Breast cancer is unpredictable in its pattern and time of distant relapse. While in some patients the disease recurs soon after removal of the primary tumor, patients continue to relapse over prolonged periods. This feature, together with the current paucity of markers, both biochemical and immunocytochemical, poses significant problems for the physician in determining the optimal form and duration of adjuvant therapy. Currently, conventional preoperative staging investigations including chest radiography, liver ultrasound and skeletal scintigraphy identify occult metastases in approximately 1% of patients1 and have been abandoned in many centers unless clinically indicated.
Greater prognostic information is gained from clinicopathologic staging. Tumor size, grade, the presence of lymphovascular invasion, axillary lymph node involvement and steroid receptor status are important prognostic indicators.2 Despite this, many patients with favorable prognostic features will relapse within 5 years and many patients with poor prognostic features survive for more than 10 years. Studies from our unit and others have shown that many patients with primary breast cancer have disseminated cancer cells.3, 4, 5, 6
The presence of disseminated cancer cells as detected by polymerase chain reaction (PCR) and immunocytochemistry (ICC) in stage I and II breast cancer patients correlates with other prognostic indicators for early relapse such as tumor size, grade, lymphovascular invasion, local lymph node infiltration and with overall survival.4, 5, 6, 7, 8, 9, 10 These different studies using either PCR or ICC to detect a number of different antigens at the time of surgery4, 5, 6, 10, 11 showed a positive detection rate ranging from 25% to 43% depending on the marker used and the number of sites aspirated.
It has been suggested that the monitoring of minimal residual disease (MRD), as defined by the presence of disseminated cancer cells, could be used to improve disease staging, as a marker for evaluating new therapeutic strategies and assessing treatment response in individual patients. Previous studies have shown a very low incidence of MRD in the peripheral blood of primary breast cancer patients with early-stage disease.12, 13, 14
There is considerable variation in the incidence of MRD in both primary and metastatic breast cancer patients. The bone marrow is positive for disseminated cancer cells in 25–50% of patients using PCR14 and 20–40% using ICC.4, 5, 7, 8, 9, 10 Although our earlier study showed a lower incidence of micrometastases immediately after surgery,15 another group obtained a further sample after chemotherapy and showed that there was still evidence of MRD.16
Cytokeratin-19 (CK19) has been used as a marker for disseminated epithelial cancer cells in a number of studies in different solid tumors.12, 13, 14, 17, 18, 19 Other markers such as mammaglobin20, 21, 22, 23, 24, 25, 26 and maspin27, 28, 29, 30, 31 have recently been put forward as breast-specific markers for disseminated malignant cells. However, to date, none has been fully evaluated clinically.
We have developed a quantitative reverse transcriptase-PCR (RT-PCR) method for CK19 measurement (QPCR).32, 33 In this, the number of CK19 transcripts is expressed as a ratio as compared to the number of transcripts from the Abelson oncogene (ABL),34, 35 enabling us to distinguish benign from neoplastic samples. Our previous study used both QPCR and ICC on 145 peripheral blood samples from 22 patients with known metastatic breast cancer.33 There was a significant correlation (p < 0.0001) between the numbers of cytokeratin-positive cells by ICC with CK19 QPCR. There was also a correlation between the 2 assays and response indicating that circulating tumor cell levels reflect changes in disease load. This observation suggested that it might be possible to use QPCR and ICC to monitor residual disease at follow-up and to evaluate any changes with systemic treatment in patients following treatment of primary breast cancer.
Material and methods
Consecutive patients diagnosed with early-stage breast cancer were invited to take part in the study. All women were attending Charing Cross Hospital, Queen Mary's Hospital, or West Middlesex Hospital (London, U.K.) between September 1997 and March 2002. All patients had cytologically confirmed primary breast cancer and no evidence of distant metastases on chest radiology, bone scanning, or liver ultrasound (one patient with an abnormal bone scan has been excluded from all analyses). The study was approved by each local ethics committee and was conducted in accordance with the Declaration of Helsinki and all patients gave written informed consent. Samples were blinded for analysis and patients understood that the results would not be made available to them.
The initial bone marrow sample was taken together with peripheral blood at the time of surgery under general anesthetic from both the right and the left posterior iliac crest. Follow-up samples were taken at 3, 6, 12, 24, 36 and 48 months after surgery under local anesthetic and taken from one side only. The iliac crest for sampling at the first follow-up was determined by the QPCR results of the initial sample and the side with the highest CK19:ABL ratio32, 33 was used. At the subsequent follow-up, the QPCR result of the previous sample determined the side to be sampled. If the initial sample was negative on both sides, the follow-up sample was obtained from either the left or the right iliac crest. If, however, the sample became positive, then subsequent samples were obtained from the same side.
The majority of patients underwent wide local excision (81%) and had clinical T136 disease (70%). Pathologic nodal status was assessed in 76% of patients, of whom 58% were N0 (Table I). All patients received adjuvant systemic therapy. Of the 31 (24%) patients who received chemotherapy, 21 received FEC:5-fluorouracil 600 mg/m,2 epirubicin 50 mg/m2 (or 75 mg/m2) and cyclophosphamide 600 mg/m2 3–4 weekly for 6 months. Four patients received CMF:cyclophosphamide 600 mg/m,2 methotrexate 40 mg/m2 and 5-fluorouracil 600 mg/m2 on days 1 and 8 of a 4 weekly cycle for 6 months; 6 patients received epirubicin as 50 mg/m2 on days 1 and 8 of a 4 weekly cycle for 6 months. Eighty-six patients (66%) with estrogen receptor-positive tumors received hormonal therapy (tamoxifen 20 mg/day for all except one patient who received prostap 3.75 mg/month) alone. Sixteen of the patients who received tamoxifen received it in combination with chemotherapy, 15 received chemotherapy alone and 14 received no therapy.
Table I. Clinical Baseline Characteristics of the 131 Patients Recruited
Wide local excision
Preparation of bone marrow samples
The skin was incised before the aspirates were taken to minimize the risk of epithelial contamination. Between 2 and 5 ml of bone marrow was aspirated from each side using disposable 15 G (1.8 mm) marrow gauge bone marrow needles (Rocket Medical, Watford, U.K.), into syringes primed with preservative-free heparin (Leo Labs, Risborough, U.K.).
Blood and bone marrow samples were prepared as previously described.32, 33 Briefly, the mononucleocytes were separated from the blood over Ficoll (Amersham, Buckinghamshire, U.K.). The interface cells were removed, washed and the red cells lysed followed by a repeated wash. The mononuclear cells were counted and aliquoted for QPCR and ICC on the basis of ideally 6 × 106 cells for each methodology but with a minimum of 3 × 106 cells for each. Those undertaking the QPCR and ICC were blind to the clinical status and the identity of the patients and their previous assay results.
Cells were cytocentrifuged at a concentration of 5 × 105 per area, air-dried and stained as previously described.32, 33, 37 A total of 6 areas were stained; 4 were stained for the presence of cytokeratin positive cells and 2 were negative controls. The primary antibody (A45-B/B3; Micromet, Munich, Germany) directed to a common epitope of cytokeratin38 was used at 2 mg/ml. The rabbit antimouse antiserum (Z259; Dako, Hamburg, Germany) and the alkaline phosphatase antialkaline phosphatase (APAAP) complex (D651; Dako) were used as recommended by the manufacturer and the reaction developed with new fuschin. An isotype IgG1 mouse myeloma antibody MOPC-21 (Sigma, St. Louis, MO) served as negative control and the MCF-7 cell line as a positive control.32, 33, 37 The cytospins were counterstained with hematoxylin and screened using the Automated Cellular Imaging System (ACIS; Chromavision, San Juan Capistrano, CA). Samples that were isotype-positive were deemed uninterpretable and therefore excluded from the results. During this study, we have screened a large number of bone marrow aspirates from subjects without breast cancer and found them to be all negative.
Synthesis of cDNA was performed as described previously.32, 33 Samples were tested initially for CK19 by nested PCR followed by quantitation for all positive samples as previously described.32, 33
All pre-PCR manipulations were performed in a PCR-only designated laminar flow cabinet and using plugged pipette tips. At least 2 negative controls were included per run and prepared last. The reaction products were electrophoresed on a 1.8% agarose gel in a separate room using dedicated pipettes. A band of 463 bp was visualized for CK19-positive samples. Ethidium bromide-stained gels were used, as Southern blotting would only have served to increase the sensitivity of both methodologies thus resulting in the same ratio.
Competitive QPCR was performed as previously described.32, 33, 34 Briefly, a titration series of independent reactions containing 2.5 μl of cDNA plus 2.5 μl of known competitor dilution were added to 20 μl first-step mix. The competitor PCR product was seen at 588 bp and equivalence points were estimated by inspection, which has a low variation (17%) with an experienced operative.34 This is comparable to the variation seen in assays using TaqMan technology where intraassay variation ranges from 11% to 24%.39, 40, 41
Quantification of ABL transcripts as an internal control for the amount and quality of cDNA32, 33, 34 was performed for all samples by a single-step QPCR. Bands were visualized at 385 bp for ABL and 486 bp for the competitor. QPCR results were expressed as the ratio of CK19:ABL. A CK19:ABL ratio greater or equal to 0.1% (1 CK19/1,000 ABL transcripts) was regarded as positive. Conversely, samples with a ratio less than 0.1% were deemed negative. In our previous studies, we have screened 43 normal bone marrows32, 33 and never observed levels in excess of this.
The results presented here are preliminary and as such are primarily descriptive. Recruitment to the cohort has stopped but follow-up of patients continues. Chi-square-based tests were used to assess associations between baseline characteristics and MRD positivity at surgery. Positivity rates at each time point are reported and trends over time of the number of patients positive by QPCR, ICC, either technique and both are presented graphically for all data and the subgroup of patients with 12-month follow-up. Matched pairwise comparisons of CK19:ABL ratio values used the Wilcoxon signed-rank test.
Patients and baseline MRD results
One hundred thirty-one patients were recruited into the study with a mean age of 58 years (range, 31–84 years). Of these, 93 (71%) agreed to continue after surgery and 74 have been followed up after 3 months (Table II). Of these, 60 subjects have been followed up for 48 months but only 31 (35%) have given bone marrow samples at 6, 12, 24, 36 and 48 months. Four patients have relapsed; of these, 2 have subsequently died.
Table II. Summary of Results of MRD Detection of All Patients in the Study by ICC and QPCR
Bone marrow sample
Number of patients
Summary of results of detection of CK19-positive cells in bone marrow aspirates at surgery to 48 months postsurgery (QPCR positivity of > 0.1% CK19:ABL). The combined column indicates the results of both aspirates at the time of surgery by ICC and/or QPCR and of combining ICC and/or QPCR positivity at follow-up (see also Fig. 1).
As this is the first attempt to use QPCR in the detection/monitoring of MRD in primary breast cancer patients, QPCR and ICC (which has been widely used for detection of MRD) were compared in the 558 samples where both results were available (110 patients at surgery-left plus 109 at surgery-right plus all ICC at follow-up; Table II). Although in the majority of cases (71%), the 2 methodologies were in agreement, this was principally because of the concordance of dual negative findings. Where a sample was found to be positive, disagreement between the 2 tests was predominantly where a positive result was seen by QPCR but the ICC result was negative. This is illustrated in Table III, which details results of 51 patients with complete data to 12 months (i.e., no missing data points or isotype-positive ICC). There were no significant relationships between the baseline characteristics of the patients (Table I) and a positive finding by ICC or QPCR.
Table III. Results of MRD Detection in Bone Marrow Aspirates from All Patients up to 12 Months
Summary of results of detection of CK19-positive cells in bone marrow aspirates from (all patients with complete data from surgery to 12 months postsurgery (n = 51). QPCR data are the ratio of CK19:ABL expressed as a percentage.
At surgery, 68/131 (51%) patients had positive results using either or both methodologies (Table II, Fig. 1a). The number of positive patients as determined by ICC (31%) was consistent with previously published studies using ICC at surgery.4, 5, 6, 10, 11
We carried out a separate analysis for those samples that were positive using both techniques (i.e., double positive). Of all samples taken, 63/558 (11%) were double positive either at surgery or on follow-up. Only 21 of 109 (19%) patients were double positive at surgery (Fig. 1, both ICC and QPCR). However, these patients did not differ in terms of their clinicopathologic characteristics from the patients who had only one or other test positive at surgery. Only 31 of 93 subjects had a double positive aspirates during follow-up, but again, there was no significant difference between these and other patients in terms of conventional prognostic indexes. Of the 165/558 (30%) samples positive by QPCR, 63 (38%) were also positive by ICC. Of the 124/558 (22%) samples positive by ICC, 63 (51%) were also positive by QPCR. This demonstrates the higher sensitivity of the QPCR as compared to the ICC in terms of samples positive and the discrepancies may be due to the 2 techniques measuring different markers and due to sampling problems arising from analyzing a single bone marrow aspirate by both ICC and QPCR.
The number of positive cells detected by ICC in the aspirates ranged from 1 to 12 cells (except for one patient who had 2 results of 51 and 148 after surgery), but for the vast majority of samples that were positive, 77% had 1 to 2 detectable cells.
Follow-up of all study patients
In contrast to the initial sample at surgery when bilateral aspirates were taken, a single follow-up aspirate was taken during follow-up. At 3 months, 14/77 (18%) were positive by ICC and 26/88 (30%) by QPCR, approximately the same as at surgery if one considers either the left or right aspirate alone. At 6 months after surgery, 11/69 (16%) were positive by ICC and 13/74 (18%) by QPCR. The decrease from surgery to 6 months was significant by QPCR (p = 0.001) but not by ICC (p = 0.09).
At 12 months after surgery, there appeared to be an increase in the proportion of positive patients. This increase, 16% to 32% (ICC) and 18% to 46% (QPCR) from 6 to 12 months, was of statistical significance (p = 0.008 for ICC and p = 0.001 for QPCR). This was initially thought to be an indication of the emergence of resistance to adjuvant therapy; however, the trend was not maintained beyond 12 months. There was no relationship between the characteristics of the breast carcinoma (node status, ERα status, etc.) and whether patient's bone marrows were positive by 12 months with either ICC or QPCR.
At 24 months, the overall proportion of positive patients fell to the 3-month level. Thereafter, the percentage of patients positive for disseminated cells stabilized at approximately 40% for the next 2–3 years if the 2 methodologies are combined (Fig. 1a), indicating that MRD persisted in a substantial proportion of patients for up to (and at least) 48 months. Figure 1(b) shows CK19:ABL ratios for all patients at different time points to illustrate the median and range of values.
Effect of adjuvant therapy
The individual pattern of positivity within a patient was examined to determine whether individual patients showed either rises or falls in MRD during adjuvant therapy. Of the 110 patients with at least 1 ICC sample, 65 (59%) had a positive result at some time and 56/80 (70%) of patients with at least 2 follow-up ICC samples had a positive result. All patients received adjuvant therapy and, in cases where there was a positive result in either the pretreatment or 3-month aspirate and some follow-up data, 32/45 (71%) showed a fall in the CK19:ABL ratio [in 11/32 (34%), this decrease was a single log; in the remaining 21 patients, there was a larger decrease] and 15/30 (50%) showed a reduction in cytokeratin-positive cells detected using ICC during follow-up.
Of interest, in one patient, 977/2 × 106 stained cells were detected in the bone marrow sample taken at surgery and the ratio of CK19:ABL in this patient was similarly very high at 1:2 (50%). Despite this, she had no sign of overt metastatic disease on routine bone scanning, commenced treatment with tamoxifen and was monitored in the same way as the other patients. One year after starting adjuvant therapy, the ratio of CK19:ABL fell to 1.6% and the number of ICC-detectable cells fell to 146/2 × 106 stained cells. This patient was excluded from the statistical analysis because she clearly had extensive metastatic breast cancer confined to the bone marrow.
Results in 51 patients with complete data (12-month results)
We wished to examine the results in the 51 patients who had a complete set of 12-month bone marrow results. These are detailed in Table III and Figure 2. The pattern of results in these patients is similar to the entire cohort (Fig. 1). Table III shows the QPCR and ICC results in this cohort of 51 patients and Figure 3 shows the frequency of the number of positive samples in these 51 patients. Only one patient was positive throughout by QPCR and none by ICC. The majority of patients had 1 or 2 samples positive: 35 (69%) by QPCR and 31 (61%) by ICC. Nine (18%) patients were negative throughout by QPCR and 18 (35%) by ICC. Of the remaining 51 patients, 2 (4%) had 3 samples positive by ICC and 6 (12%) by QPCR and 1 (2%) patient had all samples positive by QPCR. Twenty-three (45%) of these patients were negative at surgery by both QPCR and ICC. All of these had a positive result in at least one follow-up aspirate, although of these, 11/23 (48%) had only a single positive result by QPCR (Table III). This may be due to sampling problems and the splitting of the aspirate for analysis by both ICC and QPCR.
At the outset of the study, an analysis of 20 ml of peripheral blood from each patient at each time point until 12 months after surgery was planned. However, only 28/246 samples (11%) were CK19-positive by QPCR and none of 208 by ICC. This low detection rate is in agreement with previously published studies12, 13, 14, 42 and it was thus decided to discontinue analysis of these peripheral blood samples.
Following surgery for primary breast cancer, approximately 70% of women with early breast cancer now receive tamoxifen or chemotherapy for a fixed time interval but no attempt is made to ascertain whether it is effective in individual patients. Only a proportion of patients benefit and therefore many experience treatment-related side effects for little benefit. If a monitoring system could be developed, alternative therapies could be given to nonresponding patients, with the expectation that more patients would be cured, as prolonged follow-up of trials of adjuvant therapies have demonstrated that a proportion of patients are cured by systemic therapy.43
Bone marrow samples from primary breast cancer patients were analyzed for presence of micrometastases. Bone marrow was aspirated at the time of excision of the primary tumor and at 3, 6, 12, 24, 36 and 48 months after surgery. There were major losses to follow-up between surgery and 3 months (some patients were unwilling to have aspirates taken so close to surgery) and from 3 to 6 months (some patients were unwilling to return after the first aspiration carried out under local anesthetic). Withdrawals after 6 months were less frequent and collection of follow-up samples continues. Annual bone marrow aspirates appear acceptable to the majority of patients.
Our results indicate that a variable pattern of MRD changes is observed after surgery and adjuvant treatment for breast cancer. However, a significant proportion of patients shows a decline in the number of cytokeratin transcripts following surgery and adjuvant therapy. We speculate that the significant decrease in the proportion of patients positive by QPCR at 6 months is probably due to the surgical removal of the primary tumor together with the effects of adjuvant therapy. There is evidence in animal studies that tumor removal may stimulate cell proliferation in macrometastatic foci44, 45 and that removal caused a switch of micrometastatic foci to an angiogenic phenotype, thus resulting in growth of metastases.46 No direct evidence of these data has been demonstrated in human breast cancer; however, one study does show some evidence that removal of the primary tumor may induce changes in the growth kinetics of metastatic foci.47 To investigate the rise in the proportion of patients positive at 12 months and the subsequent fall, we considered the relationship between CK19 positivity and adjuvant treatment (chemotherapy alone or in combination with hormonal therapy). Although there were only a limited number of chemotherapy-treated patients in this cohort (n = 9), no treatment-specific associations were observed (data not shown).
Some patients appear to show a rise in MRD despite therapy, but it is too early to judge the significance of this observation, since only 4 patients in this series have clinically confirmed relapses (2/4 were positive at surgery by QPCR and 1/4 by ICC).
Many groups have examined sequential changes in biochemical markers. We have previously measured 12 different markers from diagnosis to relapse, but only demonstrated a lead interval (as defined by the time between the detection of markers of micrometastases and the appearance of clinically or radiologically overt metastases) in 3 markers: carcinoembryonic antigen (CEA), serum alkaline phosphatase (SAP) and γ-glutamyl-transpeptidase (GGT); for these 3, the mean lead interval was approximately 3-month duration. This result was confirmed recently by Sutterlin et al.,48 who measured CEA and carbohydrate antigen 15-3 (CA 15-3) for their predictive value. The diagnostic accuracy was only 83% for CEA and 88% for CA15-3. CEA and CA15-3 had a positive predictive value of 27% and 47%, respectively. Molina et al.49 recorded lead times of 2–5 months with circulating c-erb B2, CEA and CA15-3. Other studies have focused on defining possible markers for monitoring metastatic breast cancer; these studies have included CA15-3 and the erythrocyte sedimentation rate (ESR),50 CA15-3, CEA and TPS.51 These studies show that the rate of fall of tumor markers relate to outcome. However, to date, considerable controversy surrounds the value of monitoring patients using these markers. It has been proposed that as disseminated cells are highly resistant to chemotharapy16 and that the presence of tumor cells in bone marrow after high-dose chemotherapy is not associated with a poor prognosis,52 merely monitoring the presence of disseminated tumor cells is insufficient to predict relapse. Studies on the biology of the disseminated tumor cells may improve the prediction of relapse, in particular the proliferative potential of the cells.53, 54 The proliferative potential was demonstrated to be a better indicator of decreased patient survival than ICC; however, neither of these studies were used for monitoring follow-up samples. Therefore, this methodology requires further investigation.
Real-time PCR is now becoming the standard methodology for quantifying gene transcripts and, using Lightcycler Technology (Roche Diagnostics, Mannheim, Germany), we have developed QPCR assays for both CK19 and ABL using the artificial competitor we made previously32 in order to construct an internal standard curve for absolute quantification of CK19 and ABL transcripts (data not shown). The nested QPCR had an intraassay variability of 6–13% when 2 samples (1 borderline positive, 1 highly positive) were assayed 6 times in the same run and the interassay variability (same sample assayed 6 times on consecutive runs) was 10.5%. When the 2 QPCR methodologies were compared in 49 patient samples, there was an extremely good correlation between the 2 (p = 0.0002). Real-time PCR shows increased sensitivity and reproducibility of the test for further studies. A recent study looking at CK19 transcripts in peripheral blood of breast cancer patients55 has employed a similar strategy. With the development of real-time PCR, this may become the standard methodology for monitoring breast cancer patients. A further possibility for optimization is the development of automated imaging systems. Several are now commercially available; however, only further research will determine whether this is indeed the case. We recommend using both methodologies in the detection of bone marrow micrometastases because of the uncertainty inherent in detecting such small numbers of cells by ICC alone in a large sample.
In conclusion, we have demonstrated that in a substantial proportion of patients, MRD persists using the techniques that we have developed, and that it is possible to monitor patients, preferably using both QPCR and ICC, after breast surgery using bone marrow aspirates.
The authors thank Professor Klaus Pantel (Universitätsklinikum Eppendorf, Hamburg, Germany) for his invaluable technical assistance and advice.