Frequency and genome load of Epstein-Barr virus in 509 breast cancers from different geographical areas

Since the few data exploring a possible association between Epstein–Barr virus (EBV) and breast cancer are conflicting, we investigated this association together with the influences of geographical areas. 509 breast cancers were sampled from areas with varying risks of nasopharynx carcinoma (NPC) such as North Africa (Algeria and Tunisia, high-risk area); southern France (Marseille, intermediate-risk area); and northern Europe (northern France, the Netherlands and Denmark; low-risk areas). Polymerase chain reaction (PCR) of a subregion of EBV BamHIC encoding the EBERs demonstrated that 31.8% of the tumours contained the viral genome. No significant differences were observed among the geographical areas. However, positive samples showed higher loads of the EBV genome in the NPC high- and intermediate-risk areas than in the low-risk areas. EBV type 1 was the dominant strain. In situ hybridization studies using a35S-labelled riboprobe for EBER1 and a laser capture microdissection, combined with quantitative PCR, showed that EBV localization was restricted to some tumour epithelial cell clusters. EBV could not be detected in the stroma. Considering the whole population covered, the presence of the EBV genome was not correlated with age, menopausal status, tumour, size, nodal status or histological grade. © 2001 Cancer Research Campaign http://www. bjcancer.com

previous study (Romain et al, 1997). Patients from Denmark (n = 97) were registered in the high-risk group of the Danish Breast Cancer Cooperative Group program (Andersen et al, 1981).

Tissue specimens
Breast tumour specimens were obtained at surgery and stored in liquid nitrogen. Samples were all processed in Marseille (laboratoire de Transfert d'Oncologie Biologique, Assistance Publique-Hôpitaux de Marseille). Tumours from other areas were shipped on solid carbon dioxide to this institution. Tumours were prepared for DNA extraction (Sambrook et al, 1989) and cryosections. Fixed paraffin-embedded tissues were obtained in specific cases, according to the results of BamHIC PCR.

PCR analysis of EBV genome
Qualitative BamHIC PCR The synthetic primers, sense 5′-AACCTCAGGACCTACGCT-3′ and antisense 5′-TAGCGGACAAGCCGAATAC-3′, described by Labrecque et al (1995), were used for BamHIC PCR. The amplified fragment is 501 base pair (bp). Aliquots of each PCR product were electrophoresed on a 1% agarose gel in Tris/borate buffer containing 0.5 µg ml -1 ethidium bromide, photographed, and prepared for Southern blot (Sambrook et al, 1989). The filters were hybridized with a random-primed 200 bp fragment of EBER1 cDNA (kindly provided by BE Griffin and LG Labrecque), washed as previously described (Ouafik et al, 1989), and exposed to film to verify the identity of the bands.

Semi-quantitative BamHIC PCR
To quantify the EBV genome load in the tumours, DNA extracted from the Namalwa cell line, containing two integrated EBV copies per cell, was used. 150 × 10 3 Namalwa cells yielded 1 µg of DNA. Serial dilutions of DNA were prepared from 30.0 to 0.1 ng, equivalent to 9000-30 copies of EBV genome, respectively. Data are expressed as the number of BamHIC copies/500 ng DNA.

In situ hybridization (ISH) procedure
Frozen and paraffin-embedded sections (10 µm) were used for ISH studies. Paraffin-embedded sections were dewaxed in toluene, rehydrated in graded ethanol, digested with 10 µg/ml proteinase K (Roche Molecular Biochemicals, Meylan, France) for 30 min at 37˚C. Sections were mounted on subbed glass slides and prepared for hybridization to RNA probes, as previously described (Ouafik et al, 1990). Radiolabelled riboprobes were prepared using uridine 5′-[α-35 S-thio] triphosphate (Amersham Pharmacia Biotech, Uppsala, Sweden) and T3 or T7 polymerase (Stratagene, La Jolla, CA) to synthesize, respectively, RNA sense and antisense transcripts of 200 bp EBER1 cDNA contained in pBluscript plasmid (kindly provided by BE Griffin and LG Labrecque). Hybridized sections were exposed to Ilford K5 autoradiography emulsion for a variable length of time (1 week-1 month) to confirm the heterogeneous labelling patterns.
Images were recorded using a PixCell II system (Arcturus Engeneering, Mountain View, USA) which incorporates an Olympus IX-50 microscope and a single sheep CCD colour camera (COHU). Digitalization was performed via the colour frame grabber Matrix Meteor TM slotted to a microcomputer. The public domain software Image, written by W Rosbard at the National Institute of Heath (NIH), was used for the image processing. The mean grain density was quantified by morphological filtering and thresholding, and expressed as percentage of surface for both the tumour epithelial formations and the stroma.

Laser capture microdissection
OCT embedded tissue blocks prepared from frozen breast cancer samples were sectioned at 8 µm in a cryostat. The sections mounted on uncoated glass slides were fixed in 70% ethanol for 30 s, stained with Mayer's haematoxylin, and washed in 70% and 95% ethanol for 60 s. Subsequently, slides were stained with eosin Y. Staining was followed by two 60 s dehydration steps in 95% and 100% ethanol and two final 15 min dehydration steps in xylene. Once air-dried, the sections were microdissected with a PixCell II LCM system (Arcturus Engeneering, Mountain View, USA). For each tumour analysed, several epithelial areas (approximatively 5 × 10 3 cells) were independently captured; stromal areas without infiltrating malignant epithelial cells were pooled to provide a sufficient number of GAPDH copies. Cell populations were estimated to be homogeneous as determined by microscopic visualization. DNA from laser captured cells was extracted and subsequently used for Q-PCR.

Statistical analysis
Patients' characteristics were compared with the Chi-square test. P < 0.05 was considered significant.

PCR analysis of EBV genome
In order to detect the presence of EBV genome in breast cancers, genomic DNA was prepared and used for BamHIC PCR analysis. A 501 bp amplified product was obtained (Figure 1). The presence of the BamHIC subregion of the EBV genome was used to define EBV positivity. Among the 509 breast cancers analysed, only 31.8% were EBV positive. The EBV-negative tumours included samples shown with low, moderate and high degree of lymphoplasmocytic infiltration by histopathology. In addition, normal tissues adjacent to breast cancers (n = 3) and benign breast tumors (n = 7), were investigated by PCR and were all found to be negative.
Breast cancers showed a positive ratio from Algeria of 40.0%, and from Tunisia 32.8%, Marseille 26.7%, St Cloud 29.2%, The Netherlands 32.5% and Denmark 35.1%, respectively. Frequencies were not different among the geographical areas studied. However, EBV-positive breast cancers from the NPC high-and intermediate-risk areas showed higher loads of the EBV genome than those from the low-risk areas: that is 47.0% and 28.1% of the samples contained more than 150 BamHIC copies/500 ng DNA, respectively (P = 0.01) and 22.7% and 8.3% more than 1500 BamHIC copies/500 ng DNA (P = 0.01) ( Table 2).

Detection of EBV genome and tumour characteristics
In the overall population, no significant link was observed between the detection of the EBV genome by BamHIC PCR and age, menopausal status, tumour size and lymph node involvement. No association was also demonstrated with histoprognostic grade; in Marseille the highest EBV frequency was found in grade 3 (35.6%) and the lowest in grade 1 (22.2%), but differences were not significant. The load of the EBV genome was also not correlated with clinical and pathological characteristics. It is important   Annealing was carried out at 56˚C for 30 s followed by extension at 68˚C for 35 s. The products were fractionated on a 1% agarose gel and prepared for Southern blot analysis using a 200 bp DNA probe as described in 'Materials and Methods'. A control lacking DNA was subjected to PCR at the same time (-). Samples 1, 2, 7, 8, 12, 13, 15: <150; 3, 4, 9, 14, 16, 17: 150-1500; 5, 6, 10, 11: >1500 BamHIC copies/500 ng DNA to point out that 40.0% of each inflammatory and non-inflammatory Algerian samples were EBV positive; only inflammatory carcinomas (41.7% vs 28.6%) showed high rate of EBV positivity in Tunisia.

EBV genotyping
EBV genotyping was applied to 13 breast cancers from Algeria, Marseille, the Netherlands and Denmark containing more than 150 BamHIC copies/500 ng DNA. EBV1 was present in all the tumours examined. EBV2 was detected in 2 EBV1-positive samples from the Netherlands (not shown).

In situ hybridization
20 EBV-positive breast cancers from Algeria, Tunisia and Marseille were prepared for ISH studies in order to determine whether EBV expression occurs in the tumour epithelial compartment or in lymphoplasmocytic cells infiltrating the stroma. Positive results were obtained on the malignant epithelial compartment of those tumours with more than 1500 BamHIC copies/500 ng DNA (n = 10). Considerable heterogeneity was however observed among malignant epithelial areas. Heterogeneity was already detectable using 1 week exposure and was still evident when the exposure time was extended to 1 month. Figure 2A shows a section hybridized with an antisense probe. Silver grains were concentrated over the tumour epithelial formations, whereas the stroma displayed fewer isolated grains; mean grain density was 9.4% and 1.7%, respectively. Such a pattern was considered to express the specific localization of EBV material, since the close adjacent section hybridized with the sense riboprobe, representative of the technical background for the exposure time, also contained a low density of randomly distributed silver grains both over the tumour epithelial formations (1.9%) and the stroma (1.2%) ( Figure 2B). Figures 2C and 2D show two distant fields of the same highly heterogeneous section hybridized with the antisense riboprobe. The epithelial areas in field 2C contained a high density of silver grains (9.5%) corresponding to EBV positivity whereas the low density (1.8%) obtained in field 2D corresponds to negativity. For both fields 2C (1.5%) and 2D (0.9%), a low grain density was observed in the stroma. For the same tumours, signal intensity was lower in paraffin-embedded sections than in cryosections.

Epstein-Barr virus in breast cancer 787
British Journal of Cancer (2001)  In situ hybridization for EBV in breast cancer. Cryostat tumour sections (10 µm-thick) were hybridized with 35 S-labelled antisense and sense riboprobes, dipped in photographic emulsion, exposed for 1 month, developed, and lightly counterstained with haematoxylin. (A) shows a section hybridized with the antisense riboprobe. Silver grains are concentrated over the tumour epithelial formations. A close adjacent section from the same tumour shown in (B), was hybridized with the sense riboprobe and displays a low density of randomly distributed silver grains. (C) and (D) show two distant fields of the same section hybridized with the antisense riboprobe. The epithelial areas in field C contain a much higher density of silver grains compared to those in field (D). Calibration bar for all micrographs = 50 µm 8 of the breast cancers, shown by PCR to be EBV negative were examined by ISH and showed non-specific labelling.

Laser microdissection combined with real-time quantitative PCR
In order to perform precise molecular analysis of purified cell populations and determine whether EBV is localized in the stroma or in the tumour epithelial compartment, as suggested by ISH, two EBV-positive breast cancers from Marseille were analysed combining LCM and Q-PCR. Figure 3 shows a target tissue before laser shots ( Figure 3A) and the captured malignant epithelial cells ( Figure 3B). In both tumours some malignant epithelial areas were highly positive, with EBV loads ranging from 128 to 5816 BamHIC copies/100 ng GAPDH; however, other malignant epithelial areas independently procured from the same case were EBV-negative. GAPDH did not significantly differ between positive and negative epithelial areas. The stroma areas were EBVnegative (Table 3).

DISCUSSION
Our results demonstrate that the EBV genome can be detected in some breast cancers with a predominance of the EBV1 subtype. Overall, 31.8% of our 509 tumour samples were positive for the BamHIC non repetitive sequence. Labrecque et al (1995) reported conservatively that 21% of breast cancers were positive for the BamHIW repeat sequence, compared to 51% in Bonnet's study (1999). Elsewhere, absence of BamHIW has been reported in 10 medullary carcinomas (Lespagnard et al, 1995). In 37 tumours analysed in the current study (not shown), the concordance between BamHIC and BamHIW primer pairs for PCR analyses was 83.7%.
In our overall population, presence of the EBV genome was not correlated with age, menopausal status, tumour size, nodal status, or histological grade. In Marseille, there was however a trend toward a positive association between EBV and grade, as reported by Bonnet et al (1999). High rates of positive tumours were observed in young Algerian patients and in inflammatory samples.
The ISH using a 35 S-labelled riboprobe for EBERs demonstrated that EBV expression was exclusively localized in malignant epithelial cells. The samples found strongly positive by PCR were also positive by ISH. Within individual tumours, a high heterogeneity was however observed among epithelial cell clusters. No labelling was obtained for samples that were negative by PCR. In our study radioactive ISH signal intensities were low in paraffin-embedded sections. The absence of labelling that has been reported by others (Chang et al, 1992;Lespagnard et al, 1995;Glaser et al 1998) could be a result of technical problems, such as the fixation as reported by Penault-Llorca et al (1994), or of the low sensitivity of non-radioactive ISH.
Laser capture microdissection (LCM) was combined with Q-PCR to quantify the EBV genome in specific cell populations of the tumours. This is the first time that this technique has been used to localize EBV in breast cancers. The use of LCM clearly confirmed the epithelial localization of EBV and the heterogeneity among epithelial cells, as suggested by ISH. The highest loads of EBV genome found here in malignant epithelial populations captured from EBV-positive breast cancers are in the median range of those observed for NPC collected in Maghrebian countries, when both are expressed as the number of BamHIC copies/100 ng GAPDH (not shown). It is unlikely that our PCR-positive results are due to contamination. First, EBV was detected in 2 independent amplifications, semi-quantitative PCR with Southern blot hybridization and Q-PCR combined with LCM; secondly, Q-PCR was run in a closedtube system and required no post-amplification manipulation; finally, normal human genomic DNA (Roche Molecular Biochemicals, Meylan, France) and controls lacking DNA always remained negative in PCR analyses.
Detection of EBV in breast cancer has engendered a great deal of controversy. In a recent note, Brink et al (2000) suggested that positivity with EBV is most likely caused by the presence of some infected lymphocytes in the tumour samples. However, it is possible that the RT-PCR used by these authors was not sensitive enough to detect all the EBV transcripts (Joab, 2000). In our study, the presence of EBV in tumoral lymphoplasmocytic infiltration can be ruled out. First, ISH demonstrated no specific labelling of the stroma and the infiltrating lymphoplasmocytic cells near the positive tumoral epithelial clusters; secondly, LCM combined with Q-PCR also showed that EBV localization was restricted to some tumour epithelial cell clusters; thirdly, the tumours found to be EBV negative using BamHIC PCR included samples with low, moderate, and high degrees of lymphoplasmocytic infiltration; finally, EBV was not detected in normal tissues adjacent to breast cancers and benign breast tumours, in agreement with previous studies (Labrecque et al, 1995;Bonnet et al, 1999).
One hypothesis mentioned by Labrecque et al (1995) for the inconsistency concerning the presence of EBV in breast cancer is that EBV-positive tumours might belong to a subgroup of breast cancers. Our study shows that EBV frequencies were not different among the geographical areas analysed. In contrast, the load of EBV genome was significantly higher in NPC high-risk areas (Maghrebian countries), and intermediate-risk areas (southern France) where migrants of Maghrebian origin and French people born in Maghreb account for a large proportion of the population. Although differences in the patient characteristics among the participating centres suggest that caution should be used in making comparisons by country, environmental factors inherent to life style and food habits that are not modified by migration, and inherit susceptibility (Steinitz et al, 1990;Bouchardy et al, 1996;Parkin and Iscovich, 1997), may partly explain the observed differences.
In summary, we confirm the presence of EBV in breast cancers. Our findings demonstrate that at least 31.8% of 509 samples of invasive ductal carcinomas contain a subregion of BamHIC, encoding EBERs, and that the load of intratumoral EBV genome differs according to the geographical area. ISH and LCM studies show that the EBV is located in some tumour epithelial cell clusters, with no specific detection in lymphoplasmocytic infiltrations.