Formalin-fixed, paraffin-embedded, surgical biopsy specimens were obtained from histopathologically normal breast tissues (from patients undergoing elective breast reduction mammoplasty with no known history of breast cancer), neoplastic and non-neoplastic adjacent breast tissues from breast cancer patients from the pathology tissue procurement archives of the University of Alabama at Birmingham and the Birmingham Veterans Affairs Hospital. The ages of breast cancer patients ranged from 22-80, (mean = 48). Ages ranged from 20-57 in the normal breast (reduction mammoplasty) group (mean = 36). All samples were obtained in accordance with ethics guidelines from the institutional review board of each institution. A faculty pathologist (K. K.) re-examined all cases to confirm histological diagnosis. Since these paraffin embedded specimens were part of the patient record, we were not allowed to exhaust the specimen. Some specimens had only limited amounts of tissue available, and we were therefore not able to perform all of the immunohistochemical and in situ hybridization studies on each specimen. For this reason the total number of specimens used exceeded the number analyzed for any given reagent.

4-μm paraffin sections were procured from biopsy specimens of neoplastic and non-neoplastic breast, deparaffinized in xylene, and hydrated them in graded alcohols. Samples were processed as previously described 25 using monoclonal antibodies for immunodetection (anti-IE1/2 ["IE"; IgG₁ isotype, 1:40, Chemicon, Temecula, CA], anti-CMV cocktail, containing antibodies specific for early and late antigens ["E/L"; IgG_2α isotype,1:40, Innovex Biosciences, Richmond, CA], and anti-CMV late antigen ["L"; IgG_2α isotype, 1:40, Chemicon, Temecula, CA]). As controls we used anti-CD34 (1:40, BioGenex), and anti-smooth muscle actin (1:40; BioGenex, San Ramon, CA), and omission of primary antibody (no antibody). Immunostaining with the different antibodies was performed in a blinded fashion with respect to tissue diagnosis. One pathologist (K. K.) who was blinded to the antibody used analyzed immunostaining results and sections were called positive if specific immunoreactivity was detected.

To confirm that HCMV nucleic acids were present in the pathological sections, the investigators performed in situ hybridization with a commercially available HCMV oligonucleotide cocktail probe labeled with fluorescein (BioGenex/Innogenex, San Ramon, CA). This probe consisted of six fluorescein isothiocyanate (FITC)-conjugated 40-mer probes spanning coding regions within the HCMV IE1 gene, and did not hybridize with human DNA sequences. Positive control probe specific for human Alu DNA sequences and negative control probe specific to an insect virus genome, both provided by the manufacturer, were also used. 4-μm paraffin sections were cut, deparaffinized, and hydrated through an established series of graded ethanol. Status of fixation was assessed for all cases before proceeding, and sections were post-fixed in formalin if necessary. After treatment, the prepared slides were rinsed in distilled water, dehydrated to 100% ethanol and air dried. Prediluted probe was then placed onto the sections, a cover slip was applied and slides were denatured on a MISHA thermocycler (Shandon Lipshaw/Hybaid Omnigene) at 90°C for 8-10 minutes, and then hybridized at 37°C in humidified chamber overnight. Slides were washed in TBST buffer, subjected to probe wash (0.05% SSC buffer 20 min. at 40°C), then washed in 1× PBS. Endogenous avidin, biotin and Fc receptors were then blocked using avidin/biotin blocking kit (BioGenex, San Ramon, CA) and Fc block (Innovex Biosciences, Richmond, CA). Fluorescein-labeled probe was then detected using Supersensitive^® in situ detection system (Innogenex, San Ramon, CA) with the chromogen BCIP/NBT.

DNA was purified from paraffin sections (3-6 10 μm sections) cut from a subset of the same biopsy specimens described above using DNeasy Tissue System (Qiagen, Valencia, CA) per manufacturer's instructions. To avoid potential PCR contamination, these experiments were performed in a laboratory devoid of previous exposure to infectious or recombinant HCMV. All preparations were processed throughout in a blinded fashion; no positive controls were used in any PCR reactions and blank paraffin blocks were cut sequentially between each patient sample and processed identically. For preparation of each case, the sectioning blade was replaced and the cutting surface was cleansed with xylene and ethanol. From each sample, 100-250 ng of DNA was amplified by nested PCR using internal and external primers specific for HCMV glycoprotein B (UL55) gene as described 39 . Samples were considered positive when a band of 140 bp size could be visualized on agarose gel with ethidium bromide. Amplified DNA products were visualized on 1.5% agarose gels with ethidium bromide, bands were cut out, and DNA was extracted (gel extraction kit, Qiagen, Valencia, CA) and analyzed by automated sequencing (ABI Model 377 DNA Sequencer, Foster City, CA). Confirmation of HCMV sequence was performed using a NCBI Blast search.

Immunohistochemical data were determined in breast tumor samples, breast tumor control samples, and normal control samples. The percentages of positive sections for each specific monoclonal antibody in each group of samples were estimated. Chi-square test or Fisher's exact test was applied to compare tumor samples or tumor controls samples to normal control samples.

To determine whether HCMV was present in breast epithelium from normal controls and breast cancer patients, archived formalin fixed paraffin embedded breast tissues were utilized. Non-neoplastic breast tissues from breast reduction mammoplasty patients was used as "non-neoplastic normal" control (N = 38). To evaluate HCMV presence in neoplastic tissues, the investigators obtained tumor biopsy specimens (N = 39) and histopathologically non-neoplastic biopsy specimens from many of the same patients ("tumor control", N = 21).

To detect HCMV protein expression, we performed immunohistochemistry on as many specimens as possible with a panel of monoclonal antibodies that were specific to HCMV antigens expressed at different stages of the viral life cycle. Due to limitations on the use of archived patient specimens and/or tissue quality, not all specimens could be evaluated for all antibodies and in situ hybridization. The antibodies used are specific for HCMV immediate early ("IE"), early and late ("E/L"), or late antigens ("L") (Figures 1 and 2). As controls, we used monoclonal antibodies specific for smooth muscle actin (SMA) (data not shown) and CD34 (Figure 1, l). These antibodies react to smooth muscle cells and vascular endothelial cells, respectively, but not breast epithelial cells, and serve as IgG_2α and IgG₁ isotype positive and negative monoclonal antibody controls, respectively. As additional negative controls, we performed immunostaining in the absence of primary antibody for all cases (data not shown).

We detected HCMV-IE antigens in neoplastic epithelium from 97% of breast cancers tested while IE antigens were detected in only 63% (p = 0.0009) of normal control cases tested (Table 1, Figure 1). IE immunoreactivity occurred in normal appearing ductal epithelium (Figure 1, a-b), epithelium in DCIS, and in IDC epithelial cells (Figure 2, a) evident in tumor specimens, but IE immunoreactivity was not observed in stromal cells (Figure 1, a-f). Immunoreactivity with the HCMV-E/L monoclonal antibody cocktail was detected in 84% of breast cancer specimens tested; only 21% of normal breast controls had confirmed immunoreactivity (p < 0.0001; Table 1). The HCMV-E/L immunoreactivity was detected in a similar cellular pattern as that of IE antigen, with a staining pattern restricted to epithelium (Figure 1, g-h; Figure 2, b). Monoclonal antibody immunoreactivity with HCMV-L was detected in 56% of breast cancer specimens and 39% of normal controls (p = 0.227; Table 1). The pattern of late antigen cellular localization was also similar to IE and E/L immunoreactivity (Figure 1, j, k; Figure 2, c). HCMV late antigen was detected, in general, at a less intense level, although occasional rare cells were very intensely positive for late antigen (Figure 2, c). In some of the tumors there were areas of homogenous low level immunoreactivity to IE, E/L and L antigens, while in other tumors there were scattered foci of positive tumor cells. The intensity of immunostaining varied significantly from tumor cell to tumor cell within any given tumor (e.g., Figure 1, j, k; Figure 2).

In general, we detected little HCMV immunohistochemical staining of stromal fibroblasts in the tumor cases. We did, however, perform immunostaining on a subset of tumor cases with a monoclonal antibody specific to the HCMV pp65 tegument antigen. In several cases we detected intense immunostaining of infiltrating stromal macrophages with this antibody (see Additional file 1).

Immunoreactivity was not detected in breast epithelium when primary antibody or monoclonal antibodies specific for smooth muscle actin and/or CD34 were used (Table 1), although immunoreactivity was evident in smooth muscle cells and endothelial cells with these antibodies as expected (e.g., Figure 1, l). The HCMV-IE, E/L and L monoclonal antibodies were specific for HCMV antigens when tested against known HCMV infected lung tissues from an AIDS patient (e.g, Figure 1, i).

Our analysis also demonstrated that there was a higher incidence of HCMV immunoreactivity in "tumor control" tissues from breast cancer patients than in "normal control" tissues from breast reduction mammoplasty (Table 2). Interestingly, while the incidence of HCMV IE immunoreactivity trended to a higher level in the breast cancer patients' control tissues compared to the normal control tissues, it was not statistically significant. The incidence of HCMV E/L and L immunoreactivity was, however, significantly greater in the tumor control breast tissue compared to the normal control tissues (Table 2).

Of the tumor specimens and matched breast control specimens from tumor patients, a total of 30 specimens were analyzed with all three antibodies (25 tumor cases and 5 matched controls). 22 of the true normal breast tissues (from reduction mammoplasty) were also analyzed with all three antibodies. In all but one case where all three antibodies were tested in the tumor specimens, the specimens were positive for IE antibody staining (Table 1). In 3 of these cases both the E/L and L antibody staining was negative, otherwise it was present. In 2 other cases either E/L or L antibody staining was negative while the other two antibodies were found to be positive.

In contrast, while 15 of the 22 true normal breast specimens from reduction mammoplasty patients were positive for IE antibody staining, only 7 of these 15 cases were also immunoreactive for E/L and L. These findings suggest that HCMV early and late antigen expression is less frequently found in normal breast tissues than in tumor tissues or normal appearing breast tissues from breast cancer patients.

To determine if the HCMV positivity rate of the tissue samples was due to the differences in the mean age or ethnicity of the control versus the cancer patient populations, we analyzed the data by assessing the HCMV IE1 immuonreactivity rate of patients ≤45 years or >45 years of age. In ≤45 years group, 7 of 8 (87.5%) cancer patients had positive HCMV IE1 antigen, compared to 12 positives of 17 (70.6%) normal tissue, and the difference was not significant. In >45 years group, all 20 cancer patients were positive for HCMV IE1 antigen, compared to 4 positives out of 7 normal tissues (p = 0.012). Thus, these data remained significant for patients >45 years of age.

There were 7 African-American patients and 31 Caucasian patients in the breast cancer group. There were 11 African-American and 21 Caucasian patients in the normal breast reduction mammoplasty group. For Caucasians, there was no significant difference between HCMV prevalence (95.7% in cancer tissues compared to 88.9% in normal tissues). For African-Americans, HCMV prevalence was 100% in cancer tissues, while it was 50% in the normal tissues. However, this difference was not significant because of small numbers.

We performed in situ hybridization on breast cancer specimens and their paired control breast specimens, as well as normal controls from individuals with no history of breast cancer, to determine the presence of HCMV nucleic acids. We detected HCMV nucleic acids specifically in neoplastic or nonneoplastic epithelial cells in 16/18 randomly selected patients from the breast cancer pool. In 3/3 cases that we tested both neoplastic and control epithelium from the same patient, HCMV nucleic acids were detected in both specimens. We also performed in situ hybridization on 18 randomly selected "normal control" breast specimens, and 11/18 of these specimens confirmed specific HCMV nucleic acid detection.

The pattern of HCMV immunoreactivity was similar to the pattern of ISH staining observed. We detected HCMV nucleic acid hybridization in normal breast epithelium and in neoplastic epithelium in areas of DCIS and IDC, but HCMV nucleic acids were principally undetectable in stromal cells (Figure 3, a-d). The pattern of HCMV nucleic acid detection within the cell was distinctly different from that of the Alu positive control probe, which is entirely specific to cellular nuclear DNA Alu repeats (Figure 3, e). In HCMV positive cells, we detected HCMV nucleic acids in the nucleus but also predominantly in the cell cytoplasm (as shown in Figure 3, b-d, and see Additional file 2). A positive control DNA probe (specific to DNA Alu repeat sequences) was specific for nuclear DNA in both epithelial and stromal cells in DCIS/IDC in tumors and normal breast (Figure 3, e), while a non-specific DNA probe was non-reactive in tumor and normal tissues (Figure 3, f). The specificity of HCMV nucleic acid hybridization was confirmed by detection of HCMV infected cells in HCMV infected lung tissues from an AIDS patient (Figure 3, g), while a negative control probe was negative in these same tissues (Figure 3, h). In normal breast epithelium from reduction mammoplasty patients, the amount of signal from HCMV nucleic acid hybridization was, in general, dramatically less than that found in breast tumor cells (e.g., Figure 3, i-j)

With respect to the internal consistency of immunohistochemical staining for HCMV antigens and detection of HCMV nucleic acids by in situ hybridization, there was a high correlation between positive for IHC and ISH specimens among the breast cancer and paired controls from breast cancer patients. 16/18 cases analyzed were positive for both HCMV antigen and nucleic acids, while in 2 cases there was a discordance (see Additional file 3).

In the normal breast tissue from non-cancer controls, there was a high degree of inconsistency between IHC and ISH results. Some of the positive IHC cases were negative for ISH, and vice versa (see Additional file 3). We attribute these results to the overall extremely low levels of antigen and nucleic acids detected in these normal tissues compared to the cancer cases. At such low levels of antigen and nucleic acid detection, we suspect that limitations of detection may have resulted in lack of internally consistent results between the two groups. In addition, it is possible that latent HCMV infection may occur in normal breast epithelium, in which case nucleic acids may be detected in the absence of protein expression.

We performed nested PCR for HCMV UL55 gene using DNA that was extracted from paraffin sections of 8 tumors and 4 control cases. 6/8 tumor cases and 1/4 normal control cases demonstrated amplified HCMV UL55 gene, which was confirmed by direct sequencing of the PCR products (not shown). Since the tumor specimens that were tested were all positive for HCMV by immunohistochemistry, we hypothesize that the negative PCR specimens may have had viral nucleic acids below the level of detection for our assay, or viral genetic material may have been more degraded than cellular housekeeping genes used as controls, since the DNA quality in these assays is very variable. Alternatively, the paraffin specimens used for these assays, which were not necessarily sequential sections to the ones used for immunohistochemistry and in situ hybridization, may have had less viral genome present. The one case of normal breast control tissue that was positive for HCMV UL55 by PCR was also positive for HCMV by immunohistochemistry and in situ hybridization, while the remaining three cases were negative by all three detection methods.