• Open Access

Glucose-regulated protein 58 modulates cell invasiveness and serves as a prognostic marker for cervical cancer

Authors


To whom correspondence should be addressed. E-mail: khlin@mail.cgu.edu.tw

Abstract

Human papilloma virus infection is critical but not sufficient to cause cervical cancer. Molecular markers of cervical carcinogenesis are essential. The aim of this study was to identify aberrantly expressed proteins in cervical cancer and determine their clinical significance. A two-dimensional polyacrylamide gel electrophoresis (2-DE) proteomic strategy was used for screening candidate proteins. Immunoblotting and immunohistochemical (IHC) analyses were performed to confirm the results of 2-DE, and the clinical significance was estimated. Glucose-regulated protein 58 (Grp58) was overexpressed in 73% of cancers. The IHC staining showed that the Grp58 histoscore was significantly higher in patients with adenocarcinoma (AD) compared with squamous cell carcinoma (P < 0.05). Grp58 staining was intense in AD with a penetration depth greater than half of the cervical stroma (P = 0.033). High Grp58 expression was associated with low overall survival and recurrence-free survival (RFS) rates (P = 0.007 and P = 0.013, respectively). In multivariate analysis, high Grp58 expression (P = 0.042) and lymph node metastasis (P = 0.026) were determined as independent prognostic factors for RFS. Patients exhibiting both high Grp58 expression and lymph node metastasis displayed poorer outcomes than the other patient groups. In functional studies, knockdown of Grp58 in HeLa cells led to decreased cell invasiveness and inhibition of lung metastasis in a xenograft mouse model. In conclusion, Grp58 serves as a potent prognostic factor of cervical AD. Estimation of the Grp58 index in conjunction with the lymph node metastasis status might aid in predicting the prognosis of cervical AD. (Cancer Sci 2011; 102: 2255–2263)

Cervical cancer is one of the most common cancers among women worldwide. Approximately 500 000 diagnoses of cervical cancer are made each year, leading to 274 000 deaths, although the incidence and death rate have declined markedly in the past three decades because of effective screening.(1) Human papillomaviruses (HPV) are the causative agents of over 99% of cervical cancers containing viral sequences. Infection by high-risk HPV types is necessary but not sufficient for progression to cancer. Most HPV infections are transient without clinical manifestation and are cleared naturally; <1% of women become HPV carriers and remain at high risk of developing cervical cancer.(2) The mechanisms underlying the progression and regression of lesions caused by HPV infection are unknown, and further investigation of the molecular alterations in cervical cancer is needed.

Glucose-regulated protein 58 (Grp58) is a glycoprotein-specific thiol-oxidoreductase belonging to the disulfide isomerase family of proteins. Grp58 is a multifunctional protein.(3) Recent studies suggest that Grp58 plays a role in cancer, although related information is limited. Hirano et al.(4) demonstrated that enhanced expression of Grp58 was associated with oncogenic transformation in normal rat kidney cells. Grp58 overexpression modulates STAT3 signaling in cancer.(5,6) Conversely, loss of Grp58 expression correlates with more aggressive gastric cancer and could provide prognostic information for patients.(7) The roles of Grp58 in different tumors are still controversial, and how Grp58 is involved in cervical tumorigenesis remains to be investigated. Thus, we aimed to explore the clinicopathological significance of Grp58 in cervical carcinomas and validate its molecular function in both in vitro and in vivo studies.

Materials and Methods

Patients.  A total of 109 patients with cervical carcinoma who underwent primary definitive surgery between 2000 and 2008 at Chang Gung Memorial Hospital, Taoyuan, Taiwan, was retrieved from the hospital database and enrolled for prognosis analysis under the Institutional Review Board-approved protocol (IRB: 95-1241B) with informed consent. The patients, comprising stage I to IIB disease, were clinically staged according to the International Federation of Gynecology and Obstetrics (FIGO) staging system. Fresh clinical specimens and paraffin-embedded tissues of the 109 patients were obtained from the hospital archive. Patient and tumor characteristics are listed in Table 1. The paired adjacent normal and tumor tissues from four of the 109 patients, comprising two squamous cell carcinomas (SCC) in stage IB1 and IIA, one adenocarcinoma (AD) in stage IB2, and one adenosquamous cell carcinoma (ADSCC) in stage IB1 were used in the proteomic analysis. Fresh tumor and paired adjacent non-cancerous specimens from 37 of the 109 patients were available and used in the western blot analysis. The histological types of the 37 patients were SCC in 17 patients, AD in 13 patients and ADSCC in seven patients. The 106 available paraffin-embedded tissues from 109 cervical cancer patients were sectioned and used in the immunohistochemical (IHC) and statistical analyses (three of 109 patients who were lost to follow up were excluded). The median age of the 106 patients was 52 years (range, 33–78). The histological types were SCC in 52 patients, AD in 34 patients and ADSCC in 20 patients. The median follow-up time of the survival patients among 106 cases was 49 months (range, 4–114 months). There were 29 patients with relapse and 13 patients died during the follow-up period. Lymph node metastases were histologically confirmed in 26 (24.5%) of 106 patients.

Table 1.   Patient and tumor characteristics (n = 109)
Parametern (%)
Total no. of patients109 (100.0%)
Median age (range) (years)52.3 (32.9–78.4)
Histological type
 Squamous cell carcinoma53 (48.6%)
 Adenocarcinoma34 (31.2%)
 Adenosquamous cell carcinoma22 (20.2%)
Stage
 IA1 (0.9%)
 IB170 (64.2%)
 IB220 (18.3%)
 IIA11 (10.1%)
 IIB7 (6.4%)
Histological grade
 Well differentiated23 (21.1%)
 Moderately differentiated39 (35.8%)
 Poorly differentiated40 (36.7%)
Tumor size
 ≤4 cm51 (46.8%)
 >4 cm25 (22.9%)
Depth of penetration
 ≤50%37 (33.9%)
 >50%52 (47.7%)
Lymph node metastasis
 No82 (75.2%)
 Yes26 (23.9%)
Recurrence
 No77 (70.6%)
 Yes30 (27.5%)

Two-dimensional polyacrylamide gel electrophoresis (2-DE) and matrix-associated laser desorption/ionization time-of-flight mass spectrometric (MALDI-TOF MS) analysis.  Tumor and paired adjacent normal tissue protein samples (150 μg each) were separated by 2-DE. Details of the 2-DE and mass spectrometric analysis have been described previously.(8)

Western blot analysis.  Western blot analysis was performed as described previously.(9) Grp58-specific first antibody (1:10 000 dilution, Atlas; Sigma-Aldrich, St Louis, MO, USA) and horseradish peroxidase-conjugated, affinity-purified second antibody to rabbit (Santa Cruz Biotechnology, Inc., Santa Cruz, CA, USA) were used. The immunocomplexes were visualized by chemiluminescence with an ECL detection kit (Amersham, Piscataway, NJ, USA). The intensities of the immunoreactive bands were quantified using ImageGauge software version 3.46 (Fuji Photo Film, Tokyo, Japan).

Immunohistochemistry (IHC).  Formalin-fixed and paraffin-embedded tissues were examined by IHC according to previously described procedures.(10) The Grp58 antibody (Atlas) was used at 1:2000 dilution. The second antibody was horseradish peroxidase-conjugated anti-rabbit antibody (Santa Cruz Biotechnology, Inc.). The immunocomplexes were visualized using Envision-kit (DAKO, Carpinteria, CA, USA). Brown-colored cytoplasmic patches were considered Grp58 positive. Slides were scored separately by two independent pathologists (Y. L. and S. M. J.) blinded to all clinical data. The staining intensity was graded as absent (0), weak (1+), medium (2+) or strong (3+). The histoscore (Q) was calculated by multiplying the percentage (P) of positive cells by the intensity (I), according to the formula: Q = P × I. The mean Q of each type of cervical cancer was chosen as the cut-off value to divide the high/low expression groups as previously described.(11,12)

Cell line and cell culture.  The human cervical cancer cell line HeLa was obtained from American Type Culture Collection (ATCC; Number: CCL-2) within 6 months before the experiment and cultured as recommended. To analyze the serum-induced change in cell morphology, cells were wounded and stimulated (10% FBS) as previously described(13) following 24 h serum deprivation. Images of the cells were acquired at 0 and 90 min after stimulation.

Establishment of HeLa Grp58 knockdown stable cells.  Clones of short hairpin RNA (shRNA) targeting Grp58 were purchased from GenDiscovery (Open Biosystems, Huntsville, AL, USA). Transfection of shRNA to target the endogenous Grp58 gene was performed using electroporation. After 24 h of incubation, cells were selected in medium containing puromycin (125 ng/mL) for 2 weeks. Specific repression of the targeted gene was confirmed using western blot.

Transwell assay.  The migration and invasion activity of cells was examined using the Transwell migration assay (Corning-Costar 3494 Transwell, Lowell, MA, USA) as described previously.(9)

Xenograft mouse model.  HeLa stable cells were harvested and washed twice with PBS. The cells (2 × 106/150 μL per mouse) were injected intravenously into the tail vein of 4-week-old male severe combined immunodeficiency (SCID, C.B17/Icr-scid) mice. Six weeks later, the mice were killed and the lungs were removed, fixed with 3.7% formaldehyde and sectioned. The number of mice bearing lung foci was determined.

Real-time quantitative RT-PCR (qRT-PCR).  Total RNA was extracted from cells using Trizol. The first cDNA strand was then synthesized using the superscript III kit for RT-PCR (Life Technologies, Carlsbad, CA, USA). qRT-PCR was performed using SYBR green as described.(14) The sequences of primers used are listed in the Data S1.

Statistical analysis.  Groups were compared using the Mann–Whitney U test and Kruskal–Wallis H test. The cancer-specific overall survival (OS) and recurrence-free survival (RFS) with death or relapse as events were analyzed using the log-rank test. The OS and RFS were determined from the date of surgery to the date of the last follow up and the date of first detected recurrence respectively. The Kaplan–Meier survival curves and 5-year survival rates were used to analyze the survival outcome. Cox regression analysis (covariates including parameters associated with OS or RFS in univariate analysis) was used to identify the independent predictors of survival. The multivariate-adjusted hazard ratios (HR) and 95% confidence intervals (CI) were used. P < 0.05 was considered significant.

Results

Proteomic findings.  The 2-DE, MALDI-TOF MS analysis of cervical cancer tissues and comparison with adjacent normal cervical tissues led to the identification of an array of proteins. Ten proteins displaying identical trends in at least two patients (>2-fold change) are listed in Table 2. Alpha enolase (ENO1), maspin, Grp58, pyruvate kinase 3 isoform 1 (PKM2), peroxiredoxin 1 (PRDX1), squamous cell carcinoma antigen 1 (SCCA1) and SCCA2 were identified with an elevated level in tumor tissues; conversely, Mn-superoxiddismutase (MnSOD), PRDX6 and tropomyosin 2 (TPM2) were decreased in tumor tissues. Among these, Grp58 was upregulated in all four cancers and therefore selected for further analysis (Table 2, Fig. 1A).

Table 2.   Proteins identified in cervical tissues
CxCa patient#1#2#3#4
FIGO stageIB1IIAIB1IB2
Histological typeSCCSCCADSCCAD
SymbolDescriptionMW (kDa)PITrend†
  1. †Proteins with an intensity change greater than twofold in tumors compared with the adjacent normal tissues are listed. AD, adenocarcinoma; ADSCC, adenosquamous cell carcinoma; CxCa, cervical carcinoma; Down, downregulation; FIGO, International Federation of Gynecology and Obstetrics; Grp58, glucose-regulated protein 58; MW, molecular weight; PI, isoelectric point; Up, upregulation; SCC, squamous cell carcinoma.

ENO1Alpha enolase47.57.0UpUpUp 
Grp58Glucose regulated protein 58 kDa57.26.0UpUpUpUp
MaspinSerine proteinase inhibitor, clade B, member 542.65.7UpUp  
MnSODMn-superoxiddismutase19.27.3DownDown  
PKM2Pyruvate kinase 3, isoform 158.57.6Up  Up
PRDX1Peroxiredoxin 122.38.3Up  Up
PRDX6Peroxiredoxin 625.28.6DownDownDownDown
SCCA1Serine proteinase inhibitor, clade B, member 344.66.4UpUp  
SCCA2Serine proteinase inhibitor, clade B, member 445.05.9UpUp  
TPM2Tropomyosin 233.04.6Down  Down
Figure 1.

 Glucose-regulated protein 58 (Grp58) is overexpressed in cervical cancers. (A) Results of two-dimensional polyacrylamide gel electrophoresis (2-DE). The spots corresponding to Grp58 (left panel) are circled. (B) Grp58 protein levels were verified in 37 paired tumor and adjacent normal cervix tissues. Two paired representative results are shown. The lower panel represents the quantitative results. (C) Grp58 expression patterns were determined using immunohistochemical (IHC) analysis. (a) Paired normal squamous epithelium of (b); (b) squamous cell carcinoma (SCC); (c) paired normal columnar epithelium of (d); (d) adenocarcinoma (AD) (original magnification, ×400). ADSCC, adenosquamous cell carcinoma; CxCa, cervical carcinoma; N, adjacent normal; T, tumor; T/N, tumor-to-normal uptake ratio.

Verification of the abundance of Grp58 protein in cervical cancer.  Expression levels of Grp58 were verified using western blot analysis in 37 paired cervical tissues. The Grp58 protein was upregulated (tumor-to-normal uptake [T/N] ratio > 1.3) in 73.0% (27/37) and downregulated (T/N ratio < 0.8) in 2.7% (1/37) of the cervical cancer tissues (Fig. 1B). The average T/N ratio in 37 patients was 4.0 (median, 1.8; range, 0.3–18.3), suggestive of overexpression of the Grp58 protein in cervical cancer. Subsequently, IHC was performed to establish the expression patterns and location of Grp58. Among the 106 slides analyzed, the adjacent normal cervical epithelium was examined on 77 slides. Grp58 Q values of adjacent normal and tumor regions were scored, respectively. However, the paired normal columnar epithelium was rarely determined on a single slide of AD. The sample size was too small for valid statistical inference. Therefore, the staining of normal squamous epithelium was scored for statistical analyses. Grp58 was detected mainly in the cytoplasm and cell membranes of cancer cells. Negative to weak staining of Grp58 was observed in the normal cervical epithelium, in contrast to moderate to strong staining in cancerous tissues (Fig. 1C). Grp58 was significantly overexpressed in cancerous tissues (mean Q ± SD, 170.7 ± 54.8) relative to the non-cancerous tissues (15.5 ± 34.2; P < 0.001). In addition, the expression of Grp58 in pre-invasive lesions was determined (Table S1). We noticed that Grp58 was upregulated in cervical intraepithelial neoplasia 2-3 (CIN2-3) and AD in situ (AIS) by comparing with their adjacent normal epithelium (P = 0.037 and P < 0.001, respectively). The Grp58 Q of SCC was significantly lower than that in CIN3 (P = 0.017). No significant differences were observed when comparing CIN2 with CIN3 or comparing CIN2 with SCC. Furthermore, Grp58 Q was higher in AD than that in AIS (P < 0.001).

Grp58 expression is correlated with clinicopathological features.  The correlation between the Grp58 Q and clinicopathological parameters of 106 cervical cancerous tissues was examined. As shown in Figure 2(A), Grp58 expression differed significantly between different histological types of cervical cancer (P = 0.018). Grp58 Q of AD (n = 34) was significantly higher than that of SCC (n = 52) (mean Q ± SD, 186.2 ± 60.0 and 156.6 ± 56.6, respectively, P < 0.05). Statistical analyses were performed separately in each histological type of cancer. Grp58 Q was higher in AD patients with depth of tumor penetration greater than half of the cervical stroma (P = 0.033; Table 3). However, Grp58 Q of AD was not correlated with age, FIGO stage, histological subtype and grade, tumor size and lymph node status. In SCC and ADSCC (n = 20), Grp58 Q was not linked with any parameters examined (data not shown). These data disclosed that Grp58 overexpression is correlated with the histological AD type and might be involved in the local invasion of cervical cancer.

Figure 2.

 Survival curves of patients. (A) The scatter plot shows glucose-regulated protein 58 (Grp58) Q scores of each histological type of cancer. (B,C) Overall survival (OS) and recurrence-free survival (RFS) curves of adenocarcinoma (AD) patients. (D,E) OS and RFS curves of AD patients regrouped according to the Grp58 Q and lymph node metastasis status. The 5-year survival rates of each group are shown. ADSCC, adenosquamous cell carcinoma; SCC, squamous cell carcinoma.

Table 3.   Patient distribution according to glucose-regulated protein 58 (Grp58) histoscore in cervical adenocarcinoma
Parametern (%)Grp58 histoscoreP
  1. NS, not significant.

Age (years)
 ≤5020 (58.8)181.25 ± 64.68NS
 >5014 (41.2)193.21 ± 54.27
Stage
 ≤Ib116 (47.1)177.19 ± 68.85NS
 >Ib118 (52.9)194.17 ± 51.69
Histological subtype
 Endocervical23 (67.6)186.09 ± 63.85NS
 Intestinal4 (11.8)205.00 ± 33.17
 Signet-ring cell1 (2.9)250.00
 Minimal deviation1 (2.9)50.00
 Mixed5 (14.7)186.00 ± 21.91
Histological grade
 Well18 (52.9)186.39 ± 68.72NS
 Moderate14 (41.2)183.93 ± 54.21
Tumor size
 ≤4 cm14 (41.2)188.57 ± 72.52NS
 >4 cm13 (38.2)188.85 ± 50.17
Depth of penetration
 ≤50%13 (38.2)160.00 ± 60.100.033
 >50%19 (55.9)206.32 ± 55.60
Lymph node involvement
 No24 (70.6)176.67 ± 50.51NS
 Yes10 (29.4)209.00 ± 76.66

Grp58 expression and patient survival.  To analyze the impact of Grp58 expression on clinical outcome, patients were dichotomized according to mean Grp58 Q (mean Q values of SCC, AD and ADSCC were 156.6, 186.2 and 169.3, respectively). Patients with Grp58 Q scores above the cut-off value were classified as the Grp58 high expression group, and those with Grp58 Q scores equal to or below the cut-off value were classified as the Grp58 low expression group. The 5-year OS and RFS rates were estimated. No significant differences in the OS and RFS rates were observed between patients with SCC or ADSCC in the high and low Grp58 expression groups (data not shown). However, the Grp58 expression status was significantly associated with the outcomes of AD patients. The 5-year OS rates of patients with high and low Grp58 expression were 31.5% and 100%, respectively (P = 0.007; Table 4). The 5-year RFS rates in the high and low expression groups with AD were 45.3% and 72.9%, respectively (P = 0.013). Therefore, we focused on AD patients to determine the clinical parameters associated with survival. Lymph node metastasis was another factor associated with poor clinical outcomes (Table 4). However, the other parameters examined were not associated with patient’s prognosis after surgical treatment (Table 4). The OS and RFS curves of the high and low Grp58 expression groups are presented in Figure 2(B,C). Multivariate analysis revealed that high Grp58 expression (HR, 4.90; P = 0.042) as well as lymph node metastasis (HR, 3.76; P = 0.026) are independent factors for poor RFS (Table 4). However, both of these parameters were not independent factors for OS of AD patients (data not shown). We further hypothesized that patients exhibiting both high Grp58 expression and lymph node metastasis (Group 3, n = 7) should have worse outcomes than those with none (Group 1, low Grp58 expression without lymph node metastasis, n = 12) or one (Group 2, low Grp58 expression and lymph node metastasis or high Grp58 expression without lymph node metastasis, n = 15) of the poor prognostic factors. As shown in Figure 2(D,E), OS and RFS rates were significantly different between the groups (P = 0.002 and P = 0.001, respectively), and patients in Group 3 had the worst OS and RFS.

Table 4.   Analysis of clinicopathological parameters and survival of patients with adenocarcinoma (n = 34)
Parametern (%)Univariate analysisMultivariate analysis (RFS)
5-year OSP5-year RFSPHR†95% CIP
  1. †Multivariate-adjusted hazard ratio (HR). CI, confidence interval; Grp58, glucose-regulated protein 58; NS, not significant; OS, overall survival; RFS, recurrence-free survival.

Grp58 histoscore
 Low expression15 (44.1)100%0.00772.9%0.01311.06–22.660.042
 High expression19 (55.9)31.5% 45.3% 4.90
Age (years)
 ≤5020 (58.8)74.8%NS59.8%NS
 >5014 (41.2)64.1%55.6%
Stage
 ≤Ib116 (47.1)79.6%NS73.1%NS
 >Ib118 (52.9)58.3%50.0%
Histological grade
 Well18 (52.9)65.8%NS61.0%NS
 Moderate14 (41.2)84.6%64.9%
Tumor size
 ≤4 cm14 (41.2)70.3%NS57.1%NS
 >4 cm13 (38.2)51.3%46.2%
Depth of penetration
 ≤50%13 (38.2)100%NS84.6%NS
 >50%19 (55.9)58.9%46.1%
Lymph node involvement
 No24 (70.6)84.8%0.02174.6%0.00711.18–12.050.026
 Yes10 (29.4)43.8% 26.7% 3.76

Decreased invasive ability and metastatic capacity of Grp58 knockdown stable HeLa cells.  To explore the specific role of Grp58 in cervical cancer, HeLa cells expressing high levels of endogeneous Grp58 derived from an AD patient were selected as a model, and Grp58 knockdown stable HeLa cells were generated. The Grp58 levels in control (L#1 and L#2) and knockdown (G#1 and G#2) cells were detected. As expected, Grp58 protein was decreased in the presence of Grp58-specific shRNA, but not control scrambled shRNA (Fig. 3A). The migration and invasion abilities of these stable cells were determined. The folds of traversed cells were 0.06 and 0.05 in G#1 and G#2, respectively, relative to the control cells (L#1) in the migration assay (Fig. 3B, left), and 0.07 in G#1 and G#2, respectively, in the invasion assay (Fig. 3B, right). Reduced migration and invasion were not attributable to different cell proliferation rates (data not shown). Cell morphology was additionally determined after stimulation with several growth factors in serum. Growth factor-stimulated lamelipodia formation was visualized in the control cells, but not the Grp58 knockdown cells (Fig. 3C). In the in vivo metastatic assay, lung metastatic foci were present in three of five experimental mice injected with control cells, but none of the five mice injected with Grp58 knockdown cells (Fig. 3D). Clearly, the migration ability and metastatic capacity were inhibited in the Grp58 knockdown cells.

Figure 3.

 The migration, invasion and metastatic capacity were decreased on glucose-regulated protein 58 (Grp58) knockdown. (A) Expression of Grp58 was determined using western blot in control and Grp58 knockdown cells. (B) Trans-well assay was used to determine cell migration (left panel) and invasion (right panel) ability. **P < 0.01. (C) Arrows indicate the serum-stimulated lamelipodia structures. (D) Results of a xenograft mouse model. (a,c) Lungs removed from mice injected with the indicated cells. (b,d) Sections of lung stained with hematoxylin and eosin staining. T, tumor foci; N, normal lung tissue. Scale bar, 100 μm.

Possible mechanisms of Grp58 involvement in HeLa cell invasiveness.  To determine the genes affected by Grp58 knockdown, we performed an Affymetrix microarray (Data S1). The results are shown in Table S2. The mRNA levels of seven genes showing aberrant expression patterns, identified as tumorigenesis related, were determined using qRT-PCR. As shown in Figure 4, CD24, laminin-β3 (LAMB3), S100 calcium-binding protein A4 (S100A4), S100A10, secretory leukocyte peptidase inhibitor (SLPI) and Tax1 binding protein1 (TAX1BP1) were downregulated, and insulin-like growth factor binding protein 7 (IGFBP7) was upregulated on Grp58 knockdown. These findings coincided with those of the Affymetrix microarray.

Figure 4.

 The mRNA level of genes perturbed on glucose-regulated protein 58 (Grp58) knockdown. The mRNA levels of genes in the control and Grp58 knockdown cells were determined with qRT-PCR and the folds of expression by comparison with L#1 cells are shown.

Discussion

Identifying predictive or prognostic biomarkers for cancers are important in clinical translational research. Based on the etiology of cervical cancer, HPV DNA and viral protein detection are frequently used in the screening process. Other molecular indicators of cervical cancer, such as SCCA, CA125, p16ink4a and mini-chromosome maintenance protein 5, have also been identified. However, most of these markers are appropriate for prediction and screening, but are not suitable prognostic or independent indicators of clinical outcome.(15–17) At present, we demonstrated that Grp58 is overexpressed in cervical cancer and serves as an independent prognostic marker for RFS in patients with AD in particular. Other proteins identified in the current study are reported to play roles in cervical cancer. SCCA1 and SCCA2, known clinical biomarkers of SCC,(18) were observed overexpressed in both cervical SCC specimens. Aberrant expression of ENO1,(19) Maspin,(20) PKM2,(21) MnSOD,(22) PRDX isoforms(23) and TPM2(24) in cervical cancer have been reported. The present study supports the notion that these proteins are involved in cervical cancer and were also observed in our database.

It appears that the biological behavior between cervical AD and SCC might be different. According to epidemiological studies, the incidence of AD is increasing relative to SCC, and patients with AD are generally considered to have a worse prognosis and higher recurrence than patients with SCC. In addition, cervical AD is prone to result in radio-resistance compared with SCC.(25) Therefore, screening method, prognosis and treatment strategies specifically tailored to AD are needed. Results of the present study revealed Grp58 expression in AD was significantly higher than in SCC, and high Grp58 expression was associated with serious stromal invasion and poor outcomes for patients with AD but not SCC. Thus, Grp58 should be considered as a potentially and specifically prognostic and therapeutic target for cervical AD. Moreover, more than one-third of cervical cancer patients develop disease relapse within the first 2 years after successful primary treatment.(26) A significant part of recurrence is asymptomatic for curative treatment. Routine surveillance of women previously treated with either radical hysterectomy or radiotherapy is necessary. Nowadays, Papanicolaou (Pap) smear is still the most common tool for screening and surveillance. Nevertheless, an insufficient detection rate of AD rather than SCC using the Pap smear test remains to be one of the unsolved clinical challenges to date.(27) Therefore, identification of molecular markers to predict prognosis of cervical AD is important. We speculate that application of the Grp58 index of patients with AD might be useful in stratified prognosis, prediction of recurrence and determining follow-up strategies. In addition to the difference of prognosis for SCC and AD, Grp58 might also play different roles during CIN and AIS progression to invasive cancer. It is interesting to know how Grp58 involves cervical lesion initiation and progression and whether Grp58 is regulated by HPV oncoproteins.

Lymph node metastasis is the most important histopathological prognostic indicator of cervical cancer.(28) In the present study, lymph node metastasis is associated with worse outcomes for patients with early stage cervical AD, and acts as an independent predictor of poor RFS. However, other histopathological prognostic factors, including tumor size, depth of invasion and tumor stage, were not associated with the outcomes of patients who were in stage I–II disease underwent definitive surgical intervention. After adjusting the lymph node metastasis status, the Grp58 index still worked as a prognostic predictor of RFS. Furthermore, patients exhibiting both these factors (high Grp58 expression and lymph node metastasis) had the worst outcomes. Therefore, the Grp58 expression index might be useful as an adjunct to hysterectomy findings for assisting in prognostic evaluation. However, reduced Grp58 expression is reported to be a poor prognostic factor for gastric cancer.(7) This discrepancy might be due to tissue-specific effects. Celli and Jaiswal(29) demonstrated Grp58 upregulation in breast, uterus, lung and stomach tumors, and downregulation in colon cancer. Therefore, Grp58 might play different roles in different tumor types.

Perturbation of endoplasmic reticulum (ER) function is associated with several diseases, including cancer.(30) The microenvironment of tumors causes physiological ER stress, and cells trigger the unfolded protein response (UPR) to protect tumors from stress-induced cell death. Many UPR-related proteins, such as Grp78,(31) Grp94(32) and calreticulin,(33) are involved in tumorigenesis and tumor progression. In cervical cancer cell lines, the UPR-related proteins such as Grp78, Grp94 and activating transcription factor 4 (ATF4) are associated with tolerance to ER stress, apoptosis and sensitivities to X-ray and anticancer drugs.(34–37) However, the expression and prognostic value of UPR-related protein in cervical cancer are largely unknown. Grp58 is also a type of ER stress-responsive protein. In our experiments, knockdown of Grp58 expression reduced the migration ability and metastatic capacity of HeLa cells. This result is consistent with an earlier clinical study showing that Grp58 is highly expressed in patients with serious cervical stromal penetration, and further supports the theory that Grp58 is critical for cervical AD progression. In our microarray analysis, some genes were perturbed after knockdown of Grp58 in HeLa cells. Many of these genes are involved in the processes of tumor invasion and progression. CD24,(38) LAMB3,(39) S100A4,(40,41) S100A10(42) and SLPI(9) are overexpressed in various types of tumor to augment cell invasiveness. Conversely, IGFBP7 is downregulated in melanoma, colon cancer and thyroid carcinoma, and restoration of IGFBP7 expression reduces the migration ability of cells.(43) TAXIBP1 functions in cell proliferation, and its expression is increased in HCC(44) and ovarian tumors.(45) Both microarray analysis and qRT-PCR showed downregulation of oncogenic genes (cd24, lamb3, s100a4, s100a10, slpi and tax1bp1) and upregulation of oncosuppressor gene (igfbp7) in Grp58 knockdown cells. We propose that Grp58 might regulate HeLa cell invasiveness and metastasis via modulation of these genes. However, the detailed mechanisms underlying the Grp58-mediated regulation of tumor invasion and metastasis require further elucidation.

In conclusion, Grp58 overexpression is a potent prognostic factor in cervical AD. High Grp58 expression and lymph node metastasis are strong independent predictors of poor RFS. A combination of the Grp58 index with the lymph node metastasis index might increase the specificity of cervical AD prognosis. Our clinical and functional studies have demonstrated that Grp58 augments AD local invasion by promoting the invasion and metastasis processes in cervical AD cancer cells. Thus, the potential use of Grp58 as a prognostic factor of cervical AD might encourage consideration of a human clinical trial for more aggressive adjuvant therapy such as concurrent chemoradiotherapy (CCRT) with combined chemotherapy regimens for those with the presence of both lymph node metastasis and high Grp58 expression, in contrast with single agent, platinum-based CCRT for those with none or one risk factor in surgical specimens. Further investigation for translational research and human trials are warranted.

Acknowledgments

This work was supported by grants from Chang-Gung University, Taoyuan, Taiwan (CMRPD 160373) and the Department of Health (DOH99-TD-C-111-006).

Disclosure Statement

The authors declare that there are no financial disclosures from any one.

Ancillary