatment. In contrast to siRNA knockdown, 17‐AAG effectively inhibited proliferation of the BHP10SC_mice cell line in vitro (Fig. 4B). The concentration needed to inhibit 50% of the proliferation (IC₅₀) was 0.023 μM. Treatment with 17‐AAG also blocked cell invasion in a dose‐dependent fashion (Fig. 4C,D). The 17‐AAG treatment in an orthotopic mouse thyroid cancer model after injection of BHP10‐3SC_mice cells effectively suppressed growth of the tumor (Fig. 4E,F).

Figure 4.

 17‐N‐allylamino‐17‐demethoxygeldanamycin (17‐AAG) blockade of hypoxia‐inducible factor (HIF)‐1α and heat shock protein (HSP)90. (A) Immunoblot shows inhibition of HIF‐1α protein expression by 17‐AAG in a dose‐dependent manner, but not for HSP90. (B) 17‐AAG effectively inhibited proliferation of the BHP10SC_mice cell line (IC₅₀ = 0.023 μM). (C,D) Invasion assay using 17‐AAG. (E,F) Twenty‐eight days after injection of the cell line (yellow arrowhead, trachea; white arrow, papillary thyroid carcinoma developed from injected cell line).

Comparison between PTC cell lines.  Immunoblotting for HIF‐1α, HSP90 and RACK1 were evaluated in SNU790, BCPAP, KTC1 and KTC2 cell lines (Fig. 5A). HSP90 and RACK1 were highly expressed in all cell lines and HIF‐1α was strongly expressed in BHP10‐3SC_mice and KTC2 cell lines and KTC1 and BCPAP showed moderate expression of HIF‐1α. The proliferation assay with treatment of 17‐AAG showed a similar inhibition of cell growth among the cell lines. The IC₅₀ of KTC1, KTC2, SNU790 and BCPAP was 0.504, 0.751, 1.926, and 3.648 μM, respectively (Fig. 5B). An invasion assay performed with 0.1 μM of 17‐AAG completely blocked KTC1 and KTC2, and also effectively blocked BCPAP. The invasion of SNU790 was less affected by the 17‐AAG treatment (Fig. 5C).

Figure 5.

 Role of 17‐N‐allylamino‐17‐demethoxygeldanamycin (17‐AAG). (A) Immunoblotting of total protein. (B) Cell proliferation assay obtained 48 h after treatment with 17‐AAG. (C) Invasion assay (×200 magnification) and (D) comparison of invaded cell number with treatment of 0.1 μM 17‐AAG for 24 h. HIF, hypoxia‐inducible factor; HSP, heat shock protein.

Discussion

In the present study we have demonstrated that HIF‐1α and HSP90 play an important role in the tumorigenesis of PTC. Our results support that HIF‐1α is highly expressed in PTC, associated with invasiveness, and that inhibition of HIF‐1α and also HSP90 abrogates proliferation and invasiveness of the cells, both in vitro by knockdown and by pharmacological inhibition, which was also demonstrated in an orthotopic animal model. It is noteworthy to mention that the selection of the molecules tested in the current study was not from a hypothesis. Rather, it was carefully selected from the microarray of the whole human genome, verified at the RNA and protein levels in both cell lines and human tissue, and validated by inhibition of the specific molecule. We believe that discovery of the target molecule in this fashion renders more compelling evidence than other approaches.

Of note, there was a paper claiming that the BHP10‐3 cell line is identical with TPC1.^{( 13 )} Using short tandem repeat and single nucleotide polymorphism array analysis, they found that BHP2‐7, BHP10‐3, BHP7‐13 and TPC1 cells are genetically identical and now TPC1 is generally used to represent these cell lines. However, the name BHP10‐3 is used in this paper to maintain continuity with previous work.^{( 9 )} BHP10‐3SC_mice is a subclone of BHP10‐3, which was selected using a colony forming assay and through reculture of tumor grown in nude mice from the originally non‐tumorigenic cell line.

TTC9, lumican, TPR, humanin1, integrin‐β1, HSP105, HIF‐1α and HSP90 were selected as the candidate genes from 15 upper ranked genes because the function of these genes is known and antibodies for further study were available. All of these genes showed upregulated expression of RNA in the BHP10‐3SC_mice cell line with RT‐PCR. TPR, integrin‐1β, HSP105, HIF‐1α and HSP90 were selected for protein level expression. These genes have been shown to be associated with cancer progression and a possible application in targeted therapies.^{( 14} , ¹⁵ , ¹⁶ , ¹⁷ , ^{18 )} At the protein level, only HIF‐1α, HSP90 and TPR showed a significant increase in expression. There was no difference in the expression of HSP105 and integrin β1 between BHP10‐3SC_mice and BHP10‐3. Because the BHP10‐3SC_mice cell line showed a tendency of decreased tumorigenic potential with increasing passage numbers (data not shown), three BHP10‐3SC_mice cell lines in different passages were compared with the original BHP10‐3 cells. It is well known that culturing cancer stem cells in serum media induces differentiation of cancer stem cells to less tumorigenic cells with ongoing passages.^{( 19 )} We assumed that a similar phenomenon might be occurring in the BHP10‐3SC_mice cell line, so we tried to compare the early and late passage cells to find the molecule associated with the change of passages. The only molecule that showed some difference with passage was HIF‐1α and HSP90 in cytoplasm. This trend of change in expression suggests that HIF‐1α and HSP90 might be a key molecule that controls the tumorigenesis among the selected candidate genes.

For the next step, an immunohistochemical study in human PTC was performed for comparison of expression of the molecular markers between cancer and normal thyroid follicles. Same as the result from immunoblotting, HIF‐1α and HSP90 showed significantly stronger expression in PTC than in normal tissue. HIF‐1α is assumed to have a role in maintaining the cancer stem cells in glioma and colorectal cancers.^{( 14} , ²⁰ , ^{21 )} HSP90 is also a chaperone protein that plays a major role in regulating HIF‐1α activity.^{( 22 )} For that reason, these two molecules were selected for further evaluation to see the effect of blockade of these molecules.

The immunoblotting after transfection with siRNA showed a significant decrease of HIF‐1 by HIF‐1α siRNA and also by HSP90AA1 siRNA. The expression of HSP90 did not change by HSP90AA1 siRNA transfection. This can be explained by the presence of various functional genes encoding HSP90. The gene that showed a difference in the microarray was HSP90AA1, but there are another four genes, HSP90AA2, HSP90AB1, HSP90B1 and TRAP1, that encode HSP90 isoforms. As HSP90A is also composed of two isoforms, the overall expression of HSP90A did not show any difference by HSP90AA1 siRNA transfection. However, the decrease in expression of HIF‐1α suggests some role of HSP90AA1 in stabilization of HIF‐1α. Proliferation of BHP10‐3SC_mice cells was not influenced by blocking HIF‐1α or HSP90AA1, but showed significant inhibition of the invasiveness of cells. A combination of HIF‐1α and HSP90AA1 siRNA did not show an additive effect, which could also be observed in immunoblotting. The immunoblot results show that nuclear HIF‐1α expression was more strongly inhibited by HSP90AA1 siRNA than the combined transfection with HIF‐1α and HSP90AA1. As the HSP90AA1 blockage did not alter the expression of HSP90 protein in cells, the effect of HSP90AA1 seems to exert its effect by regulating the expression of HIF‐1α. Blocking of HIF‐1α changes the invasiveness of tumor cells without inhibiting the proliferation. Because invasiveness is an important behavior of tumor‐initiating cells, the result also supports the key role of HIF‐1α in tumorigenesis.

17‐AAG (geldanamycin) is a specific inhibitor of HSP90 and also of HIF‐1α via blocking the chaperone protein HSP90. Blocking both molecules at the same time, which has been shown to be important in the tumorigenesis of PTC in the present study, seems to be a promising target for molecular therapy. Previously, there has been one investigation of geldanamycin in thyroid carcinoma cell lines.^{( 23 )} Six cell lines were tested in the study and one of which was TPC‐1, which is identical to the original BHP10‐3 cell line that we have used. In that study, the effect of geldanamycin on proliferation was different among the cell lines. Six days of treatment with 50 nM geldanamycin inhibited 29.4% of TPC‐1 cells. In the present study, BHP10‐3SC_mice, which is derived from the BHP10‐3 cell line, showed an IC₅₀ of 23 nM, which might perhaps be due to the upregulated expression of HIF‐1α and HSP90. In the immunoblot after treatment of 17‐AAG, HIF‐1α was decreased dose dependently and the effect remained up to 24 h after treatment. However, the protein amount of HSP90 did not change. These results reconfirm the effect of HSP90 blocking on HIF‐1α. An invasion assay using a Matrigel chamber showed significant inhibition of invasion by 17‐AAG. An in vivo experiment in the orthotopic animal model also supports the possible therapeutic effect of 17‐AAG in PTC.

To investigate the effect of 17‐AAG in other PTC cell lines, we tested SNU790, BCPAP, KTC1 and KTC2 cell lines. The expression of HIF‐1α was most prominent in the BHP10SC_mice and KTC2 cell lines. The KTC2 cell line is established from anaplastic carcinoma and is assumed to have a more aggressive phenotype. Therefore, this result also supports the role of HIF‐1α in tumorigenesis. There was a report that showed that the RACK1 plays an important role in regulating HIF‐1α in a competitive manner with HSP90.^{( 24 )} However, the expression of RACK1 was not different between cell lines. The WST assay showed inhibition of cell proliferation with a range of IC₅₀. As expected, the KTC1 cell line, which showed strong expression of HIF‐1α, had the lowest IC₅₀ level among the cell lines. The inhibition of invasion was also observed in all cell lines by treating 17‐AAG. However, the effect was stronger in cell lines with HIF‐1α expression.

In a recent paper, HIF‐1α shRNA and various HIF inhibitors abrogated the colony‐forming activity of hematological malignant cell lines.^{( 25 )} In this study, Wang et al., suggested the important role of HIF‐1α in maintaining cancer stem cells. Hypoxia and its role in cancer stem cells has also been reported by other researchers.^{( 26} , ^{27 )} Our data also demonstrates that HIF‐1α is a promising molecule among the selected genes found upregulated, which has the potential to influence tumorigenesis of PTC. HIF‐1α expression in thyroid carcinoma was previously investigated by another group.^{( 28 )} They also reported that HIF‐1α was expressed in thyroid carcinoma and not in normal tissue, and that HIF‐1α exhibited high levels in the nucleus of anaplastic carcinoma.^{( 28 )} They showed that some thyroid carcinoma cell lines express HIF‐1α and the expression can be induced by hypoxia. They concluded that HIF‐1α has a strong association with aggressive disease and therapeutic resistance, which is concordant with our results.

In conclusion, microarray analysis of two cell lines with different potential of tumorigenesis revealed HIF‐1α and its chaperone protein HSP90 to be associated with invasiveness of the tumorigenic cell line. By blocking HIF‐1α through inhibiting both HIF‐1α and HSP90 in vitro via siRNA and 17‐AAG, the invasiveness of BHP10‐3SC_mice was abrogated, which was also demonstrated in vivo with an orthotopic animal model. These findings suggest that HIF‐1α might be an important target for identification and treatment of aggressive tumors.

Disclosure Statement

The authors have no conflict of interest.