| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Original Article |
1 Department of Surgery, University of Vermont College of Medicine, Given Building, 89 Beaumont Avenue, Burlington, Vermont 05405
2 Charleston Area Medical Center Institute, 3110 MacCorkle Avenue SE, Charleston, West Virginia 25304
3 Department of Surgical Pathology, Fletcher Allen Health Care, 111 Colchester Avenue, Burlington, Vermont 05405
4 Department of Orthopaedics and Rehabilitation, University of Vermont College of Medicine, 430A Stafford Hall, 95 Carrigan Drive, Burlington, Vermont 05405
Correspondence: Address correspondence and reprint requests to: David N. Krag, MD, FACS; E-mail: david.krag{at}uvm.edu.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Methods: We compared ICC and IF side by side for detection of cancer cells and false-positive staining events on bone marrow aspirates from breast cancer patients, bone marrow from healthy donors, and healthy donor blood spiked with cancer cells. The primary target for isolated tumor cell detection was CK for both methods. IF used an additional set of antibodies to label hematopoietic cells (HCs).
Results: The detection rate of CK+ events in breast cancer patient bone marrow aspirates was 18 (58%) of 31 for ICC and 21 (68%) of 31 for IF. However, with IF, 17 of 21 CK+ cases were stained with HC markers and thus were identified as false-positive events. A surprisingly high CK+ event rate was observed in healthy donor blood and marrow. In all healthy donor samples, CK+ events were readily identified as HCs by IF. Detection sensitivity of spiked cancer cells in donor blood was similar for both methods.
Conclusions: There is a high frequency of CK+ events in blood and marrow, and it is important to note that this is observed both in patients with and those without cancer. IF with multiple HC markers allows straightforward discrimination between CK+ cells of hematopoietic and nonhematopoietic origin.
Key Words: Immunocytochemistry • Bone marrow • Cytokeratin • Immunofluorescence
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Detection of disseminated tumor cells in bone marrow is an area of research designed to identify metastatic breast cancer at an earlier and potentially treatable stage. The presence of isolated tumor cells in bone marrow aspirates, as demonstrated by immunocytochemistry (ICC), ranges from 11% to 43% and has been reported to be significantly associated with survival.2–15 Other researchers have shown an absence of isolated tumor cells or little correlation with overall patient survival.16–20 This variation of detection rates and the subsequent ambiguity of clinical significance may be in part due to differences in the detection techniques and patient populations. Rare cell detection requires processing the specimen to reduce background cells, labeling the target cell with antibodies, and interpreting cell morphology. These methods vary in different studies. Criteria for positive scoring of candidate-labeled cells are not standardized and frequently not well defined.16,21 In addition, a significant problem seems to be that nontumor cells can be labeled with antitumor antibodies. For example, hematopoietic cells (HCs) have been shown to stain positive with certain epithelial cell antibodies, such as MUC-1.16,22 In addition, clinical control samples from healthy donors can have cells that are stained positive with cytokeratin (CK) antibodies. It is important to note that cells from cancer patient samples can even stain positive with alkaline phosphatase alone.23
These issues are important, particularly because clinical applications of staging with this technology may be based on only one or a few detectable target cells. To address these variables, we used an expanded profile of antibodies with the goal of increasing our ability to specify the cell type of candidate-labeled cells. This method should maintain current levels of sensitivity for detection of disseminated tumor cells but, importantly, allow improved discrimination of labeled cells by reducing false-positive events. In this article, we report our experience using immunofluorescence (IF) with an expanded profile of antibodies side by side with ICC in samples from breast cancer patients and healthy donors.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Samples from the primary breast cancer patients were obtained by using touch preparation of the exposed face of the freshly excised tumor or with fine-needle aspiration passed directly into the solid tumor nodule. Slides were air-dried and stored at –80°C.
Negative control marrow and blood specimens were obtained from healthy donor volunteers. Sternal bone marrow was collected from patients undergoing cardiac surgery. Marrow from other locations was obtained from patients undergoing elective orthopedic operations. Peripheral blood was collected in Vacutainer tubes containing EDTA as an anticoagulant. The first tube of 3 to 5 mL was discarded to diminish the possibility of skin contaminants from the specimen.
Positive control clinical samples were obtained from breast cancer patients with documented stage IV metastatic breast cancer. As with the healthy controls, the first tube of blood was discarded to avoid contaminating squamous cells.
Specimen Processing
Bone marrow or peripheral blood samples were diluted 1x with phosphate-buffered saline and layered on a Ficoll-Paque Plus gradient (Amersham Biosciences, Uppsala, Sweden). Samples were centrifuged for 30 minutes at 400 x g. The mononuclear cell layer was carefully harvested and washed twice with phosphate-buffered saline. The cell count of the mononuclear cell layer was determined by using the Unopette Microcollection System (Becton Dickinson). Cells at a concentration of 1.5 million cells per slide were cytocentrifuged onto FisherPlus (Fisher Scientific, Pittsburgh, PA) glass slides at 1500 rpm for 5 minutes by using a Hettich Cytocentrifuge (Tuttlingen, Germany). Slides were air-dried and then frozen at –80°C until staining.
Staining Techniques
Bright-Field ICC
Slides were thawed for 15 minutes at room temperature before staining. All staining steps were conducted at room temperature and included 5-minute washes of Tris-buffered saline between steps. Cells were fixed in acetone (or, alternatively, 10% formalin and Triton X-100) for 10 minutes, and then two blocking steps were used: the first was Protein Block Serum-Free, and the second block was Dual Endogenous Enzyme Block (DakoCytomation, Carpinteria, CA), each for 15 minutes. When blocking was complete, slides were then incubated with a cocktail of mouse primary CK antibodies that included CAM5.2 (Becton Dickinson) and AE1 (Zymed Laboratories, South San Francisco, CA) for 30 minutes. Control slides were incubated with an irrelevant mouse immunoglobulin G antibody (1/60) to simulate immunoglobulin G concentration of the primary antibody (DakoCytomation). All subsequent reagents used were part of the DakoCytomation alkaline phosphatase anti–alkaline phosphatase (APAAP) kit. Slides were labeled with a link rabbit anti-mouse antibody and then incubated with the APAAP immune complex. Fast Red chromagen was used to visualize the enzyme substrate reaction. Counterstaining was performed with 4',6-diamidino-2-phenylindole (DAPI; Molecular Probes, Eugene, OR). Coverslips were applied after treatment with DakoCytomation Glycergel Mounting Media.
Fluorescence ICC
Slides were thawed for 15 minutes at room temperature before staining. All staining steps were conducted at room temperature and included 5-minute washes of phosphate-buffered saline between steps listed below. Cells were fixed in 100% methanol (or, alternatively, 10% formalin and Triton X-100) for 10 minutes and then blocked with Protein Block Serum-Free. All incubations were 30 minutes long. A cocktail of primary antibodies (1/100) including CD45 (Neomarkers, Fremont, CA), CD38, and CD34 (DakoCytomation) were incubated to label a variety of white blood cells, including plasma cells and stem cells. Secondary red fluorescent goat anti-mouse Alexa Fluor 568 (Molecular Probes) was then applied at a 1/400 dilution. Slides were then incubated with an irrelevant mouse antibody (DakoCytomation) to prevent subsequent cross-labeling. Antibodies to CKs 8/18 (Neomarkers) were custom-conjugated to green fluorescent Alexa Fluor 488 at 1.3 mg/mL used at 1/400 (Molecular Probes) and applied to the cells. Nuclear staining was performed with DAPI (Molecular Probes) for 10 minutes. Coverslips were applied after treatment with DakoCytomation Glycergel Mounting Media.
Cell-Detection Sensitivity
The sensitivity of each detection method—ICC and IF—was evaluated by adding known numbers (n = 10–100) of BT474 breast cancer cells to healthy donor specimens. Samples were processed, and the number of labeled cells was determined.
Slide Evaluation
Bright-Field ICC
Four slides (for an approximate total of 6 million cells) were evaluated for the presence of CK+ cells. Four slides were also evaluated for nonspecific staining with an irrelevant antibody. A cell was scored as positive if it exhibited red staining, the pattern of staining was consistent with cytoplasmic CK location, and it contained a visible nucleus by phase microscopy or UV light. Anucleated squamous cells and debris were excluded.
Bone marrow specimens were also scored according to published criteria by Naume et al.8 Briefly, cells were classified into four categories based on morphological criteria: (1) tumor cell, (2) uninterpretable cell, (3) probable HC, and (4) HC.
Fluorescence ICC
Four slides were screened for the presence of CK+ green fluorescent staining events. Digital images of all positive-staining events were obtained. Staining intensity was scored as the mean pixel value for each stained cell. A positive green fluorescent staining event was defined as staining with twice the intensity of background cells. Candidate green-stained cells were then evaluated for staining with hematopoietic-specific antibodies in the red fluorescent spectrum, for blue fluorescent DAPI staining of nuclei, and for morphology in bright-field phase contrast. A cell was interpreted as a tumor cell if it was positive for CK, was negative for the HC markers, contained a DAPI-stained nucleus, and had morphology consistent with that of a cell.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
). We also demonstrated similar sensitivity between the two techniques by staining slides with BT474 cells only, without the possible confounding presence of HCs. Nearly 100% of intact BT474 cells were stained with CK antibodies by both ICC and IF.
|
Blood and Bone Marrow Samples From Healthy Donors
The prevalence of CK+ cells was evaluated with both IF and ICC in 10 bone marrow and blood samples from healthy donors. Table 1 shows that the rate of CK+ staining events with both ICC and IF methods was relatively high in the 10 healthy marrow and blood control specimens. The number of CK+ cells ranged from 1 to 7 cells per sample.
|
is to present the surprisingly high prevalence of false-positive CK+ staining events. All of the CK+ events also stained positive with HC markers with the IF method.
Blood Samples From Stage IV Breast Cancer Patients
Blood from breast cancer patients with metastatic disease provided positive clinical controls for IF CK staining events. With the IF method, 11 of 19 cases had CK+ cells. However, in five of these cases, the CK+ cells were also positive for HC markers. Thus, 6 of 19 cases served as positive controls that had CK+ and HC–cells.
Breast Cancer Patient Bone Marrow Samples
Thirty-one breast cancer patients had marrow samples evaluated for the presence of CK+ cells by using both IF and ICC. Table 2 shows that the rate of CK+ staining events was relatively high and was similar for both ICC (58%) and IF (68%). The ICC method does not provide counterstaining of HCs, so determination of false-positive staining was more difficult. Inferential interpretation was aided by matched-slide staining with an irrelevant antibody.8,9 The rate of positively stained cells with an irrelevant antibody was high (48%). This casts doubt on the interpretation of CK+ events as cancer cells with the ICC method. CK+ events were only 10%higher than the rate of positive staining with irrelevant antibody (58%vs. 48%).
|
In the set of breast cancer marrow cases evaluated by ICC, we also interpreted the status of CK+ cells according to previously published staining and morphology criteria8 (Table 3). Six of the 18 CK+ cases were scored as tumor cells, 5 were scored as uninterpretable cells, and 7 were scored as probable or definite HCs.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Our own experience and that of others7,8,16,21–23,25 demonstrates that false-positive events with ICC analysis of marrow aspirates are not rare. The vast majority of cells in a marrow aspirate are hematopoietic and are therefore the most likely candidates for being falsely stained with anti-CK antibodies. Detection of disseminated tumor cells in the marrow is an exercise in rare event detection. The luxury of abundant target cells is absent, and interpretation must be made on a very limited population of target cells. Few pathologic studies are ultimately based on the interpretation of a patient sample as small as one or two cells. This sets the bar high for correct interpretation of marrow aspirates of breast cancer patients. False interpretation of a single cell could lead to alteration of staging and treatment planning. This limitation alone may keep this technology from mainstream application.
Our goal in this study was to document conditions in which false-positive CK staining events were observed and to apply conventional cell-staining methods to better understand the types of cells that are stained positive with CK antibodies. We chose an IF approach because fluorescence microscopy is highly sensitive, and different wavelengths of light can be readily evaluated. By using different IF reagents with matched excitation and emission filter sets, a variety of different antigens could be labeled and readily imaged. Viewing samples with narrow bandpass filters allowed the reagent of choice to be visualized with a nearly black background. This was particularly important for ease of detecting rare events, because a stained cell in a nearly black background stands out to the examiner’s eye. In addition, this method lends itself to digitization of the image and quantification of intensity levels. This moves the procedure away from the usual subjectivity of interpretation to at least a somewhat more quantitative approach.
CK antibodies were used for detection of CK+ cells. Both the ICC and IF methods demonstrated equal sensitivity for detecting cancer cells in controlled spiking experiments. This was expected, because both methods target CK. We chose a set of antibodies to broadly label HCs. We initially stained HCs with anti-CD45 alone. This left many cells unstained (data not shown), so the set was expanded to include anti-CD38 for staining plasma cells and anti-CD34 for staining stem cells. It would be attractive to assign each antibody to different excitation/emission spectra to obtain a better understanding of the nature of false-positive events. However, for simplicity of matching antibodies with staining reagents, the entire set of antibodies to HCs was assigned one color spectrum.
Positive staining events were observed in a surprisingly high proportion of blood and marrow specimens from healthy donors (Table 1). Staining events were observed both with anti-CK antibodies and irrelevant antibodies; this indicates that such binding is likely to be nonspecific. It is important to note that positive staining events were also seen with irrelevant antibodies in samples of marrow from breast cancer patients. The rate of positive staining with irrelevant antibodies was similar in healthy donors and breast cancer patients. By definition, these are false-positive events. Our observed false-positive staining rate is higher than those in articles from other investigators.7,8,22–24 For example, Braun et al.6 reported an incidence of 1% CK+ staining events in healthy control samples, and Wiedswang et al.7 reported 4%. The reason for this discordance is unclear. One possibility is that we examined more marrow cells than other investigators (6 x 106 vs. 2 x 106 mononuclear cells). In breast cancer patient marrow samples, the positive staining event rate for CK+ events was only slightly higher than the rate of staining with an irrelevant antibody. This is of concern and suggests that a significant proportion of CK+ staining events are nonspecific.
The ICC method did not allow discrimination of CK+ events on the basis of staining criteria. However, with IF, all CK+ events in bone marrow from breast cancer patients, bone marrow from healthy donors, and blood from stage IV breast cancer patients and healthy donors could readily be cross-checked as possible HCs on the basis of positive HC marker staining. IF was particularly helpful in identifying false-positive events in marrow aspirates from breast cancer patients, in which most CK+ events were HCs. Counterstaining with a cocktail of antibodies to label a range of HC markers (including CD45, 
, 
, glycophorin A, and CD68) has been used previously with some success for identification of false-positive events.23 Borgen et al.23 used an APA-AP technique for primary tumor labeling, and HC labeling was performed with a gold secondary antibody with a silver enhancement kit. This required removing the coverslip, monitoring silver stain development under the microscope, and relocalizing the target cell. A significant portion of cells labeled with the primary antibody were observed to label positively with at least one of the HC markers. Only a minority of the positively stained HCs were CD45+. This is particularly important given the recent clinical study of tumor cells in the blood of breast cancer patients, in which the definition of circulating tumor cells was "nucleated cells lacking CD45 and expressing cytokeratin."26
As has been noted by Lagrange et al.,16 most positive staining events could be interpreted as false positive on the basis of careful morphological evaluation. Other investigators have noted that discrimination of false-positive events on the basis of morphology can be very difficult.23 Use of conventional recommended scoring, based on morphology of our CK+ cases, resulted in identification of several such events as HCs. However, considerable uncertainty remained. Scoring based on multiple markers for both cancer cells and HCs may alleviate some of the subjectivity and difficulty associated with relying solely on morphology. In addition, flow cytometry and newly introduced automated imaging devices are better suited to detection of markers than to interpretation of morphology. Morphology is an important feature of interpreting rare cells and is complementary to appropriate staining of antigens.
The IF method performed well with bright staining in three measures of positive staining. This included cancer cells spiked into healthy blood specimens, stage IV breast cancer patients, and cancer cells obtained from the primary breast cancer. In the samples from the primary tumor, we did not see any cancer cells stained with the HC markers. This indicates that staining of cancer cells with our set of HC markers would be very unlikely.
It is unclear why rare HCs in the marrow or blood are positively labeled with either CK antibodies or irrelevant antibodies. Possibilities include cross-reactive binding to cells that express different but similar antigen binding sites, unusual expression of CK in cells that still have hematopoietic markers (for example, undefined stem cells), cells that have binding sites for the Fc portion of the antibody, and enzymatic reactions that result in positive coloration of a cell. The last possibility has been described and seems to be the result of an enzymatic reaction of the reagents with unblocked endogenous alkaline phosphatase.23 This would not apply to the false-positive cells by our IF method because this method is not based on an enzymatic reaction with alkaline phosphatase and the problem of endogenous enzymes cross-reacting is eliminated.
Given the wide variety of cells that populate the marrow, it is reasonable to expect that rare HCs may bind to species-specific antibodies on the Fc or another non–complementarity-determining region of an antibody. We have observed individual HCs bind goat anti-mouse antibody (control slides in which only secondary antibody is applied), but not rabbit anti-mouse antibody. Other cells demonstrated the opposite staining pattern.
Detection of disseminated tumor cells in the marrow or blood is a promising tool for improving the ability to establish prognosis in individual patients. Several clinical studies support this concept, but there is considerable variation in the observed results and the methods used. We have attempted to validate the methods used in reported clinical studies and have observed a surprisingly high rate of false-positive events. Our results indicate that using only a tumor-targeting antibody is not specific enough. We conclude that a set of counterstaining antibodies for HCs will allow identification of a substantial portion of CK+ staining events as HCs.
|
ACKNOWLEDGMENTS |
|---|
Received for publication December 7, 2004. Accepted for publication May 11, 2005.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
 |
|
 |
|
  I. Langer, U. Guller, O. R. Koechli, G. Berclaz, G. Singer, G. Schaer, M. K. Fehr, T. Hess, D. Oertli, L. Bronz, et al. Association of the Presence of Bone Marrow Micrometastases with the Sentinel Lymph Node Status in 410 Early Stage Breast Cancer Patients: Results of the Swiss Multicenter Study Ann. Surg. Oncol., June 1, 2007; 14(6): 1896 - 1903. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
