SEARCH

SEARCH BY CITATION

Keywords:

  • hedgehog pathway;
  • cervical cancer;
  • chemoradiation;
  • gene expression;
  • heterogeneous

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SOURCES
  8. REFERENCES

BACKGROUND:

Hedgehog (Hh) signaling was assessed in patients with primary cervical carcinoma who were receiving chemoradiation. Because the up-regulation of Hh has been reported in response to hypoxia, the authors examined associations between Hh gene expression and measurements of HP5 (the percentage of oxygen pressure readings in each tumor <5 mm Hg) and interstitial fluid pressure (IFP).

METHODS:

Sonic hedgehog (SHH), Indian hedgehog (IHH), patched 1 and 2 (PTCH1 and PTCH2), smoothened (SMO), and glioma-associated oncogene family zinc finger 1 (Gli1) expression levels were determined using quantitative reverse transcriptase-polymerase chain reaction analysis on 85 frozen samples of primary cervical carcinoma and on 16 normal cervical samples. Clinicopathologic data were collected prospectively. Possible correlations between Hh expression and tumor hypoxia (HP5 and IFP) measured at the time of biopsy, the time to local recurrence, and disease-free survival (DFS) were examined.

RESULTS:

At least 1 member of the Hh pathway was elevated in all but 1 tumor compared with normal tissue (P < .0001). Hh gene expression was heterogeneous with SHH, IHH, and GLI exhibiting bimodal distribution. Elevation of SHH expression (P = .04) and low SMO expression (P = .0007) were associated with HP5. The risk of local recurrence was associated with the up-regulation of SMO (hazard ratio [HR], 2.41; 95% confidence interval [CI], 1.00-5.82; P = .044), the up-regulation of >3 Hh genes (HR, 2.56; 95% CI, 1.09-6.00; P = .026), tumor size (HR, 1.41; 95% CI, 1.14-1.74; P = .0015), and lymph node-positive disease (HR, 2.82; 95% CI, 1.16-6.86; P = .022).

CONCLUSIONS:

The proportion of tumors that expressed Hh genes in cervical cancer was very high. The current data support a role for the Hh pathway in repopulation after chemoradiation and suggest that SMO may be a valid therapeutic target. The authors concluded that further investigation into this pathway after radiation and Hh inhibition are warranted. Cancer 2012;118: 3105–15. © 2011 American Cancer Society.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SOURCES
  8. REFERENCES

Worldwide, cervical cancer is the second most common malignancy in women and a major cause of morbidity and mortality.1 Standard treatment for patients who are diagnosed with stage IIB through IVA cervical cancer consists of cisplatin-based chemotherapy and radiation. However, the 5-year survival is only 66%.2-4 Current options for patients with recurrent and metastatic disease are limited, and their median overall survival is <12 months.5-7 To date, biologic therapy has had little impact on survival.8, 9 The identification of potential new targets are urgently required to develop novel therapeutic strategies.

The hedgehog (Hh) signaling pathway controls cell proliferation and differentiation during embryonic development but is largely suppressed in the adult.10 Pathway activation is initiated by binding of 1 of the 3 Hh ligand proteins: sonic hedgehog (SHH); Indian hedgehog (IHH) or desert Hedgehog, and patched (PTCH). In the absence of Hh ligand, PTCH functions as an inhibitor of smoothened (SMO), a transmembrane protein with homology to G-protein coupled receptors. Binding of any of the 3 Hh ligands to PTCH relieves the suppression of SMO, resulting in downstream activation, including the glioma-associated oncogene (GLI) zinc finger family of transcription factors.10, 11 Hh pathway cellular interactions are not fully understood, but aberrant activation of this pathway has been described in a range of human tumors.12-16 Activation may occur as a result of either mutation12 or deregulation within the tumor cell.17, 18 An alternate model proposes a paracrine interaction between tumor cells and their microenvironment.19-21 In this model, Hh-expressing cancer cells of epithelial origin are regarded as refractory to Hh ligand, whereas stromal cells are ligand responsive.19 Thus, cancer cells may usurp the normal functions of Hh during development, mediating epithelial and mesenchymal cell interactions to further tumorigenesis. Recent in vitro studies have suggested a role for the Hh pathway in the response to hypoxia. Hypoxia significantly enhances expression of SHH and hypoxia-inducible factor 1 (HIF-1), facilitating translocation of GLI1 into the nucleus.22, 23 In pancreatic cell lines, hypoxia increases transcription of SMO in a ligand-independent manner, resulting in increased tumor invasiveness.24

Squamous carcinoma of the cervix is a tumor of epithelial origin that has served as a model for studying the effects of tumor hypoxia and cellular microenvironment on tumor growth. Tissue hypoxia is associated with a poor outcome for women with lymph node-negative cervical cancer who are receiving radiation as primary treatment.25-28 Therefore, investigation of the Hh pathway in women who are receiving radiation for cervical cancer is of interest. Furthermore, recent reports suggest that the Hh pathway may play a role in tumor resistance and repopulation in patients who are receiving chemoradiation for squamous cell carcinoma of the esophagus.29, 30 In addition, in a large trial-based series of patients with head and neck squamous cell carcinoma, GLI1 was associated with a significantly worse outcome for patients who received radiation alone.31 Similarities between cervical squamous carcinoma, head and neck squamous cell carcinoma, and esophageal cancer exist in terms of both etiology and approach to treatment. Therefore, investigation of the Hh pathway in women receiving chemoradiation for cervical cancer is warranted.

Fresh-frozen tumor and hypoxia measures were collected from women who were receiving primary chemoradiation for cervical cancer as part of a prospective translational research program evaluating hypoxia and tumor microenvironment.26 The objectives of this study were: 1) to investigate and compare Hh gene expression using quantitative reverse transcriptase-polymerase chain reaction (qRT-PCR) analysis in cervical tumors compared with normal cervical tissue; 2) to explore the relation between the expression of individual Hh genes and clinicopathologic variables, markers of hypoxia (the percentage of oxygen partial pressure [pO2] readings in each tumor <5 mm Hg [HP5] and interstitial fluid pressure [IFP]); and 3) to explore the relation between Hh gene expression and outcome (time to local recurrence and disease-free survival).

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SOURCES
  8. REFERENCES

Patients and Tissue Samples

Biopsy specimens were obtained from sequential patients who had locally advanced cervical carcinoma (International Federation of Gynecologists and Obstetricians [FIGO] stage IB through IVA), were planning to receive primary treatment with chemoradiation, and were enrolled in our translational research program. After written informed consent was obtained according to a protocol approved by the University Health Network/University of Toronto Research Ethics Board, patients had an examination under anesthesia as part of their pretreatment evaluation. Punch biopsies were obtained according to a standardized research protocol that involved rapid freezing in cryovials containing Optimum Cutting Temperature (OCT) medium (Sakura Finetek, Torrance, Calif) followed by storage in liquid nitrogen. Immediately before taking the biopsies, measurements were made of tumor oxygenation using an Eppendorf polarographic oxygen electrode, and interstitial fluid pressure (IFP) was measured using a wick-in-needle technique, as described previously.32, 33 The results from the pO2 measurements were expressed as the percentage of pO2 readings <5 mm Hg (HP5) in each tumor, and IFP was expressed as the mean elevated tissue pressure (in mm Hg). Samples were collected between December 1999 and 2007.

In addition to the examination under anesthesia (to determine stage according to FIGO guidelines), staging investigations included chest x-ray, computed tomography scans of the abdomen and pelvis, and pelvic magnetic resonance imaging (MRI). Pelvic and para-aortic lymph nodes were classified as positive for metastatic disease if the MRI short-axis dimension was >1 cm and equivocal if it was 8 to 10 mm.34 Although the sensitivity of MRI for detecting small-volume lymph node metastases is low, the specificity for lymph nodes ≥10 mm in the short-axis dimension approaches 100%,35 and our previous cervical cancer studies demonstrated that the classification of lymph nodes in this manner is a strong, independent predictor of disease-free survival after treatment with radiotherapy.36 All patients received external-beam radiotherapy to the primary cervical tumor and pelvic lymph nodes regardless of whether or not enlarged pelvic lymph nodes were identified on pretreatment imaging studies. Patients with enlarged pelvic lymph nodes also received treatment of low para-aortic lymph nodes because of the high risk of occult para-aortic lymph node metastases in this setting. A 4-field box technique was used to treat the pelvis, and anterior-posterior parallel-opposed fields were used for the pelvis and para-aortic lymph nodes. An external-beam radiation dose from 45 to 50 grays (Gy) in 1.8-Gy to 2-Gy daily fractions was delivered using 18-megavolt to 25-megavolt photons. During external-beam radiation, patients received weekly doses of cisplatin 40 mg/m2 (total, 5 doses). External radiation was followed by a single intrauterine brachytherapy treatment using low-dose-rate or pulsed-dose-rate equipment. A brachytherapy dose from 35 to 40 Gy was prescribed to a point 2 cm lateral to the applicator at the level of the external cervical os.26, 32 Clinicopathologic and follow-up data were collected prospectively in a predefined manner.

Tissue layers (200 μm thickness) were cut using a cryostat through the frozen biopsy samples. Samples were kept frozen on dry ice after sectioning. Sections 5 μm in thickness were cut from above and below these tissue layers and stained with hematoxylin and eosin. Biopsy specimens that measured >4 mm in greatest dimension and contained >70% tumor were included in this study. This was based on previous work demonstrating that biopsies with these characteristics were necessary to account for intratumoral heterogeneity in biomarker expression.37 Normal cervical tissue samples were obtained from patients who underwent hysterectomy for benign causes. All samples were reviewed by a specialist gynecologic pathologist (B.C.) to ensure that no dysplasia or malignancy was present in normal tissue.

RNA Extraction and Real-Time Quantitative Polymerase Chain Reaction Analysis

Total RNA was extracted from all frozen tissue sections (normal and tumor) using the Qiagen RNeasy Mini Extraction kit (Qiagen, Mississauga, Ontario, Canada) according to the manufacturer's instructions. From 0.5 to 1.0 μg of DNase-treated total RNA, first-strand complementary DNA (cDNA) was reverse-transcribed using OmniScript (Qiagen) according to the manufacturer's instructions. For real-time PCR detection, 10% of the cDNA was mixed with primers (0.3 μm final concentration), double-distilled H2O, and SYBER Green Master Mix (Applied Biosystems, Foster City, Calif) for PCR amplification with a final well volume of 10 to 20 μL. Sequences for primers for the Hh genes were obtained from Robert L. Yauch of Genentech (South San Francisco, Calif) and were synthesized by Invitrogen (Carlsbad, Calf) (Table 1).

Table 1. Primers for Hedgehog Genesa
Gene NameForward (5′) PrimerReverse (3′) Primer
  • Abbreviations: Gli1, GLI family zinc finger 1; IHH, Indian hedgehog; L32, ribosomal protein L32; PTCH1, patched 1; PTCH2, patched 2; SHH, sonic hedgehog; SMO, smoothened, frizzled family receptor.

  • a

    The primers were sequenced by Robert L. Yauch (Genentech, South San Francisco, Calif) and synthesized by Invitrogen (Carlsbad, Calif).

IHH5′-CGGCTTTGACTGGGTGTATT-3′5′-CCGAGTGCTCGGACTTGAC-3′
SHH5′-CGGCTTCGACTGGGTGTACT-3′5′-GCAGCCTCCCGATTTGG-3′
PTCH15′-CCACGACAAAGCCGACTACA-3′5′-GCTCTGCTGCCGGGATT-3′
PTCH25′-CGCCCGCTATCAGTTTGC-3′5′-CACGATGGCCTTAGCATGTG-3′
SMO5′-TGAGTGGGAGCGCAGCTT-3′5′-GCCCGATGGTCACATTGG-3′
Gli15′-CAGGGAGGAAAGCAGACTGACT-3′5′-GGCTTGGCTGTGGCTTCA-3′
L325′-AACCCAGGCATTGACAACA-3′5′-TGTTGGCTCCCATAACCGATGT-3′

The real-time PCR protocol consisted of 40 cycles at 50°C for 2 minutes (pUNG deactivation to prevent carry over contamination), 95°C for 10 minutes (hot start), 95°C for 15 seconds (melting), and 60°C for 1 minute (annealing and extension), as recommended by the manufacturer (Applied Biosystems, Foster City, Calif). The reactions were run and analyzed with an ABI PRISM 7900 Sequence Detector (Applied Biosystems). Human L32, a ribosomal protein, was used as an endogenous control for normalization. Samples were run in triplicate, and relative levels are provided in the figures with values normalized against L32 (bars represent the mean ± standard error; n = 3). All analyses were performed blinded to study endpoints.

Statistical Analyses

Outcome analyses were based on local relapse (LR) and disease-free survival (DFS), defined as the time between diagnosis and the first failure, either relapse or death. There were 31 events, including 28 failures, and only 3 were death without a previous failure. At the time of analysis, the median follow-up as 6 years (range, 0.76-10.6 years). Associations between individual Hh gene expression and survival (LR and DFS) were investigated by applying either a log-rank test when the expression was dichotomized at the median value or the Cox proportional hazards model for continuous values of Hh gene expression. For the continuous values of Hh gene expression, the linear assumption also was inspected visually by using plots of Martingale residuals. To clearly identify possible trends of Hh gene expression in relation to 3 levels of stage (I vs II vs III/IV) or lymph node status (negative vs equivocal vs positive), the Jonkheere-Terpstra test was used.

Because this was an exploratory study and the distribution of the expression of the Hh genes appeared bimodal, we also undertook a novel analysis in which we dichotomized the expression levels using the expectation-maximization (EM) algorithm.38 An attraction of this approach was that the lower peak of the generated bimodal distributions (see Fig. 1B) was at expression levels quite similar to those in the normal cervix, allowing a categorization of (quasi)normal or up-regulated. The expression levels of all 6 Hh genes, dichotomized based on median values or the EM algorithm, were tested individually for their significance in terms of LR and DFS. Because the Hh genes may interact, we also tested the significance of the number of up-regulated genes, comparing samples that had >3 or <3 (median number) of the tested genes up-regulated. Note that the P values need to be interpreted with caution given the number of analyses undertaken.

thumbnail image

Figure 1. (A) Gene expression profiles are illustrated. SHH indicates sonic hedgehog; PTCH1, patched 1; SMO, smoothened, frizzled family receptor; L32, ribosomal protein L32; IHH, Indian hedgehog; PTCH2, patched 2; Gli1, GLI family zinc finger 1. (B) The distribution of hedgehog (Hh) gene expression and the expectation-maximization (EM) algorithm are illustrated. Red lines represent the cutoff between normal and abnormal values.

Download figure to PowerPoint

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SOURCES
  8. REFERENCES

Patient Characteristics

An analysis of Hh gene expression was conducted on samples taken from 85 patients with cervical carcinoma and 16 normal cervical samples. The majority of tumors (87%) were squamous carcinomas, and 8 (9%) women had adenocarcinoma. Seventy-four percent of women completed all 5 doses of cisplatin, and an additional 12% of women received 4 doses of cisplatin (Table 2). Additional patient characteristics, including IFP and HP5, are provided in Table 2.

Table 2. Population Characteristics
CharacteristicNo. of Patients (%)
  1. Abbreviations: FIGO, International Federation of Gynecology and Obstetrics; HP5, the percentage of oxygen partial pressure readings <5 mm Hg; IFP, interstitial fluid pressure.

Median age [range], y47 [26-79]
FIGO stage 
 IB24 (28)
 IIA2 (2)
 IIB39 (46)
 IIIB19 (22)
 IV1 (1)
Lymph node status 
 Negative40 (47)
 Equivocal15 (18)
 Positive30 (35)
Grade 
 Well differentiated2 (2)
 Moderate differentiated51 (64)
 Poorly differentiated27 (34)
Histology 
 Adenocarcinoma8 (9)
 Adenosquamous3 (4)
 Squamous74 (87)
No. of cisplatin doses 
 ≥473 (86)
 310 (12)
 <32 (2)
Median tumor size [range], cm5 [2-9]
IFP: Median [range], mm Hg16.7 [3.0-37.5]
HP5: Median [range], %59.5 [0.0-98.0]

Gene Expression Analysis

The expression of Hh pathway genes in normal cervix compared with cervical carcinoma tissue is illustrated in Figure 1A. Hh gene expression was heterogeneous in the tumor samples and was significantly higher than in normal cervix (P < .0001). The expression profiles for each gene are plotted from the smaller to larger values for both tumor tissues (black) and normal tissues (red) in Figure 1A. The expression profiles from the normal cervical tissue overlap the values in tumor tissue at low levels, suggesting that some tumors exhibit normal levels of these genes. Further analyses of the distributions suggested a largely bimodal relation (Fig. 1B, black line) that was particularly apparent for SHH, IHH, and GLI1. These distributions of the expression levels are consistent with the tumors exhibiting a mixture of normal or elevated levels of the individual genes. By using the EM algorithm, we dichotomized the values for each Hh gene. In Figure 1B, the red, solid vertical line is the cutoff point chosen by the algorithm between normal and elevated levels. The red lines in the figure are the expected distributions identified by the EM algorithm. What is noteworthy about these distributions is that most tumors had normal levels of PTCH1/PTCH2 and SMO, whereas the opposite was true for SHH, IHH, and GLI. Up-regulation was observed in all histologic subtypes. Figure 2 illustrates up-regulation patterns of the genes within the same tumor based on the EM algorithm, and only 1 patient had a tumor in which there was no up-regulation of at least 1 of the Hh genes. This patient remained alive without evidence of disease recurrence 4.4 years after diagnosis.

thumbnail image

Figure 2. Patterns of hedgehog (Hh) gene up-regulation and recurrence are illustrated. SHH indicates sonic hedgehog; IHH, Indian hedgehog; PTCH1, patched 1; PTCH2, patched 2; SMO, smoothened, frizzled family receptor; GLI1, GLI family zinc finger 1; +/−, with or without.

Download figure to PowerPoint

Correlation Between Hedgehog Gene Expression and Known Prognostic Factors

The data were analyzed using both the cutoff point suggested by the EM algorithm and the more conventional median. When the median was used as a cutoff point, decreased SMO expression was associated with advanced stage (P = .0074) and with a trend toward an association with poorly differentiated cancers (P = .06). Higher expression of IHH was associated with larger tumor size (P = .016) and positive pelvic lymph nodes (P = .0083), although it should be noted that the limited sensitivity of MRI for detecting microscopic lymph node involvement in patients with cervical cancer35 precludes a more detailed analysis of the relation between Hh signaling and metastases in this series. Up-regulation of SHH was associated with tumor hypoxia (HP5; P = .04), whereas the reverse was true for SMO (P = .0007) (Table 3). When the data were analyzed using the EM algorithm, the association between SMO and HP5 was maintained (P = .04) together with an association between GLI and early stage disease (P = .023). No relation was observed between the Hh genes and IFP.

Table 3. The Association Between Hedgehog Gene Expression and Hypoxia Oxygen Partial Pressure (pO2) Measurements Expressed as the Percentage of pO2 Readings in Each Tumor <5 mm Hg (HP5)
 HH Value 
VariableLowHighWilcoxon P
  1. Abbreviations: Gli1, GLI family zinc finger 1; HH, hedgehog gene; IHH, Indian hedgehog; L32, ribosomal protein L32; PTCH1, patched 1; PTCH2, patched 2; SHH, sonic hedgehog; SMO, smoothened, frizzled family receptor.

HP5 by SHH50.662.9.0399
HP5 by IHH62.256.55.7131
HP5 by PTCH159.8559.5.5897
HP5 by PTCH257.561.9297
HP5 by SMO64.542.7.0007
HP5 by GLI162.557.25.3313

Local Relapse and Disease-Free Survival

The median follow-up for the patients in this study was 6 years (range, 0.76-10.6 years). Correlations between LR, DFS, and postulated prognostic factors for the group of patients studied are provided in Table 4. Tumor size and the presence of positive lymph nodes were associated with an increased risk of LR. Poorly differentiated tumors, lymph node-positive disease, and tumor stage approached significance for DFS. There was no association between completion of chemotherapy and LR or DFS, although there was a trend toward improved overall survival for those who received >3 doses of cisplatin (hazard ratio [HR], 0.57; P = .09). These figures should be interpreted with caution, because only 12 patients (14%) received <4 doses of cisplatin. Markers of hypoxia (HP5 and IFP) were not associated with outcome in this group of patients. With the individual genes dichotomized based on the median as the cutoff point, we tested for associations with LR (Fig. 3A), and higher levels of SMO were associated with an increased risk of locoregional recurrence (HR, 2.41; 95% confidence interval [CI], 1.00-5.82; P = .044). A similar pattern was observed with the individual Hh genes when levels were dichotomized based on the EM algorithm (Fig. 3B). No significant association using the median (Fig. 3C) or EM cutoff points for individual Hh genes was observed for DFS. However, there was a trend toward worse survival for patients whose tumors had higher levels of SMO (Fig. 3C).

Table 4. Clinicopathologic Factors and Recurrence-Free and Disease-Free Survival
 Recurrence-Free SurvivalDisease-Free Survival
VariableHR (95% CI)PHR (95% CI)P
  1. Abbreviations: CI, confidence interval; HP5, the percentage of oxygen partial pressure readings <5 mm Hg; HR, hazard ratio; IFP, interstitial fluid pressure.

Age1.00 (0.96-1.03).821.01 (0.99-1.04).33
Tumor size1.41 (1.14-1.74).00151.23 (0.97-1.57).09
Lymph node status    
 Equivocal vs negative0.74 (0.16-3.44).700.81 (0.26-2.48).71
 Positive vs negative2.82 (1.16-6.86).0221.88 (0.89-3.95).1
Disease stage    
 IIB vs I/IIA1.6 (0.58-4.4).361.79 (0.74-4.32).2
 III/IV vs I/IIA1.64 (0.53-5.03).391.62 (0.59-4.47).35
Tumor differentiation    
 Poor vs well/moderate1.75 (0.75-4.08).192.08 (1.01-4.29).047
HP5, %0.99 (0.98-1.01).290.99 (0.98-1.01).3
IFP, mm Hg0.98 (0.92-1.04).50.99 (0.94-1.04).58
thumbnail image

Figure 3. (A) The probability of local relapse (Rel.) is divided into high and low expression categories using the median value for each hedgehog (Hh) gene. SHH indicates sonic hedgehog; HR, hazard ratio; CI, confidence interval; PTCH1, patched 1; SMO, smoothened, frizzled family receptor; IHH, Indian hedgehog; PTCH2, patched 2; GLI1, GLI family zinc finger 1. (B) The probability of local relapse is divided into high and low expression categories using expectation-maximization (EM) algorithm cutoff values for abnormal and normal gene expression. (C) Disease-free survival (DFS) is divided into high and low expression categories using the median value for each Hh gene.

Download figure to PowerPoint

Figure 2 illustrates the patterns of recurrence based on combinations of Hh genes that were up-regulated using the EM algorithm. Because the Hh genes may have a synergistic or antagonistic effect, we performed an analysis using a cutoff point of 3 up-regulated genes. Figure 4 indicates that women who had tumors with >3 Hh genes up-regulated were at higher risk of LR (HR, 2.56; 95% CI, 1.09-6.00; P = .026). However, there was no association with DFS.

thumbnail image

Figure 4. The probability of relapse (Rel.) recurrence is illustrated based on the number of up-regulated hedgehog (HH) pathway elements (cutoff point, >3 up-regulated HH genes). HR indicates hazard ratio; CI, confidence interval.

Download figure to PowerPoint

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SOURCES
  8. REFERENCES

Aberrant activation of the Hh pathway is increasingly being recognized as an important oncogenic signaling pathway in many epithelial tumors. Hh has been implicated in repopulation after radiotherapy29-31 and in the response to hypoxia.22-24 To date, published series have used immunohistochemistry to evaluate the Hh pathway in clinical samples. This approach has limitations, particularly the quantification of low levels of antigen expression and the possibility that some antibodies may have cross reactivity. In the current study, we used RT-PCR to demonstrate the up-regulation of 1 or more Hh pathway members in all but 1 of 85 cervical carcinoma samples compared with normal cervical tissues. Elevated SMO was associated with LR in both statistical models used. In an exploratory analysis, down-regulation of SMO and up-regulation of SHH appeared to be associated with tissue hypoxia. To our knowledge, this is the first study to investigate the Hh pathway in relation to hypoxia in clinical samples. Our data suggest that Hh may be a valid therapeutic target in cervical cancer and support other data suggesting a potential therapeutic role for targeting Hh in patients who receive chemoradiation.

All 6 Hh pathway members had a wide range of expression in the 85 tumors analyzed. The density of the tumor values for SHH, IHH, and GLI suggested a bimodal (mixture of 2 normal) distribution. Because of this observation, in addition to a conventional (using median values) statistical approach, we performed a unique analysis of the data by classifying the expression of each gene as likely to be normal or abnormal using the EM algorithm. RT-PCR is limited, in that it does not distinguish the origin (stroma vs tumor) of the up-regulated Hh members; however, to our knowledge, this is the first study to report on the patterns of Hh pathway up-regulation in individual tumors. A previous study of the Hh pathway in cervical cancer established increased expression using immunohistochemistry, but no association with clinicopathologic variables or outcome was included in the analysis.39 In our series of samples evaluated by RT-PCR, 89% of carcinoma samples had elevated levels of SHH, 85% had elevated levels of IHH, 25% had elevated levels of PTCH1, 22% had elevated levels of PTCH2, 33% had elevated levels of SMO, and 79% had elevated levels of GLI1. This extent of up-regulation was higher than that reported in other gynecologic malignancies examined by immunohistochemistry.13, 14 Up-regulation was observed in both squamous and adenocarcinoma histologic subtypes. The high levels of expression in cervical carcinoma (associated with human papillomavirus) are of interest, because viral infection has recently been implicated as a factor in Hh pathway activation.40

The correlations between Hh pathway genes and LR and DFS were analyzed using both median values and the EM algorithm cutoff points. Up-regulation of SMO was associated with an increased risk of LR (within the irradiated field; HR, 2.41; 95% CI, 1.00-5.82; P = .044). Although it did not reach statistical significance, the separation of the curves for LR (Fig. 3A,B) suggest that further investigation of SHH, GLI, and PTCH1 also may be warranted. In patients who had >3 genes up-regulated, there was an increased risk of LR (HR, 2.56; 95% CI; 1.09-6.00; P = .026); and, in all but 3 of these patients, both GLI and SMO were up-regulated. It is, however, possible that the number of genes up-regulated is simply a measure of the genetic instability and aggressiveness of the disease. Given the small numbers of any 1 combination of Hh genes in this analysis, it is not possible to comment on the effect of individual combinations, and we are exploring this further in primary xenograft models. Because the majority of patients in this study had squamous or adenosquamous carcinomas (only 8 women had cervical adenocarcinoma), it was not possible to determine whether the effect of Hh expression on outcome was the same across histologic subtypes. To ensure that our patients were representative, our clinical data and DFS associations were compared and were confirmed as consistent with observations in a larger series of 200 patients who received chemoradiation at our institution. No significant association between the Hh pathway (or other clinicopathologic variables, other than the degree of differentiation) (see Table 4) and DFS was observed, although the patients who had tumors with higher levels of SMO trended toward a worse DFS (Fig. 3C). Unlike studies in head and neck and esophageal cancers, GLI was not associated with outcome.30, 31

Reports from in vitro studies have implicated up-regulation of SMO and SHH in response to hypoxia.22-24 We observed an association between SMO and HP5 using both the median cutoff point (P = .0007) and EM algorithm models (P = .04). In our series, however, hypoxic tumors (high HP5) were associated with lower levels of SMO, suggesting that alternative mechanisms are responsible for up-regulating SMO. HP5 was not associated with outcome in this patient population, possibly reflecting the addition of cisplatin to the radiation. Conversely, higher levels of SHH were associated with hypoxic tumors, supporting the model in which hypoxia up-regulates SHH.

The outcome of women with cervical carcinoma is poor.5-7 Identification of those women at increased risk of recurrence who may benefit from more aggressive treatment strategies and the identification of new therapeutic targets are high priorities. The interactions of the Hh pathway in tumor cells are complex, and our study adds to the immunohistochemistry-based literature in clinical samples, with the EM model suggesting a novel methodology for analyzing data in tumors. Exploration of the relation between cervical cancer cells and stroma in xenograft models and tumor samples will help to further elucidate the biology involved. The level of up-regulation in the cervical cancers, combined with our data on local relapse, suggests that the Hh pathway is a potential therapeutic target in this disease, and we plan to investigate these observations further in primary xenograft models. Drugs targeting the Hh pathway are emerging as a new class of therapeutic agents. Drugs targeting SMO currently are entering the clinic in both phase 1 and 2 clinical trials.41 Our data support a role for the Hh pathway in local tumor recurrence; and this, together with data in other tumor types, suggests that inhibiting the Hh pathway alone or in combination with chemoradiotherapy may be a valid therapeutic strategy for further investigation in cervical carcinoma, and possible therapeutic trials are warranted in this area.

FUNDING SOURCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SOURCES
  8. REFERENCES

This work was supported by the Princess Margaret Hospital Foundation.

CONFLICT OF INTEREST DISCLOSURES

The authors made no disclosures.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SOURCES
  8. REFERENCES