Mammaglobin Expression in Leukapheresis Products Is a Predictive Marker of Poor Prognosis in Women with High-Risk Breast Cancer

The tumor-node-metastasis (TNM) classification offers a fairly good chance of evaluating the probability of relapse in patients with large breast tumors and positive nodes. Expression of estrogen receptor (ER)/progesterone receptor (PgR), age, and menopausal status, as well as overexpression of oncogenes, provide other criteria for postoperative decision-making (1).

Unfortunately, more than 20% of low-risk breast cancer patients relapse at 5 years, and almost 70% of high-risk patients develop metastatic disease within 10 years despite adjuvant chemotherapy (2, 3). A partial explanation for these data may be the presence of occult micrometastatic disease at diagnosis. The identification of tumor cells in the bone marrow (BM), peripheral blood progenitor cell collections (PBPCCs), or peripheral blood (PB) is not yet considered a routine procedure in determining adjuvant treatment, but it does represent a valid prognostic tool worth further investigation.

One issue that remains controversial is the lack of standardized techniques and markers used by groups studying this problem. In particular, the use of immunocytochemical techniques grants considerable discretion in identifying different cell subpopulations and distinguishing epithelial from nonepithelial cells. Reverse transcription-polymerase chain reaction (RT-PCR) gives a higher percentage of sample positivity but, when used for cytokeratin (CK), also increases the number of false positives caused by pseudogene amplification or illegitimate transcription.

Some investigators have correlated micrometastatic disease in BM with other known prognostic factors and have determined it to be an independent predictor of relapse, as evaluated by identification of CK-expressing cells using sensitive immunocytochemical techniques (4, 5, 6, 7, 8, 9). In turn, a recent meta-analysis has concluded that independent prognostic impact remains to be substantiated by further study relying on standardized protocols (10).

Patients who are affected by disseminated disease and who undergo high-dose chemotherapy will have approximately 20% to 25% of granulocyte colony-stimulating factor (G-CSF)–mobilized PBPCC contaminated by tumor cells using routine procedures (11, 12, 13). Many others, even those with histologically negative BM, indeed have 10% to 15% of epithelial cells in the autologous BM or PB progenitor cells (PBPCs) used for transplantation (14, 15).

The extent of the disease and the sites of metastases seem to correlate with contamination of the hematopoietic support as well as with the number of tumor cells detected (13). Studies using clonogenic assay techniques have also shown that micrometastatic tumor cells maintain their viability and may be related to disease relapse, as demonstrated by gene marking experiments (15, 16).

To conduct a detailed study on the possible prognostic role of PBPCC in breast cancer patients undergoing adjuvant high-dose chemotherapy, we introduced a new approach based on a concomitant evaluation of a multiple panel of genes, including MAS and MAM, and an additional step for CK RT-PCR amplification with hybridization of an internal probe to enhance specificity without affecting sensitivity.

Both MAS and MAM represent good candidates for typing breast cancer disease.

MAS is a protein related to the serpin family of protease inhibitors and is widely expressed in epithelial tissues. In vitro, it shows tumor suppressor activity, acts directly on endothelial cells to stop their migration toward basic fibroblast growth factor and vascular endothelial growth factor treatment, and limits mitogenesis and tube cellular formation. In a xenograft mouse model of human prostate cancer, MAS blocks tumor growth and reduces tumor-associated microvessel density working as an antiangiogenesis modulator (17, 18). The role of MAS as an epithelial marker of micrometastasis is controversial. Some reports evaluating BM have correlated it to poor prognosis; others have shown a possible effect as a protection factor from relapse (19, 20, 21). In breast cancer, Maas et al. (22) report on the loss of MAS expression with increasing malignancy (from ductal carcinoma in situ to invasive carcinoma to lymph node metastasis) by mechanisms of transcriptional deregulation. On the contrary, Umekita and Yoshida (23) have shown up-regulation with the acquisition of an aggressive phenotype.

MAM, a mammary-specific member of the uteroglobin gene family, is a glycoprotein detected only in the mammary gland that is overexpressed in a high percentage of human breast cancers (24, 25). Its expression seems to be related to progression from localized to locally advanced and metastatic disease (25). MAM is detectable in PB and apheresis products of metastatic breast cancer patients, but a correlation between MAM expression and known prognostic factors for breast cancer has not yet been demonstrated (26, 27, 28). Leone et al. (29) developed a real-time quantitative polymerase chain reaction showing MAM to be the more affordable marker in detecting epithelial cell contamination in hematologic samples.

In our institution, patients with stage III breast cancer underwent surgery, followed by three courses of adjuvant myeloablative high-dose chemotherapy, each supported by reinfusion of PBPCs previously harvested after G-CSF mobilization. The regimens applied were a combination of cytotoxic drugs, such as taxanes, anthracyclines, ifosphamide or cyclophosphamide, carboplatin, and etoposide, as required by our experimental protocols [epirubicin and cyclophosphamide (EC); taxotere, epirubicin, and cyclophosphamide (TEC); taxotere, ifosphamide, carboplatin, and etoposide (T-ICE); and ifosphamide, carboplatin, and etoposide (ICE)].

Cells obtained from the PBPCCs harvested before chemotherapy administration to the patients were analyzed by RT-PCR for molecular expression of CK, MAS, and MAM to check their sensitivity and specificity. Patients reinfused with tumor-contaminated leukapheresis were then evaluated for relapse-free survival (RFS) and overall survival (OS).

One hundred and ninety-four patients were enrolled in the study from September 1998 to May 2001 after informed consent was obtained according to the Helsinki Declaration. They had a performance status of 0–1 (Eastern Cooperative Oncology Group scale) and histologically proven breast cancer classified as T_1–3, N_1–3, M₀, identifying stage III disease according to the most recent classification. All of the patients had undergone surgery, followed by three courses of adjuvant high-dose chemotherapy, each supported by reinfusion of PBPCC previously harvested after G-CSF mobilization.

One hundred and ninety-four patients (90 for MAM) were analyzed for the presence of epithelial cells on 481 PBPC apheresis (216 for MAM). Five milliliters of PBPCs were collected from apheresis after 4 days of BM stimulation by G-CSF and lysed by NH₄Cl to remove red cells, resuspended in buffered saline, washed twice, and counted. Part of the cell suspension (12–16 × 10⁶ cells) was cytospun on slides; part was stored with guanidine isothiocyanate at −80°C for further mRNA analysis.

Total cellular RNA was extracted from the cells by the QIAamp RNA kit (Qiagen, Hilden, Germany) and treated with a reverse transcriptase enzyme (SuperScript II; Gibco, Gaithersburg, MD). The resultant cDNA was amplified to detect CK, MAM, and MAS expression according to Moschinski et al. (30), Zach et al. (31), and Luppi et al. (32), respectively. Single-round RT-PCRs using β-actin–specific primers confirmed the presence of intact RNA, adequate cDNA synthesis, and the absence of inhibitors. CK, MAM, and MAS amplified products were visualized on 2% agar gel stained with ethidium bromide as single bands of 108, 201, and 175 bp, respectively. A hybridization step with an internal radiolabeled probe (Figs. 1 and 2) was introduced to reduce false-positive results in CK amplification.

The characteristics of the patients were summarized according to the following factors: age, stage, lymph node status, receptor status, c-ErbB2 status, Ki67 status, and treatment schedule. We studied the distribution of CK, MAS, and MAM expressions; their relationship with the previously mentioned factors; and their potential prognostic role.

The number of neoplastic lymph nodes was considered as both a continuous variable and a discrete variable grouped as follows: 2 to 9 neoplastic lymph nodes, 10 to 19 neoplastic lymph nodes, and ≥20 neoplastic lymph nodes. c-ErbB2 status was dichotomized into −/+ versus 2+/3+. Receptor expression of >10% was considered positive; Ki67 expression of >20% was considered positive.

The association between known and potential prognostic variables was assessed using the χ² test and Fisher’s exact test. RFS and OS were estimated with the Kaplan-Meier method; time was calculated from the start of chemotherapy to relapse or last follow-up for RFS and to death or last follow-up for OS. A Cox regression model (33) was fitted to adjust for age and account for other characteristics relevant in assessing the prognostic role of known and potential factors on RFS and OS.

All of the analyses were done with S-Plus software (34).

We evaluated a total of 481 PBPCCs obtained from 194 stage III breast cancer patients with a median age of 46 years (range, 24–71 years) and a median number of neoplastic lymph nodes equal to 13 (range, 3–63 neoplastic lymph nodes). Table 1 shows the distribution of the patient’s main prognostic factors, together with the high-dose chemotherapy schedule administered.

Patients were classified into three categories based on the number of lymph nodes involved: 61 patients (31%) with 2 to 9 lymph nodes involved, 86 patients (44%) with 10 to 19 lymph nodes involved, and 47 patients (24%) with ≥20 lymph nodes involved. As far as receptor status is concerned, 110 (57%) patients were ER positive, and 87 (45%) were PgR positive; Ki67 was positive in 123 (63%) patients, whereas c-ErbB2 status, available for only 131 cases, was 2+ or 3+ in 53 (28%) others.

The follow-up was quite homogeneous, with a median time of 36 months (first quartile, 32 months; third quartile, 41 months). To date, we have observed 78 relapses and 35 deaths. After 3 years, the estimated proportion of patients free of relapse was 67% [95% confidence interval (CI), 60.1–74.6], and the estimated proportion of patients alive was 83% (95% CI, 77.1–88.8). Fig. 3 shows the Kaplan-Meier estimated curves for OS (Fig. 3,A) and RFS (Fig. 3 B) in the whole sample, with 95% CIs.

Specificity was tested on 12 healthy donors’ BM and PB samples, 13 BM samples from patients affected by nonepithelial malignancies, and 12 cell lines of epithelial and nonepithelial malignancies (TF1, CEM, Namalwa, Molt, Huvec, K-562, Sultan Neo, Rap1, Dohh2, Karpass 299, Bonna 12, and Jurkat).

CK, MAS, and MAM amplified transcripts revealed single bands of 108, 175, and 201 bp, respectively. The specificity of the CK amplified samples was further confirmed by hybridization with a specific, ³²P-labeled internal probe, which showed that not all of the RT-PCR–amplified products were CK19 transcripts. About 30% of them did not actually anneal to the probe and were not revealed by autoradiography (Fig. 1), providing evidence that pseudogene expression and/or illegitimate transcription is a real but avoidable problem. Apart from the K-562 leukemic cell line, which expressed MAS RNA, no other false-positive results were detected in the 12 cell lines. No false positives were found in the BM and PB samples used as controls of specificity in any of the amplification procedures for CK, MAM, and MAS, as reported in Table 2.

The sensitivity of the three markers was determined by mixing decreasing cell numbers of MCF-7 and CG-5 breast cancer cell lines with PB mononuclear cells obtained from healthy donors. The sensitivity of CK, MAM, and MAS RT-PCR amplification was 1 × 10⁻⁷, 1 × 10⁻⁶, 1 × 10⁻⁶ cells, respectively (Fig. 2).

Table 3 reports the distribution of CK, MAS, and MAM, the latter of which was available on a subset of 90 patients. In the tests for associations among the three, only that of CK with MAS was significant (P < 0.01); no significant association was found between these three factors and the other characteristics described in Table 1 (age, stage, lymph nodes, receptor status, c-ErbB2, Ki67, and schedule).

CK- and MAS-expressing cells were detected in 138 (71%) and 70 (36%) of the samples, respectively. When the two factors were added, individually and then together, to a Cox regression analysis including variables of known clinical relevance, no statistically significant prognostic effect was shown for either RFS or OS (Table 4). PgR and c-ErbB2 were not included in the evaluation because they added no additional information; c-ErbB2 implied the exclusion of some cases because of missing values. Fig. 4 A and B show the similarity in RFS and OS between patients with a positive and negative status for the two markers.

The only relevant effect on prognosis we could find was a positive ER status, which, as expected, was associated with a significant reduction of risk of both relapse and death.

The previously described Cox model was applied to a subset of 90 patients who could also be evaluated for MAM expression on 216 PBPCCs. The main characteristics of these patients, as described in Table 1, were very similar, with practically the same distribution as in the overall sample. The role of maspin and CK19 was also similar because the estimated coefficients reveal no significant effect on prognosis.

Fourteen of 90 (15.5%) patients expressed the MAM gene as evaluated on 216 PBPCCs, and 12 of them relapsed in different sites within the following 3 years. Additional information for this subgroup showed a statistically significant (P = 0.003) increased risk of relapse for a positive status (Table 5).

The CIs get wider because the set of patients is smaller, but the primary consideration was to observe the effect of MAM expression while confirming that the other factors had a similar behavior. There was no evident association for OS. Fig. 4 C shows the difference in RFS curves for MAM-positive versus MAM-negative samples.

Mobilization of tumor cells together with hematopoietic progenitor cells into the PB of patients with solid tumors is possible after chemotherapy + G-CSF or G-CSF alone and frequently occurs in PBPCC (13, 14, 15). Despite the limited data, there is some evidence of apheresis tumor contamination and its impact on prognosis (7, 9, 11, 12, 13, 14). In particular, the presence of occult tumor cells in the graft has been shown to have a negative influence on prognosis when evaluated by immunocytochemistry for CK gene expression (35, 36). Kasimir-Bauer et al. (37) have also shown that tumor cells are able to survive dose-intensive or high-dose chemotherapy, when evaluated by the same techniques in BM aspirates after therapy.

Unfortunately, methods have not been standardized, and more specific markers are needed to better understand the real impact of occult micrometastasis on the prognosis of breast cancer patients. MAS and MAM represent good options for evaluation.

MAS is widely expressed in epithelial tissues, but the role of MAS as an epithelial marker of micrometastasis is controversial because some data correlate it with poor prognosis, whereas others have shown a possible effect as a protection factor from relapse (17, 18, 19, 20, 21).

MAM seems to be much more specific for breast cancer because it is also detectable in the PB of metastatic breast cancer patients, and it has been shown to be the more affordable marker in detecting epithelial cell contamination in hematologic samples (29).

Our study was designed to verify the possible significance of tumor cell detection in PBPCCs using different standard and innovative markers. Cells obtained from leukapheresis were analyzed by RT-PCR for molecular expression of CK, MAS, and MAM to check their sensitivity and specificity. Patients reinfused with tumor-contaminated leukapheresis were then evaluated for RSF and OS.

Our data confirm that a large proportion of high-risk breast cancer patients undergoing leukapheresis have tumor cell contamination in their autografts as evaluated by CK, MAS, and MAM expression. In fact, 176 of 194 patients (91%), evaluated on a total of 481 PBPCCs, had at least one positive sample for any of the markers used. Such a high percentage could be explained by concomitant evaluation with the three different markers and the sensitivity reached by each of the techniques developed. However, the presence of CK and MAS molecular signals did not show any prognostic relevance as a possible marker of evolving metastatic disease and therefore has no impact on clinical outcome.

When CK was evaluated by RT-PCR, sensitivity reached very high levels without losing specificity. Using our hybridization strategy, we were able to eliminate up to 30% of false-positive results, depending on erroneous pseudogene amplification or illegitimate transcription. Apart from that, differences between CK-positive versus CK-negative curves show no differences in RFS and/or OS advantage, probably due to the low number of events observed. MAS shows similar behavior on evaluation of PBPCC and prognosis.

We have to emphasize that the data on prognosis we report are related only to leukapheresis and not to BM, where a micrometastatic disease could be much more easily identified and related to patient outcome. Instead, we are evaluating PBPCCs that had been harvested after growth factor stimulation in a nonphysiologic setting. This situation could influence the capability of the tumor cells to induce relapse, even though it has been demonstrated that the tumor cells maintain good viability (15, 16). In conclusion, MAM expression on leukapheresis, as well as expressions of CK and MAS, could reflect the presence of residual cancer cells in the BM and their mobilization by growth factor treatment. However, prognosis could be indirectly related to BM involvement by the disease and/or to reinfusion of tumor cells together with leukaphereses.

However, 48 patients included in our study were also tested for expression of mammaglobin on BMs, revealing a statistically nonsignificant trend toward bad prognosis. Moreover, there was no statistically significant relationship between the two sites of evaluation (leukapheresis and BM; data not shown).

The strength in assessing the role of these variables in OS and RFS analyses depends on the number of patients and on the number of events.

We analyzed an adequate number of patients (194) and in particular a very large number of leukaphereses (481 in total), but to date, we have only observed 78 relapses and 35 deaths. A longer follow-up will allow us to observe more events and therein increase the possibility of detecting a difference in terms of survival or relapse. Nonetheless, the estimated coefficients appear fairly stable when fit to different models, even if they are not very precise (e.g., large CIs). As such, we believe that these results, although limited in quantifying the effect, provide a useful investigation of the role of these markers.

MAM does seem to be a good marker for specificity because it is closely linked to the identification of mammary-derived cells and has been related to progressive disease. Real-time polymerase chain reaction experiments have shown that high levels of expression are warranted to obtain an effect on prognosis (29). In our data, however, a nonquantitative RT-PCR evaluation of MAM expression on PBPCs also showed a statistically significant correlation with an increased risk of relapse for high-risk breast cancer patients. No correlation was found with the sites of relapses, although there is a relative prevalence of metastasis to the skeleton and liver (data not shown).

Previously published information has shown that epithelial tumor cells, which are present in the apheresis, are clonogenic and may cause relapse (15, 16). We found clinical evidence that reinfusing MAM-expressing tumor cells could worsen prognosis.

In conclusion, the multigene panel we created to study PBPCC influence on prognosis is a significant step toward the highest possible sensitivity and specificity. When evaluated on apheresis, MAM does indeed work as a negative prognostic marker of disease and, together with CK and MAS, could even be used on BM and PB to follow our breast cancer patients more thoroughly.

We thank Drs. Thomas Foutz and Marlene Klein for helpful comments and critical reading of the manuscript.