• gestational trophoblastic neoplasia;
  • suppression subtractive hybridization;
  • cDNA microarray;
  • IGFBP1;
  • FTL


  1. Top of page
  2. Abstract


Hydatidiform mole (HM), the most common type of gestational trophoblastic diseases, can be considered as placenta with abnormal chromosome composition with potential of malignant transformation. Few biologic markers can predict subsequent development of persistent gestational trophoblastic neoplasia (GTN) requiring chemotherapy.


Suppression subtractive hybridization (SSH) combined with cDNA microarray was used to compare the differential expression pattern of HM that spontaneously regressed and that subsequently developed metastatic GTN. Tissue-specific chips were constructed from the subtracted cDNA libraries, followed by cDNA microarray analysis. Verification by quantitative RNA analysis by real-time polymerase chain reaction (PCR) and immunohistochemical analysis was performed in 23 genotyped complete HM.


Sixteen differentially expressed transcripts were identified. Quantitative RNA analysis confirmed down-regulation of ferritin light polypeptide (FTL) (P = 0.037) and insulin-like growth factor binding protein 1 (IGFBP1) (P = 0.037) in HM that subsequently developed GTN when compared with those HM that regressed. Immunohistochemical analysis further confirmed reduced IGFBP1 protein (P = 0.03) expression in HM that developed GTN.


Findings showed that reduced expression of genes related to cell invasion and immunosuppression, especially FTL and IGFBP1, were associated with development of GTN, and this finding may provide a better understanding of the pathogenesis of GTN. The potential application of FTL and IGFBP1 in management of patients with HM should be explored. Cancer 2005. © 2005 American Cancer Society.

Hydatidiform moles (HM) are characterized by cystic swelling of the chorionic villi and abnormal trophoblast proliferation.1–3 As the most common member of gestational trophoblastic disease, HM can be further classified into complete mole and partial mole based on cytogenetic, histopathologic, and clinical characteristics. Clinically, HM is managed by suction evacuation, followed by close monitoring of the level of human chorionic gonadotropin (hCG). Most cases of HM regress spontaneously after suction evacuation. However, about 8–30% of patients with HM, mostly complete mole, will develop persistent disease or gestational trophoblastic neoplasia (GTN) and will need chemotherapy.1–4

Serial hCG assay is thus the main parameter for diagnosis of GTN after identification of HM.5, 6 The whole monitoring period may last for more than one year and involves significant community resources and affects psychosocial well being of the patient.5 Therefore, identification of reliable and specific markers that could assist prediction of GTN will be beneficial to the management of patients with HM. Our previous studies demonstrated that lower apoptotic activity,7, 8 as well as higher telomerase activity in HM,9 was associated with development of persistent GTN. The differential gene expression pattern between regressive moles and persistent moles was then explored. Our study using the Atlas™ Human Apoptosis Array (Clontech; Palo Alto, CA), which includes 205 cDNA related to apoptosis, identified Mcl-1 to be a predictive factor for GTN.10

By using combined cDNA suppression subtractive hybridization (SSH) technique and microarray analysis in the current study, we attempted to expand the identification of differentially expressed genes in HM that subsequently developed GTN when compared with HM that spontaneously regressed.11 The SSH technique is sensitive and efficient for generating cDNA highly enriched for differentially expressed genes with both high and low abundance, because this technique combines both the suppression PCR amplification and cDNA subtraction, leading to generation of two subtractive libraries.12, 13 The efficient construction of tissue specific libraries from HM with different prognoses enables exploration of novel genes that may play a role in development of GTN. To the best of our knowledge, this is the first differential expression study adopting such a combined approach to compare HM that spontaneously regress and with those that develop aggressive disease.


  1. Top of page
  2. Abstract

Tissue Preparation and RNA Extraction

Hydatidiform moles were treated by suction evacuation in Queen Mary Hospital, the University of Hong Kong, after clinical and ultrasonographic diagnosis. The collection of trophoblastic tissue was approved by the Ethics Committee, the University of Hong Kong. Fresh molar vesicles were dissected and snap frozen. Histologic diagnosis was reviewed,1–3 and clinical follow-up data was retrieved. Persistent GTN was diagnosed if there was a plateau in hCG level for 4 weeks or if there was a further increase in hCG for 3 consecutive weeks after evacuation.4 The clinical and pathology data of the patients involved in different experiments of this study are summarized in Table 1. Previous fluorescent microsatellite genotyping and chromosome in situ hybridization studies revealed that all of these cases were complete moles.14, 15

Table 1. Clinical and Pathologic Parameters of Patients with Hydatidiform Moles
Case no.TestsAgeDiagnosisGestational ageDevelopment of GTNMetastasis
  1. S: SSH + microarray; R: real-time PCR; I: immunohistochemistry; CM: complete mole; GTN: gestational trophoblastic neoplasia.

2S, R, I51CM8RegressedNo
3S, R, I16CM7RegressedNo
4S, R, I45CM6GTNYes
5S, R, I29CM17GTNYes
6S, I34CM9GTNYes
10R, I48CM8RegressedNo
11R, I31CM10RegressedNo
12R, I31CM9RegressedNo
13R, I35CM15RegressedNo
14R, I37CM24RegressedNo
15R, I38CM12GTNNo
16R, I28CM6GTNNo

The histology and purity of the frozen samples was assessed before RNA extraction, and microdissections were performed to ensure that > 90% of tissues were trophoblasts. Total RNA was extracted from frozen tissue by Trizol (Invitrogen, Carlsbad, CA) with integrity checked by gel electrophoresis.10 Only samples with rRNA (28S/18S) ratio between 1.8 and 2 were used.

Self-Made Tissue-Specific Chips with SSH Subtracted Libraries

To reduce case bias, three cases of regressed moles and three cases of metastatic moles were pooled together in equal amounts for SSH by using the PCR-Select cDNA subtraction kit (Clontech, Palo Alto, CA).16 Forward and reverse SSH subtracted libraries were constructed from regressive moles and persistent moles. The regressed moles, in another SSH study, have also been used to compare with normal placentas to generate two separate sets of cDNA libraries.11 The subtracted cDNA were then cloned into pT-Adv vector (Clontech, Palo Alto, CA) and transformed into competent Escherichia coli. Individual white colonies (n = 260) from every library were further cultured. The individual clones were subjected to PCR amplification of the insert using the flanking primers (forward: 5′- TCG AGC GGC CGC CCG GGC AGG T -3′ and reverse: 5′- AGC GTG GTC GCG GCC GAG GT -3′). The PCR products containing the positive inserts were confirmed by agarose gel electrophoresis and purified with the QIAquick 96 PCR purification Kit (Qiagen GmbH, Hilden, Germany) before dotting. In total, 384 dots were printed onto the GAPS™ slides (Corning, Acton, MA) in duplication format with Cartesian Pro 3510 (Cartesian Technologies, Ann Arbor, MI), including PCR products of GAPDH for normalization.

cDNA Microarray Analysis

Microarray analysis was performed with MICROMAX TSA Labeling & Detection Kit (PerkinElmer Life and Analytical Sciences, Boston, MA) as previously described.11, 17 The signal intensity of each cDNA spot after subtraction of the background signal was normalized by using the intensity of the cDNA spots of GAPDH, bearing in mind that possible variant in GAPDH expression in individual samples might exist.18 Forward and reverse labeling and detection were performed to eliminate experimental bias. The consistent clones with high differential-expression ratio were sequenced and checked by the BLAST program.

Quantitative Real-Time PCR

Quantitative PCR19 was performed in duplicates with total RNA extracted from seven regressive moles and seven persistent moles using an iCycler iQ Multi-Color Real Time PCR Detection System (Bio-Rad, Hercules, CA). Primer sequences used for the PCR amplification and expected product size are listed in Table 2. The primer set of CGB, encoding the β subunit of hCG, was designed to amplify three major forms of CGB (3, 5, and 8).20, 21 The relative quantitative value was expressed by the 2-ΔΔCT method,22 representing the amount of specific gene expression with the same calibrators.

Table 2. Primer Sets for and Results of Quantitative Real-Time PCR Analysis
GenePrimer sequenceAnnealing temperature in °CExpected PCR product size in base pairsUniGeneReferenceP values
CGBTTTATACCTCGGGGTTGTGGGG56199NM_033043Xu et al.210.2593
IGFBP1AGCCGGCGCTCCGTGGCAGTG60298NM_000596Wulbrand et al.420.0379


Immunohistochemistry was performed on paraffin sections of 10 regressive moles and 10 persistent moles, by using the streptavidin-biotin complex immunoperoxidase method (Dako, Glostrup, Denmark), after antigen retrieval by microwave in citrate buffer. Polyclonal rabbit antibody for IGFBP-1 (Upstate, Lake Placid, NY) was applied at a dilution of 1:100. Negative control was prepared by replacing the primary antibody with Tris-buffered saline. A case of endometrial cancer with high IGFBP-1 expression was used as the positive control. The extent of immunochemical staining was categorized in four groups: 1) 0–25%; 2) 26–50%; 3) 51–75%; and 4) 76–100%.

Statistical Analysis

In cDNA microarray analysis, the difference between forward and reverse labeling and detection or between duplicated analysis in each run were performed by using an ANOVA test with normalized dot intensity. For the mRNA quantitative analysis, the relative gene expression (2-ΔΔCT) was compared with unpaired t test (Mann–Whitney test) on log-transformed value. Correlation of immunohistochemical data with clinical outcome of HM were analyzed by chi-square test. All statistical analyses were performed by using the PRISM 3.0 statistical software (GraphPad Inc, San Diego, CA). P-value of < 0.05 was taken to be statistically significant.


  1. Top of page
  2. Abstract

cDNA Microarray with Tissue Specific Chips and Sequence Identification

Similar scatter plot configurations were observed in regressive mole and persistent mole suggesting comparable gene expression patterns in regressive and persistent moles (Fig. 1). The ANOVA analysis (P > 0.1) indicated that there was no difference between the duplicated hybridization, and between different fluorescent labeling and detection assays. Between regressive and metastatic moles, 84 differentially expressed clones with ratio more than three were identified. After alignment with the GenBank database, 16 nonredundant transcripts were found, of which 15 corresponded to known cDNA, and 1 was classified as uncharacterized cDNA. All identified cDNA sequences showed > 96% homology to the known genes in GenBank with BLAST tools and are listed in Table 3. DNA sequencing indicated that some came from different regions of one complete transcript. The differentially expressed ratios between two types of mole were, in general, less than 5-fold except the ratio of IGFBP1 (−17-fold).

thumbnail image

Figure 1. Screening of subtracted libraries for differentially expressed genes in regressive and persistent moles. The scanned hybridization images and scatter plot of their quantitative result are shown together. For comparison, the hybridization was repeated with (A) first labeling and (B) reverse labeling.

Download figure to PowerPoint

Table 3. Differentially Expressed Genes Between Regressive Mole and Persistent Mole
Gene symbolGene nameTimes of sequencingAverage fold (PM/RM)aAverage Intensity of PM (cy5)Average Intensity of RM (cy3)UniGene
  • RM: regressive mole; PM: persistent mole.

  • a

    The average fold is from the ratio of all cDNA fragments from the same Unigene. The ratio is calculated from the division of normalized intensity of persistent moles by intensity of regressive moles. The positive number stands for the overexpression; the negative number is for down-regulation in persistent mole.

CGB5Homo sapiens chorionic gonadotropin, beta polypeptide 5 (CGB5)63.6203226531NM_033043
CGB3Homo sapiens chorionic gonadotropin, beta polypeptide 3 (CGB3)72.9195406741NM_000737
CGB8Homo sapiens chorionic gonadotropin, beta polypeptide 8 (CGB8)52.8201849108NM_033183
ESTHomo sapiens, hypothetical protein FLJ11273, clone MGC:33727 IMAGE:5263860, mRNA, complete cds1-2.913193707NM_018374
ANXA2Homo sapiens annexin A2 (ANXA2), mRNA1-3.0287676NM_004039
RPS13Homo sapiens ribosomal protein S13 (RPS13)3-3.14301262NM_001017
RPL31Homo sapiens ribosomal protein L31 (RPL31)1-3.216985408NM_000993
MTCO2Homo sapiens mitochondrion cytochrome c oxidase subunit II2-3.26251980NM_173705
RPS7Homo sapiens ribosomal protein S7 (RPS7)4-3.217145327NM_001011
SERPINE2Homo sapiens serine (or cysteine) proteinase inhibitor, clade E(nexin, plasminogen activator inhibitor type 1), member 2(SERPINE2)1-3.34671660NM_006216
B2MHomo sapiens beta-2-microglobulin (B2M)3-3.4329910735NM_004048
HSP90AHuman heart mRNA for heat shock protein 901-3.75832151NM_005348
A2MHomo sapiens, alpha-2-macroglobulin, clone MGC:47683 IMAGE:6056126,1-4.165245NM_000014
FTLHomo sapiens ferritin, light polypeptide (FTL)1-4.16802769NM_000146
SPP1Homo sapiens gene for osteopontin,4-4.44892041NM_000582
IGFBP1Homo sapiens insulin-like growth factor binding protein 129-17.135751NM_000596

Validation of Differentially Expressed Genes with Quantitative Real-Time PCR

A subset of 8 differentially expressed genes was chosen for validation by quantitative real-time PCR analysis (Table 2), after taking into consideration additional data from Stanford Human cDNA array analysis (Data not shown). The eight genes included chorionic gonadotropin beta polypeptide (CGB), annexin A2 (ANXA2), mitochondrion cytochrome C oxidase subunit II (MTCO2), β-2-microglobulin (B2M), heat shock protein 90 (HSP90A), ferritin (FRN) light polypeptide (FTL), osteopontin (SPP1), and insulin-like growth factor binding protein 1 (IGFBP1). The quantitative real-time PCR validation showed essentially similar expression trend with microarray analysis. However, a statistically significant difference between regressive mole and persistent moles (GTN) was demonstrated only in the expression of FTL (P = 0.037) and IGFBP1 (P = 0.037) (Fig. 2) but not the other genes (Table 2). Commercially available antibody was available for IGFBP1 but not for FTL. Thus, the expression of IGFBP1 was selected for further immunohistochemistry analysis.

thumbnail image

Figure 2. Relative expression levels of differentially expressed genes (A) FTL and (B) IGFBP1 in regressive moles and persistent moles as evaluated by quantitative real-time PCR and compared with a reference gene (GAPDH) to correct for variation in the amount of RNA. The significance of any difference in expression level was tested by the Mann–Whitney units test.

Download figure to PowerPoint

Down-Regulation of IGFBP1 in Persistent Mole by Immunohistochemistry

IGFBP1 protein expression was demonstrated in cytotrophoblast (CT), syncytiotrophoblast (ST), and extravillous trophoblast (EVT) in HM, but it was predominately stronger in syncytiotrophoblast (Fig. 3, Table 4). There was significant down-regulated expression of IGFBP1 (P = 0.03) in persistent moles.

thumbnail image

Figure 3. Immunohistochemical analysis of expression of IGFBP-1 demonstrated weaker expression in (A) persistent mole when compared with (B) regressive mole.

Download figure to PowerPoint

Table 4. Analysis of Extent of Immunochemical Staining of IGFBP-1
 2 (26-50%)3 (51-75%)4 (76-100%)
  1. ST: syncytiotrophoblast, CT: cytotrophoblast, EVT: extravillous trophoblast.

Regressive mole (n = 10)10
Persistent mole (n = 10)325
Regressive mole (n = 10)10
Persistent mole (n = 10)5-5
Regressive mole (n = 10)10
Persistent mole (n = 10)46


  1. Top of page
  2. Abstract

In this differential expression study with combined SSH and cDNA microarray analysis, eight genes were selected for confirmatory quantitative RNA study but statistical significance was only demonstrated in FTL and IGFBP1. Reduced IGFBP1 protein expression in HM that developed GTN was further confirmed. This study may be limited by the sample size and availability of antibody.


IGFBP-1 is a glycoprotein in the extracellular matrix involved in the regulation of trophoblast invasion by means of two opposite mechanisms. IGFBP-1 can stimulate the migration of human trophoblast cell migration23, 24 through phosphorylation of focal adhesion kinase and the mitogen-activated protein kinase pathway.25 Conversely, IGFBP1 may interfere with the effect of IGF-II, inhibit adenylyl cyclase through IGF-RII, and cause a decrease in intracellular cAMP, which is necessary for cell invasion of trophoblast.26

In an in vitro coculture model, IGFBP-1 from decidua was found to inhibit the invasion of trophoblast.27 In addition, raised maternal plasma concentrations of IGFBP-1 were demonstrated in preeclampsia patients, a disease known to be related to faulty trophoblast invasion.28 In HM with pure or dominant paternal genome, the over-expression of paternally transcribed growth factor, IGF2, was postulated to enhance the invasion of trophoblast and contribute to the malignant potential of HM.29 The down-regulation of IGFBP1 in persistent mole, as demonstrated in the current study, may promote the trophoblast invasion triggered by endogenous IGF2 and contribute to the subsequent relatively aggressive behavior.

Besides IGFBP1, down-regulation in other invasion related genes, including ANXA2, A2M, SERPIN2, and SPP1, were also detected in this SSH-cDNA microarray study, although statistical significance could not be reached. Annexin A2, a Ca2+-binding protein, is involved in remodeling extracellular matrix and modifying cell migration.30 SPP1, encoding osteopontin, may regulate implantation.31 Actually, reduced expression of SPP1 and its protein osteopontin in HM when compared with normal placenta has been suggested to contribute to the pathogenesis of molar pregnancy.11, 32 Similarly, our earlier study has also demonstrated reduced expression in association with promoter hypermethylation of tissue inhibitor of metalloproteinase three (TIMP3) in HM and choriocarcinoma.18, 33

B2M and FTL

Two genes involved in the immunosuppression of placentation, B2M and FTL, were found to be down-regulated in persistent mole, although statistical significance was only reached in FTL. The genome of HM was either entirely or predominantly paternally derived, representing an allograft that might provoke additional maternal immune response. FTL is a light polypeptide of FRN and may play a role in the development of inflammation and malignancies,34 including that of trophoblast. FRN may protect trophoblast from free radical formation through sequestering free Fe2+.35 Acidic isoferritin could diminish the immunosuppressive activity through combination with activated CD4+ and CD8+ T lymphocytes.36 Down-regulation of FTL, together with down-regulation of B2M,37 may alter immunosuppression in persistent moles, facilitating the progression of trophoblast.


Serial hCG assay remains the basis of diagnosis of GTN, and pretreatment hCG level is one of the criteria determining staging of GTN.38 Up-regulation of CGB3/5/8, encoding β subunit of hCG20 was detected in aggressive HM by SSH-microarray study but did not have statistical significance in quantitative RNA study. The findings concurred with the belief that it is serial assay of hCG that correlates with development of GTN and not a one-time level.

MTCO2 (Mitochondrion Cytochrome C Oxidase Subunit II)

The modulation of MTCO2 has been suggested to reflect an altered energy metabolism and has been implicated in tumorigenesis. Indeed, mitochondrial genomic aberration was commonly found in cancer cells including those arising from the female genital tract.39, 40 Our previous study has also demonstrated mtDNA instability in choriocarcinoma and HM.41

In summary, the differentially expressed genes identified in HM that subsequently developed GTN, seem to be involved in modulating tumor invasion and immunosuppression. Further study involving larger numbers of clinical samples and more quantitative analysis may need to be conducted to explore the potential application of these markers in clinical management of patients.


  1. Top of page
  2. Abstract