Fax: (011) 30-26510-99394
The neutrophil, not the tumor
Serum CA 15-3 elevation as a result of granulocyte—Colony-stimulating factor–induced neutrophil MUC1 overexpression and neutrophilia in patients with breast carcinoma receiving adjuvant chemotherapy
Version of Record online: 27 AUG 2004
Copyright © 2004 American Cancer Society
Volume 101, Issue 8, pages 1767–1775, 15 October 2004
How to Cite
Pentheroudakis, G., Malamou-Mitsi, V., Briasoulis, E., Damala, K., Vassou, A., Vartholomatos, G., Kolaitis, N. and Pavlidis, N. (2004), The neutrophil, not the tumor. Cancer, 101: 1767–1775. doi: 10.1002/cncr.20581
- Issue online: 1 OCT 2004
- Version of Record online: 27 AUG 2004
- Manuscript Revised: 12 JUL 2004
- Manuscript Accepted: 12 JUL 2004
- Manuscript Received: 8 JUN 2004
- CA 15-3;
- breast carcinoma;
- granulocyte—colony-stimulating factor (G-CSF)
Patients with resected breast carcinoma who received granulocyte—colony-stimulating factor (G-CSF)–supported adjuvant chemotherapy exhibited an increase in their serum CA 15-3 levels. The authors investigated the role of G-CSF–induced neutrophil MUC1 expression in this setting.
Twenty-two patients with resected early breast carcinoma and 6 patients with high-grade lymphoma received chemotherapy cycles with or without G-CSF support. When given, G-CSF was administered for either 5 or 10 days per cycle. Immunocytochemical staining and flow cytometric analysis of peripheral blood neutrophils and bone marrow myeloid cells for MUC1 epitopes were performed during treatment.
At baseline, the median serum CA 15-3 was 16 U/mL, with weak MUC1 expression in peripheral neutrophils (median immunocytochemical score [ICCS] = 40, flow cytometric score [FCS] = 211 antibody molecules per neutrophil). For patients receiving chemotherapy cycles with 5-day G-CSF support, median CA 15-3 levels increased moderately (median = 28 U/mL; P = 0.016) and absolute neutrophil counts (ANC) did not increase, whereas ICC staining showed a moderate increase (median ICCS = 105; P = 0.015). For patients receiving chemotherapy cycles with 10-day G-CSF, serum CA 15-3 levels increased 2–4-fold from baseline levels and reached abnormal levels (median = 47; P < 0.0005) and the ANC increased (median = 21,400/mm3; P = 0.007), whereas significant induction of MUC1 expression occurred in the cell membrane and mostly in the cytoplasm of neutrophils (median ICCS = 162; P = 0.001). Flow cytometry detected increased cytoplasmic, but not surface, neutrophil MUC1 expression in the 10-day G-CSF group only (baseline median FCS = 3975, 4th cycle median = 6327 molecules per cell; P = 0.028). In the bone marrow, induction of MUC1 expression was observed in the 10-day G-CSF group only in band forms and neutrophils, but not in more immature myeloid cells. Serum CA 15-3 levels and ICC score were found to demonstrate a linear relation. When ICCS and ANC were incorporated in a combined score, its relation with serum CA 15-3 levels demonstrated a parabolic (cubic) pattern. Serum CA 15-3 levels, ANC, and neutrophil MUC1 staining returned to baseline after the completion of therapy. No excess of malignant recurrences were observed.
Women with resected breast carcinoma who received G-CSF–primed chemotherapy showed serum CA 15-3 elevation due to an increase in peripheral blood neutrophil number and induced neutrophil cytoplasmic MUC1 expression, which was caused by G-CSF. Physicians should be aware of this interaction. Cancer 2004. © 2004 American Cancer Society.
Breast carcinoma is the malignancy diagnosed most frequently in women living in developed countries.1 With increasing public awareness and screening mammography practice, the majority of women can be treated with surgical resection of the primary and axillary lymph nodes. Adjuvant chemotherapy is widely applied postoperatively to prevent malignant disease recurrence, often in dose-dense schedules supported by granulocyte—colony- stimulating factor (G-CSF) administration.2 CA 15-3, an epitope of the MUC1 mucin, is a tumor marker secreted by malignant mammary epithelia and determination of its serum levels is one of the battery of biochemical tests obtained routinely in the follow-up of women with resected, early breast carcinoma.3 Although to our knowledge the cost-effectiveness of such a strategy has been neither shown nor disproved, it is widely used in Europe. A reversible increase in serum CA 15-3 levels in women undergoing G-CSF–supported adjuvant chemotherapy has been reported, often leading to a futile and expensive diagnostic workup for detection of tumor recurrence.4 The absence of malignant disease recurrences as well as close correlation of serum CA 15-3 levels with absolute neutrophil count (ANC) and with serum lactate dehydrogenase and alkaline phosphatase levels pointed to the polymorphonuclear neutrophil as a potential cause of elevated CA 15-3 levels. We sought to investigate the mechanisms of the increase in serum CA 15-3 levels in a cohort of women undergoing adjuvant chemotherapy after breast carcinoma resection.
MATERIALS AND METHODS
Women with AJCC TNM Stage II/III invasive ductal breast adenocarcinoma were eligible for inclusion in the current prospective study. Eligibility criteria included the following: previous surgical tumor ablation by modified radical mastectomy or wide local excision and level II or III axillary lymph node dissection within the previous 8 weeks from registration; the absence of locally recurrent or metastatic disease after staging with computed tomography scans of the chest and abdomen/pelvis and technetium 99m isotope bone scans; adequate bone marrow (ANC > 1.5 × 109/L, platelet count > 100 × 109/L), renal (serum creatinine < 1.5 × the upper limit of normal [ULN]), and hepatic function (serum levels of bilirubin, aspartate aminotransferase, alanine aminotransferase < 1.5 × the ULN); the absence of any severe comorbid or other neoplastic conditions; an Eastern CooperativeOncology Group (ECOG) performance status (PS) of 0–1; normal baseline CA 15-3 levels; and the need for adjuvant cytotoxic chemotherapy. According to their risk profile (tumor size, number of involved axillary lymph nodes), age, and ECOG PS, these patients were assigned to receive adjuvant treatment according to any 1 of 3 schedules: three times weekly combination chemotherapy without G-CSF support, dose-dense twice-weekly combination chemotherapy cycles with G-CSF support, or a combination of the 2 (three times weekly courses followed by twice-weekly G-CSF–boosted courses). These chemotherapy regimens reflected standard practice patterns in our center (Ioannina University Hospital, Ioannina, Greece). Patients receiving G-CSF support were further randomized to receive either 5 or 10 days of G-CSF per cycle. Accrual was structured in this way so as to provide comparable groups of women (i.e., women not receiving G-CSF, women receiving 5-day G-CSF, and women receiving 10-day G-CSF support per cycle), as well as a group of women who would be switching from no G-CSF to G-CSF support and, therefore, would be controls for themselves. The G-CSF received was filgrastim at a daily dose of 5 μg/kg of body weight subcutaneously.
Patients with Stage II–IV Hodgkin disease or with World Health Organization high-grade non-Hodgkin lymphomas receiving first-line cytotoxic combination chemotherapy were enrolled in the study as well. These patients formed an additional comparable patient group characterized by the absence of malignant epithelial cells. They substituted for healthy controls, overcoming the ethical problems that prohibit G-CSF administration in healthy subjects. Informed consent was obtained from all patients for inclusion in the study.
All patients were followed with physical examination; evaluation of full blood counts; serum biochemistry tests and CA 15-3 levels; and monitoring of drug use, infections, malignant disease recurrence, and other medical conditions on Day 1 of each chemotherapy cycle. Analysis of peripheral blood neutrophils for CA 15-3 and parent MUC1 peptide expression by means of immunocytochemistry and flow cytometry was performed at baseline, every 3 cycles of chemotherapy, and 3 months after the completion of treatment. Immunocytochemical analysis of CA 15-3 levels and MUC1 expression was performed in bone marrow myeloid cells at baseline and in the 4th cycle of chemotherapy in a group of patients receiving 10-day G-CSF support per cycle. These patients were simply the last six to be enrolled in the current study and bone marrow aspiration was performed solely for the purpose of the study with their consent. All study parameters also were analyzed in three healthy controls at baseline. The study was approved by the local ethics committee and written informed consent was obtained from all subjects involved with the study.
Serum CA 15-3 assessment was performed using the commercially available quantitative Abbot AxSYM System (Abbot Laboratories, Abbot Park, IL). The AxSYM CA 15-3 assay is an automated two-step sandwich assay that uses 115D8 as the capture antibody (reacting with MAM6 antigen of the human milk-fat-globule membrane) and DF3 (reacting with a membrane-enriched fraction of a metastatic human breast carcinoma) as the tracer antibody, based on microparticle enzyme immunoassay technology. The upper normal level of CA 15-3 with this assay is 31 U/mL, which applies to > 99% of the healthy female population. The median value in the same population is 15.5 U/mL.
Immunocytochemical staining based on the principle of indirect immunodetection was applied. Five milliliters of whole blood collected in ethylenediaminetetraacetic acid (EDTA)-containing tubes were processed with Ficoll-Hypaque (Pharmacia, Uppsala, Sweden) gradient centrifugation and red blood cell lysis (0.15 mM NH4Cl, 1 mM KHCO3, 0.1 mM EDTA [pH 7.3], and lysis buffer) before neutrophils were collected as sediment and stored in 30 mL of PreserveCyt® solution (phosphate-buffered saline [PBS] plus methanol) (Cytyc Corporation, Boxborough, MA). The suspension was passed through a ThinPrep® processing machine (ThinPrep 2000 method; Cytyc Corporation) and neutrophils were collected on a glass slide. The advantage of this technique is that the cells are evenly distributed in a defined circular area. Collected cells were incubated with a 96% alcolholic solution, followed by application of buffered casein solution (Power Block, BioGenex Laboratories, Inc., San Ramon, CA). The slides were then successively incubated with the anti-CA 15-3 primary antibody, biotinylated anti-immunoglobulin, streptavidin conjugated with alkaline phosphatase followed by naphthol phosphate/Fast Red substrate and Mayer's hematoxylin for nuclear staining. All slides were washed with PBS after each antibody application. All antibodies were used at an optimal working dilution of 1:100 buffer solution. Only cell-bound alkaline phosphatase activated the substrate to an intense red color staining of the cytoplasm and cell membrane. The slides were examined within 48 hours from venipuncture and neutrophils in 10 high-power fields (× 40) were graded as bearing no staining (0+), weak staining (1+), moderate staining (2+), or strong staining (3+). The percentages of cells with each staining pattern multiplied by the appropriate factor were summed to produce an immunocytochemical score (ICCS) ranging from 0 to 300.
Flow cytometric analysis was performed with the FACScan™ flow cytometer (Becton-Dickinson, Mountain View, CA) using CellQuest software (Becton Dickinson). The antibodies used were anti-CD45/CD14 double stained with fluorescein isothiocyanate (FITC) and phycoerythrin (PE) for the detection of leukocytes, IgG1/G2 double stained with FITC and PE for definition of nonspecific antibody-cell binding as negative control, anti-CA 15-3 antibody stained with FITC for study of MUC1/CA 15-3 expression on the cell surface of neutrophils, and anti-CD15 stained with FITC for definiton of positive control. After incubation in the dark of 100-μL blood aliquots with each antibody, red blood cell lysis with FACS lysing solution (Becton Dickinson) and centrifugation, each patient sample was gated for neutrophils and analyzed in the FACScan within 4 hours from venipuncture. When applied, permeabilization of neutrophil cellular membrane was achieved with incubation of neutrophils with 0.5 mL of FACS permeabilizing solution (Becton Dickinson) for 15 minutes. A regression analysis curve was constructed with the use of Quantibrite beads (Becton Dickinson) conjugated with defined numbers of PE molecules. Although our anti-MUC1 antibody-to-fluorochrome binding ratio was not 1:1, we used the same FITC–anti-MUC1 antibody lot for the analysis of all patient samples, thus introducing the same error in all measurements. This allowed the calculation of a quantitative flow cytometric score (FCS) with the use of QuantiCalc software (Verity Software House, Topshame, ME) as a quantitative parameter that may not describe the actual number of antibody molecules bound per cell, but allows direct, reliable comparisons to be made among FCS of our patient samples.
Two commercially available monoclonal antibodies (MoAb) were used for immunocytochemical analysis: mouse anti-MUC1 glycoprotein (IgG1, clone Ma695; NovoCastra Laboratories, Newcastle-upon-Tyne, UK) and mouse anti-MUC1 core (IgG1, clone HMPV; Becton Dickinson). The MoAb used for flow cytometry was FITC-conjugated mouse anti-MUC1 core (IgG1, clone HMPV; Becton Dickinson). Clone HMPV reacts with the APDTR amino acid sequence in each of the 20 amino acid tandem repeat (variable number of tandem repeats [VNTR]) units that comprise the extracellular domain of the MUC1 core peptide, whereas clone Ma695 reacts with the sialylated epitopes on the oligosaccharide side chains of the MUC1 core peptide. All antibodies used in the four aliquots of each patient sample (anti-CD45/14, IgG1/G2, anti-CD15) were commercially available mouse MoAbs conjugated with FITC, PE, or both (Becton Dickinson).
The Mann–Whitney U nonparametric test was used to compare different parameter means, whereas parameter means of paired samples were compared with the Wilcoxon signed rank test. Correlation was examined with the Pearson correlation coefficient and regression models were built with SPSS statistical software, Version 8 (SPSS Inc., Chicago, IL). Survival was calculated with the Kaplan–Meier product limit method and curves were compared with the log-rank Cox–Mantel test.
Thirty-one subjects with a median age of 50 years (range, 19–76 years) were included in the protocol. For three healthy controls, only determination of serum CA 15-3 levels, along with immunocytochemical and flow cytometric analysis of neutrophil MUC1 expression, was performed at baseline, without further follow-up. Twenty-two women with Stage II/III resected breast adenocarcinoma received adjuvant chemotherapy whereas 6 patients with Hodgkin disease or non-Hodgkin lymphoma received cytotoxic chemotherapy. Of all patients, 3 never received G-CSF support, 17 received G-CSF in each cycle of chemotherapy, and 8 patients with breast carcinoma initially received 4 courses of three times-weekly epirubicin/paclitaxel without G-CSF support followed by 3 courses of a twice-weekly cyclophosphamide/methotrexate/5-fluorouracil combination with G-CSF support. When used, G-CSF was given to each patient for the same number of days per cycle. Eleven patients received 5 days of G-CSF per cycle whereas 14 patients received 10 days of G-CSF per cycle. All patients completed their scheduled treatment plan uneventfully. At a median follow-up of 14 months from diagnosis, the 1-year disease recurrence or progression-free survival rate was > 90% with 4 malignant disease recurrences having been documented. Two disease recurrences occurred in the patient group that experienced no increase in serum CA 15-3 levels during treatment and two occurred in patients in whom the serum CA 15-3 levels increased during treatment. The recurring malignancy was Hodgkin disease in one patient, non-Hodgkin lymphoma in two patients, and breast carcinoma in one patient. Patient characteristics are summarized in Table 1.
|Median age (yrs) (range)||50 (19–76)|
|High-grade non-Hodgkin lymphoma||4|
|Chemotherapy and G-CSF status|
|Patients without G-CSF use throughout chemotherapy||3|
|Patients with G-CSF use throughout chemotherapy||17|
|Patients receiving chemotherapy cycles without G-CSF, followed by chemotherapy cycles with G-CSF||8|
|Duration of G-CSF use|
|Patients receiving 5 days of G-CSF per cycle||11|
|Patients receiving 10 days of G-CSF per cycle||14|
|No. of chemotherapy cycles administered||200|
|With G-CSF support||151|
|With 5-day G-CSF support||71|
|With 10-day G-CSF support||80|
|Without G-CSF support||49|
|Chemotherapy regimens administered in patients|
|FAC or AC||3|
|EC followed by T||2|
|E followed by T followed by CMF||8|
|ET followed by CMF||7|
|Median follow-up||14 mos|
|1-yr disease-free survival||91%|
Serum CA 15-3 Levels
Baseline serum CA 15-3 levels ranged from 12 to 24 U/mL in all patients. In the group of patients receiving chemotherapy cycles without G-CSF, no increase in serum CA 15-3 levels was noted during treatment or at follow-up. Patients receiving chemotherapy cycles with 5-day G-CSF support experienced a small increase in serum CA 15-3 levels during Cycles 4–7, although marker levels were well within the normal reference range. Conversely, patients receiving chemotherapy cycles with 10-day G-CSF support showed an increase in serum CA 15-3 levels beyond the normal range during Cycles 4–7 (Table 2). This increase was self-limited (≤ 2 × ULN), reversible (marker levels decreased to normal at follow-up), and occurred in both patients with breast carcinoma and those with lymphomas. The median ANC did not appear to increase during treatment in chemotherapy cycles without G-CSF or even with 5-day G-CSF support (baseline, ANC = 5100/mm3; Cycles 4–7, ANC = 4020/mm3, two-sided P = 0.271). As expected, a significant increase in the neutrophil count was observed in chemotherapy cycles with the administration of 10-day G-CSF (baseline, ANC = 5260/mm3; Cycles 4–7, ANC = 21,400/mm3, two-sided P = 0.007).
|Characteristics||No G-CSF group||5-day G-CSF group||10-day G-CSF group|
|Baseline||16 (12–23)||16 (12–22)||23 (17–24)|
|Cycles 4–7||22 (16–32)||28 (21–40)||47 (42–63)|
|Follow-up||17 (15–21)||17 (14–22)||19 (14–22)|
|Two-sided P value for baseline–peak difference||0.1||0.016||< 0.0005|
Immunocytochemical analysis of peripheral blood neutrophils for MUC1/CA 15-3 expression showed either no or weak staining at baseline (median ICCS = 40–60). This pattern did not change when chemotherapy cycles without G-CSF were administered. Chemotherapy with 5-day G-CSF support was associated with a moderate increase in neutrophil MUC1 staining (median ICCS = 105 at Cycles 4–7). When chemotherapy cycles with 10 days of G-CSF per cycle were administered, a significant intensification of the neutrophil MUC1 positive staining was observed, with the median ICCS reaching 160 during Cycles 4–7. Moderately or strongly positive MUC1 staining was diffuse in the cytoplasm and, to a lesser extent, in the cellular membrane of neutrophils (Fig. 1). MUC1 staining reverted to negative or weakly positive at follow-up (3 months after treatment completion) in all G-CSF treatment groups. Data regarding the ICCS by G-CSF status are shown in Table 3.
|Characteristics||No G-CSF group||5-day G-CSF group||10-day G-CSF group|
|Baseline||40 (26–74)||60 (44–80)||50 (35–70)|
|Cycles 4–7||60 (43–93)||105 (86–161)||162 (108–191)|
|Follow-up||60 (8–89)||40 (16–64)||60 (28–79)|
|Two-sided P value for baseline–peak difference||0.68||0.015||0.001|
Flow cytometry revealed weak MUC1 expression on the surface of neutrophils in comparison with the IgG1/G2 negative control in all patients at baseline (median FCS = 200–300 antibody molecules bound per cell [ABC]). No induction of MUC1 expression was noted on the surface of neutrophils during chemotherapy in groups of patients receiving treatment courses with or without G-CSF. Even with 10-day G-CSF support per cycle, the median FCS declined during Cycles 4–7 (169 ABC) and remained lower than baseline. FCS data are shown in Table 4.
|Characteristics||No G-CSF group||5-day G-CSF group||10-day G-CSF group|
|Cell surface FCS|
|Two-sided P value for baseline–peak difference||0.7||0.02||0.02|
|Two-sided P value for baseline–peak difference||0.028|
In view of the discordance of data provided by immunocytochemistry and flow cytometry and the evidence for cytoplasmic MUC1 staining in the former, neutrophils were processed with Becton Dickinson permeabilizing solution before flow cytometric analysis in 7 patients receiving chemotherapy with 10-day G-CSF support. This should allow the FITC-conjugated anti-MUC1 antibody to cross the cellular membrane and reach the cytoplasm. A significant induction of cytoplasmic MUC1 expression during 10-day G-CSF–boosted chemotherapy was observed in 5 of 7 patients. There was no change noted in one patient and cytoplasmic MUC1 expression declined in another. MUC1 expression returned to baseline levels after the completion of therapy. Quantitative data on median cytoplasmic FCS are shown in Table 4.
Combined Score Correlations
When serum CA 15-3 levels and ICCS from all patients were plotted together, there was evidence of linear correlation (Pearson correlation coefficient r = 0.776; two-sided P value < 0.0005), as shown in Figure 2. However, as the increase in serum CA 15-3 levels may be due not only to increased neutrophil MUC1 expression but to an increase in the total number of neutrophils as well, a new parameter was constructed. Multiplication of the ICCS by the ANC divided by 1000 yielded the combined score (CS), which incorporated both the intensity of MUC1 staining per neutrophil and the total number of peripheral blood neutrophils. The median CS increased significantly during treatment in patients receiving chemotherapy cycles with 10-day G-CSF support (baseline median CS = 285, cycle 4–7 median CS = 3930; two-sided P value = 0.001), but not in those receiving either no G-CSF (baseline median CS = 256, Cycle 4–7 median CS = 156; two-sided P value = 0.052) or only 5-day G-CSF support (baseline median CS = 362, Cycle 4–7 median CS = 561, two-sided P = 0.205). CS values returned to baseline levels as soon as G-CSF–boosted chemotherapy was completed. The change of the CS during treatment by G-CSF status is shown in Figure 3. When serum CA 15-3 levels and CS values in all patients were plotted together, there was evidence of a cubic association (multiple r = 0.779, r2 = 0.61). For example, initial CS increases produced significant elevation in serum CA 15-3 levels, but further CS increases (beyond a cutoff CS of 4000) were associated with a stabilization (plateau) or even a slight decline in serum the CA 15-3 levels.
Bone Marrow Sample Analysis
A bone marrow aspiration was performed at baseline and in the 4th cycle of chemotherapy selectively in 6 patients receiving chemotherapy with 10-day G-CSF administration. Immunocytochemical analysis for MUC1 expression in the myeloid lineage yielded negative or weak staining in both immature myeloid cells as well as band forms and mature neutrophils at baseline. Although there was no evidence of induction of MUC1 expression after 4 courses of G-CSF–boosted chemotherapy in all myeloid cells (baseline ICCS = 51, Cycle 4 ICCS = 58, two-sided P = 0.89), this lack of activation was observed mainly in progenitor and immature cells. In contrast, there was induced MUC1 expression in bone marrow band forms and neutrophils at the peak of chemotherapy, although this did not reach statistical significance, possibly because of the small sample. When analyzed separately, the median ICCS for mature cells increased from 39 at baseline to 71 during Cycle 4 (two-sided P value = 0.06). As in the periphery, MUC1 staining was observed mainly in the cytoplasm and, to a lesser extent, in the cellular membrane. An unexpected finding was the universal strong cytoplasmic MUC1 staining observed in megakaryocytes, both at baseline and during Cycle 4.
CA 15-3 is one of several epitopes exposed on MUC1, a mucin normally present in the apical surface of bronchial, mammary, gastrointestinal, pancreatic, ovarian, and uterocervical glandular epithelial cells.5 Existing data favor an immunoregulatory and intercellular adhesion modulatory role for this mucin.6 The MUC1 molecule is comprised of a cytoplasmic tail, a transmembrane segment, and a rod-like extracellular domain that consists of a VNTR of 20 amino acids. In the Golgi apparatus, carbohydrate side chains are added in serine or threonine residues of the polypeptide core antigen. In breast carcinoma, the MUC1 molecule is overexpressed in all surfaces of the malignant cell as well as being aberrantly glycosylated.7 The latter results in certain epitopes such as CA 15-3 being exposed in contrast to masking or lesser exposure in normal epithelia, a basis for its use as a tumor marker in the follow-up of patients who received resection or patients with advanced breast carcinoma.
In the current study, we investigated an unexplained elevation in serum CA 15-3 levels during the course of intensive adjuvant chemotherapy with G-CSF support in women with resected early breast carcinoma. Although self-limited (≤ 2 × ULN), the tumor marker remained significantly elevated throughout the treatment period and subsided to normal levels after the cessation of treatment. In patients receiving 10 days of G-CSF per chemotherapy cycle, a 2–4-fold increase was observed in comparison with baseline. Given that recording of two or more serial, elevated marker levels has been shown to herald clinical evidence of metastatic spread, such a finding could easily alarm the physician as well as the patient.8 Because dose-dense, G-CSF–supported adjuvant chemotherapy for patients with early breast carcinoma is being prescribed more often in view of studies demonstrating its superior efficacy versus conventional chemotherapy,2 the need to investigate the mechanisms of serum CA 15-3 increase is likely to become pressing so as to avoid unnecessary costs and anxiety.
Possible causes of CA 15-3 abnormalities include the recurrence of breast carcinoma and the administration of cytotoxic chemotherapy or the G-CSF cytokine. The first interpretation is unlikely because serum CA 15-3 levels did not continue to increase beyond twice the ULN and reverted to baseline levels after the completion of treatment. Moreover, no excess of disease recurrences was observed in the group of patients who experienced CA 15-3 elevation. Finally, an increase in serum CA 15-3 levels was reported in the patients in the current study with lymphoma who were receiving G-CSF–boosted chemotherapy, patients in whom no malignant mammary epithelial cells are present by definition. Massive tumor lysis induced by chemotherapy or even spontaneously, resulting in the release of tumor markers in the bloodstream, may apply to patients with bulky metastatic, chemosensitive malignancies such as lymphomas, leukemias, or germ cell tumors.9, 10 However, our study population was, by definition, disease free after the initiation of chemotherapy with baseline serum CA 15-3 levels comparable to those of healthy females. Even if some patients might have harbored micrometastases, the very low, if any, tumor burden makes the tumor lysis effect theory difficult to support. Of note, one study that reported false-positive recordings of CA 15-3 in the course of adjuvant chemotherapy in patients with breast carcinoma showed a higher false-positive rate in patients receiving G-CSF support.11
Tyshler et al.11 reported an increase in serum CA 15-3 levels in tumor-free patients with resected breast carcinoma receiving adjuvant 5-fluorouracil, doxorubicin, and cyclophosphamide chemotherapy without G-CSF support, although less frequently than when G-CSF was given. The toxic effects of chemotherapy on normal epithelia or rebound neutrophilia after chemotherapy-induced bone marrow ablation may be a possible cause. This observation prompted us to include the control cohort who received adjuvant chemotherapy cycles without cytokine support. We failed to observe an elevation in the serum CA 15-3 level or the frequent occurrence of hand-and-foot syndrome, as Tyshler et al. did, despite the administration of dose-intensive anthracycline and taxane-based chemotherapy.
The serum CA 15-3 increase was found to be both clinically and statistically significant in patients receiving chemotherapy cycles with G-CSF support, but not in those receiving chemotherapy cycles without cytokine administration. Moreover, the marker increase was more significant when the cytokine was given for 10 days rather than for 5 days per cycle. These findings were universally observed in patients with a variety of malignant disorders who were receiving a variety of chemotherapeutic regimens, provided the G-CSF cytokine (filgrastim) was administered for 10 days per cycle. These data point to the cytokine as the causative factor for the marker increase. However, although CA 15-3 shedding from tumor cells can be excluded, the question of its source remains open. Normal epithelial or hematopoietic tissues could be stimulated by G-CSF toward MUC1 overexpression and shedding of soluble CA 15-3. G-CSF receptors are present in the developing human fetus, intestinal mucosa, breast, and endometrium, in addition to pluripotent bone marrow stem cells and neutrophil precursors.12–14
Although we did not study MUC1 expression in normal epithelial tissues, the results of the current study do present strong evidence for the induction of MUC1 overexpression in peripheral blood and bone marrow neutrophils, caused by G-CSF. At baseline, all patients had either negative or weak MUC1 cell surface and cytoplasmic expression as detected by immunocytochemistry and flow cytometry. Weak cytoplasmic MUC1 staining in hematopoietic cells was reported as early as 1983 but at that time was attributed to nonspecific antibody binding.15 Recently, low-level MUC1 expression was observed by reverse transcriptase–polymerase chain reaction (RT-PCR) and Western blot assays in peripheral blood granulocyte and mononuclear cells from healthy donors.16 When MUC1 was employed as a tumor-specific marker for the molecular detection of micrometastases in the bone marrow and lymph nodes of patients with resected lymph node-negative breast carcinoma, it was shown to have low specificity or resulted in overestimation of the frequency of microscopic tumor dissemination.17, 18 These findings very well could have been due to low-level MUC1 expression in normal hematopoietic cells. Other investigators reported MUC1 expression in erythroblasts and mature B lymphocytes.19
In the current study, 5-day and mostly 10-day G-CSF administration resulted in the induction of cytoplasmic and, to a lesser extent, surface MUC1 expression in neutrophils of the peripheral blood and bone marrow. This phenomenon was not observed when chemotherapy cycles without G-CSF support were administered. In keeping with the immunocytology findings, flow cytometric analysis confirmed increased cytoplasmic, but not membranous, MUC1 neutrophil overexpression in patients receiving 10-day G-CSF support. MUC1 expression in the cytoplasm, but not on the cell surface, of normal human intestinal and breast tissues has been reported by investigators who used MoAbs reacting with the APDTR sequence of the VNTR region or the oligosaccharide side chains.20 Any effort to interpret this peculiar finding is speculative at this point in time (e.g., the presence of crossreactive epitopes in the cytoskeleton, cytoplasmic localization of MUC1 molecules resulting from alternate mRNA splicing, migration of MUC1 molecules through the endoplasmic reticulum and Golgi apparatus, sites of glycosylation, and cellular membrane or “packing”of MUC1 molecules for subsequent exocytosis).21, 22 The cytoplasmic localization of MUC1 in neutrophils could reflect a protracted “stay” in the cellular glycosylation compartment as a result of the G-CSF effect or simply the timing of the venipuncture (3–7 days from the last G-CSF injection).
Irrespective of possible contribution by normal epithelial tissues, the increase in serum CA 15-3 levels appears to be due to induced MUC1 overexpression per neutrophil as wellas to an increase in the bone marrow output of granulocytes, as shown by the marked induction of the CS in the 10-day G-CSF group. Both are G-CSF mediated. Studies of MUC1 expression in accessible normal epithelial tissues (such as exfoliated cells from oral, vaginal, or intestinal mucosa) and RT-PCR evidence of increased numbers of MUC1 mRNA transcripts in neutrophils will allow definitive assessment of the role of normal epithelial and myeloid cells in ectopic production of MUC1 epitopes under G-CSF stimulation. In any case, evidence presented in the current study point to the neutrophil as a source and to G-CSF as the trigger for the CA 15-3 increase in patients with resected breast carcinoma. This finding highlights the limitations and consequences of tumor marker use and abuse, because serum CA 15-3 monitoring is used in Europe not only for evaluating the course of advanced disease but also for following tumor-free patients either receiving or not receiving adjuvant chemotherapy. Physicians aware of this interaction can reassure their patients and avoid imposing unnecessary burden to the medical service as well as financial costs to limited budgets.
- 2Randomized trial of dose-dense versus conventionally scheduled and sequential versus concurrent combination chemotherapy as postoperative adjuvant treatment of node-positive primary breast cancer: first report of Intergroup trial C9741/CALCB 9741. J Clin Oncol. 2003; 21: 1–9., , , et al.
- 5Tissue and tumor distribution of human polymorphic epithelial mucin. Cancer Rev. 1988; 11: 55–101., , , , .
- 13Human endometrial expression of granulocyte colony-stimulating factor (GCSF) and its receptor, stimulation of endometrial GCSF production by interleukin-1 beta, and GCSF inhibition of choriocarcinoma cell proliferation. Am J Reprod Immunol. 1996; 36: 278–284., .