Lung cancer is the most common cause of death by

Transcription

1 Predicting the Sites of Metastases From Lung Cancer Using Molecular Biologic Markers Thomas A. D Amico, MD, Thomas A. Aloia, MD, Mary-Beth H. Moore, BS, Debbi H. Conlon, HT(ASCP), James E. Herndon II, PhD, Michael S. Kinch, PhD, and David H. Harpole, Jr, MD Division of Cardiothoracic Surgery, Duke University Medical Center, Durham, North Carolina Background. The use of molecular markers in staging non-small cell lung cancer (NSCLC) has been supported in retrospective prognostic models but has not been evaluated in predicting sites of metastases. Methods. Pathologic specimens were collected from 202 patients after complete resection for stage I NSCLC, who were subsequently found to have no metastases at 5 years (n 108), isolated brain metastases (n 25), or other distant metastases (n 69). A panel of eight molecular markers of metastatic potential was chosen for immunohistochemical analysis of the tumor: p53, erbb2, angiogenesis factor viii, EphA2, E-cadherin, urokinase plasminogen activator (UPA), UPA receptor, and plasminogen activator inhibitor. Results. Patients with isolated brain relapse had significantly higher expression of p53 (p 0.02) and UPA (p 0.002). The quantitative expression of E-cadherin was used to predict the site of metastases using recursive partitioning: 0 of 92 patients with E-cadherin expression of 0, 1, or 2 developed isolated cerebral metastases; 0 of 33 patients with E-cadherin expression of 3 with UPA of 1 or 2 and ErbB2 of 0 developed brain metastases. Of the remaining patients at risk (UPA 3), the risk of isolated cerebral metastases was 21 of 57 patients (37%). Conclusions. This study demonstrates that molecular markers may predict the site of relapse in early stage NSCLC. If validated in an ongoing prospective study, these results could be used to select patients with isolated brain metastases for adjuvant therapy, such as prophylactic cranial irradiation. (Ann Thorac Surg 2001;72:1144 8) 2001 by The Society of Thoracic Surgeons Lung cancer is the most common cause of death by malignancy in both men and women in the United States [1]. The current staging system for non-small cell lung cancer (NSCLC) considers the size and location of the primary tumor (T), the involvement of regional lymph nodes (N), and the presence of distant metastases (M) [2]. The standard treatment of patients with stage I NSCLC (T1-2, N0) is resection of the primary tumor alone (no adjuvant therapy) [3]. However, even after complete resection, 5-year survival is only 55% to 72% in this group of patients, due predominately to the development of distant metastases [2 5]. In order to select a subgroup of patients with stage I disease who might benefit from adjuvant therapy, investigators have attempted to identify factors which predict poor prognosis, including analysis of performance status, histologic subtype, size of the primary tumor, the degree of tumor differentiation, mitotic rate, and evidence of lymphatic or vascular invasion [6 8]. These factors are indirect measures of tumor aggressiveness and, to date, have not identified a group of stage I patients who would benefit from adjuvant therapy. Recent studies have focused on the identification of Presented at the Thirty-seventh Annual Meeting of The Society of Thoracic Surgeons, New Orleans, LA, Jan 29 31, Address reprint requests to Dr D Amico, Division of Cardiothoracic Surgery, Duke University Medical Center, Box 3496, Durham, NC 27710; damic001@mc.duke.edu. biologic markers that predict early recurrence and death in patients with NSCLC. Molecular biologic substaging the use of oncogenes and other oncogenic factors to improve risk stratification of the TNM staging system had been applied in the development of a prognostic model of recurrence in stage I NSCLC patients [9 14]. These molecular markers may be classified into subgroups based on their mechanism of action in the metastatic process: growth regulation (erbb2) [9], cell cycle regulation/apoptosis (p53) [9], angiogenesis (factor viii) [9], cellular adhesion (EphA2, E-cadherin) [15, 16], and basement membrane invasion (UPA, UPAR, PAI-1) [17, 18]. In this study, the ability of molecular biologic markers to predict the site of metastases is assessed in a population of patients with pathologic stage I NSCLC. Although postresection adjuvant therapy has not been demonstrated to be effective in patients with stage I disease [3], it is postulated that risk stratification will identify a group of patients with sufficiently elevated risk of death to justify adjuvant therapy. In order to select a subgroup of patients with stage I disease that might benefit from adjuvant therapy, measurable molecular markers must be correlated with the development of specific metastatic patterns. To address this problem, this study focuses on the identification of biologic markers that predict cerebral metastasis in patients after complete resection of stage I lung cancer. It has been demonstrated that molecular 2001 by The Society of Thoracic Surgeons /01/$20.00 Published by Elsevier Science Inc PII S (01)

2 Ann Thorac Surg D AMICO ET AL 2001;72: LUNG CANCER METASTASES 1145 biologic markers may differentiate patients with stage I NSCLC who eventually develop brain metastases from patients who do not have recurrence [19]. This study is designed to determine whether molecular markers may be used to predict the site of metastases after resection among a population that includes patients with no metastases, patients with isolated cerebral metastases only, and patients with undifferentiated distant metastases. Material and Methods Following approval by the Human Subjects Review Committee, pathologic specimens were collected from 202 patients after complete resection for stage I NSCLC, who were subsequently found to have no metastases at 5 years (n 108), isolated brain metastases (n 25), or other distant metastases (n 69). Each block was sectioned and underwent immunohistochemical analysis using antibodies to p53, erbb2, angiogenesis factor viii, EphA2, E-cadherin (E-cad; BioGenex Laboratories, San Ramon, CA), urokinase plasminogen activator (UPA), UPA receptor (UPAR), and plasminogen activator inhibitor (PAI-1; American Diagnostica, Greenwich, CT), using automated immunostainers (BioGenex) and a standard horseradish immunoperoxidase technique as previously described [9, 10]. Briefly, each tissue block underwent paraffin microtome sectioning (4 to 6 m), slide labeling, and deparaffinization with xylene and ethanol. Antigen retrieval was completed after microwaving and phosphate buffered saline (PBS) washing. After quenching endogenous peroxidase in 3% hydrogen peroxide, the slides were gradually brought to water and then underwent Antigen Retrieval (BioGenex, U.S. Patent # 5,244,787) in citrate buffer, ph 6.0. The slides were placed on the OptiMax PLUS automated slide stainer (BioGenex) where they were rinsed in three washes of PBS, preincubated in Power Block (BioGenex) for 8 minutes, and then incubated with primary antibody for 1 hour at room temperature. The reaction product was developed by incubation with a biotinylated, affinity purified secondary antibody for 20 minutes, followed by incubation with biotinylated horseradish peroxidase complex for 20 minutes. The slides were developed with the chromogen, diaminobenzidine, and counterstained with hematoxylin. Known positive tumors and normal lung tissue were used as positive and negative controls, respectively. All slides were read independently by two experienced observers who were blinded to tissue source. Observers scored each slide based on the property of the antibody employed. For p53, erbb2, EphA2, E-cad, UPA, UPAR, and PAI-1, each slide was scored on a scale of 0, 1, 2, and 3. In this scale, 0 denotes no staining; 1 denotes staining in less that 20% of tumor cells; 2 denotes staining in 21% to 50% of tumor cells; and 3 denotes staining in greater than 50% of tumor cells [11]. Scores of 2 or 3 were considered positive for the purpose of statistical analysis. Differences in immunohistochemical scores were rare, but when present, were resolved by consensus. Angiogenesis factor viii was scored using two methods [11]. First, a Peri 8 score was determined by counting the number of positively stained microvessels (MV) in 10 consecutive fields (10 ) at the periphery of the tumor. The area of highest MV density was then determined by scanning the tumor at 10 power. The Hot 8 score was determined by counting the stained microvessels in three consecutive fields (20 ) within the area of highest MV density. Using this method, there was little intraobserver variability in either the Hot 8 or Peri 8 score. Reported scores are the average of the two observers individual scores. Exact 2 tests were used to determine whether the expression of a tumor marker was related to the site of relapse. Binary recursive partitioning was used to examine the relationship between tumor markers and the later occurrence of brain metastases [20]. Sometimes referenced as classification or regression trees, the goal of this analysis is to develop an algorithm for classifying or predicting patient outcomes, such as the development of isolated cerebral metastases, based on marker expression. The analytic procedure involves the iterative division of a single heterogeneous patient group into two less heterogeneous subgroups. A node or patient subgroup is considered homogeneous if all or none of the patients have brain metastases. At each step in the iterative partitioning process, a patient subgroup is chosen to split to maximally decrease the group heterogeneity or variability in the left and right branches of the classification tree. Results In this series of 202 patients who underwent complete resection for stage I NSCLC, there were 108 patients who were subsequently found to have no metastases at 5 years (No mets). Of the patients who eventually developed metastatic disease, there were 25 patients with isolated cerebral metastasis (Brain mets), and 69 patients with other distant metastases (Distant mets). The mean time to development of cerebral metastases was 16 months (range 0 to 6 years). The expression of molecular markers according to site of metastases, including all patients who are considered to have positive expression (2 or 3), is described in Table 1. The expression of p53 was significantly higher in patients with distant mets or brain mets, compared to those with no metastases. The expression of UPA was significantly higher in patients with brain metastases than in patients with only distant metastases. The quantitative expression of E-cad was used to predict the site of metastases (Table 2). All of the patients who developed isolated cerebral metastases had tumors with E-cad expression 3. Using recursive partitioning (Fig 1), the site of metastases was analyzed using the expression of other markers. None of the 92 patients with E-cad 0, 1, or 2 developed brain metastases. Of the 109 patients with E-cad 3, those with UPA less than 3 and ErbB2 0 did not develop cerebral metastases. Of the remaining patients at risk (UPA 3), the risk of isolated cerebral metastases was 21 of 57 (37%).

3 1146 D AMICO ET AL Ann Thorac Surg LUNG CANCER METASTASES 2001;72: Table 1. Positive Expression of Molecular Markers According to Site of Metastases Variable No Mets (n 108) Distant Mets (n 69) Brain Mets (n 25) p Value a p ErbB NS Factor viii NS EphA NS E-cad NS UPA UPAR NS PAI NS a p Value from 2 testing which assesses whether the percentage of patients with marker expression is the same among patients with brain mets, distant mets, or no mets. Brain mets patients with isolated cerebral metastasis; Distant mets patients with extracranial metastases; E-cad E-cadherin; No mets patients with no metastatic disease; NS not significant; PAI-1 plasminogen activator inhibitor type 1; UPA urokinase plasminogen activator; UPAR UPA receptor. Comment Adjuvant therapy after complete resection of stage I NSCLC has not been proven to improve survival, compared to surgery alone [3]. Nevertheless, the use of adjuvant therapy may prove to be a beneficial strategy among subgroups of patients with a high risk of recurrence. The analysis of molecular marker expression has been demonstrated to provide important additional prognostic information in retrospective studies of patients with stage I NSCLC [9 14], and molecular biologic substaging may eventually be used to assess prognosis within the TNM staging system. This concept is currently being tested in a multi-institutional, prospective, randomized trial by the Cancer and Leukemia Group B. Currently, the optimum therapy for patients with who present with isolated cerebral metastases after resection of NSCLC includes resection of the metastasis and adjuvant cranial irradiation [21, 22]. Although this strategy does improve the outcome in this group of patients, the overall survival is less than 20%. Prophylactic cranial irradiation (PCI) has gained acceptance in the management of patients with small cell lung cancer, where the risk of cerebral recurrence is high, even among patients who present with limited disease [23]. Although the use Table 2. Quantitative Expression of E-cad According to Site of Metastases E-cad No Mets Distant Mets Brain Mets Total Total Brain mets patients with isolated cerebral metastasis; Distant mets patients with extracranial metastases; E-cad E-cadherin; No mets patients with no metastatic disease. Fig 1. Predicting the development of brain metastases using recursive partitioning according to marker expression: E-cadherin (E- Cad), urokinase plasminogen activator (UPA), and growth regulation (ErbB2). of PCI has been suggested for patients with NSCLC, this practice has not gained acceptance, as the incidence of isolated cerebral metastases does not justify the risks of PCI [24]. The strategy of using adjuvant therapy for patients with stage I NSCLC after complete resection depends on the ability to predict recurrence or death; in particular, the use of PCI should be limited to patients with a high risk of developing isolated cerebral metastases. The use of molecular markers has not been heretofore used to predict the patterns of failure after resection for early stage NSCLC. This study utilized a panel of eight molecular markers, chosen to comprise multiple pathways in the metastatic process: growth regulation (erbb2), apoptosis (p53), angiogenesis (factor viii), cellular adhesion (EphA2, E- cadherin), and basement membrane invasion (UPA, UPAR, PAI-1). The protooncogene erbb2, which shows extensive homology to erbb1 (the oncogene that is responsible for epidermal growth factor receptor), encodes a membrane-associated tyrosine kinase, that also serves as a growth factor receptor. Mutations in recessive oncogenes that code for apoptosis have been demonstrated to allow unregulated cell growth and contribute to the proliferation of malignant tumors. Normal p53 encodes a nuclear phosphoprotein required to regulate cell growth; mutated p53 stimulates unregulated cell growth and promotes malignancy [9]. Tumor cells initially obtain nutrition from central diffusion from surrounding pulmonary parenchyma. Once a colony outgrows its blood supply, central necrosis occurs. This ischemic process signals angiogenesis in the tumor, allowing capillary ingrowth into the tumor and

4 Ann Thorac Surg D AMICO ET AL 2001;72: LUNG CANCER METASTASES 1147 further growth and invasion. The presence of angiogenesis factor viii has already been demonstrated as a powerful negative prognostic factor [9]. E-cad, a mediator of cell cell interaction, is involved in metastatic invasion; reduced expression of E-cad is associated with advanced malignancy and reduced survival [16]. EphA2 is a member of the Eph family of tyrosine kinases; activation of EphA2, either by E-cad or other antibody-mediated aggregation, decreases the extracellular matrix adhesion [15]. UPA, UPAR, and PAI-1 are involved with tumor invasion and metastasis [17]. UPA and UPAR influence tumor cell migration and invasion through the activation of plasminogen, which acts to break down the extracellular matrix. PAI-1 inhibits both free UPA and receptor-bound UPA. Recently, UPA and UPAR have been demonstrated to have strong prognostic value in predicting tumor recurrence and survival [18]. In this study, the expression of p53 was significantly higher in patients with distant mets or brain mets, compared to those with no metastases. Thus, p53 may be used to identify patients who develop metastatic disease but is not useful in predicting pattern of failure. The expression of UPA was significantly higher in patients with brain metastases (92%) than in patients with only distant metastases (59%); however, since the majority of both groups have positive UPA expression, the ability of UPA to discriminate between the 2 groups is limited. The differences in the expression of p53 and UPA among patients with isolated brain metastases is statistically significant, but the magnitude of risk stratification is not powerful enough to justify altering therapy in this group of patients. Binary recursive partitioning was therefore used to improve the ability of marker expression to identify groups of patients with high or low risk of isolated cerebral metastases. After recursive partitioning, 2 groups are identified: 1 group with a significantly lower risk of developing isolated cerebral metastases and 1 group with a significantly higher risk of developing isolated cerebral metastases, compared to the entire study group. All of the patients who developed isolated cerebral metastases expressed E-cad. When the site of metastases was analyzed using the quantitative expression of E-cad, the ability of the molecular markers to characterize the pattern of failure was apparent. Using the recursive partitioning exploratory analysis, subgroups with very low risk of cerebral metastases are described, and the remaining patients are at significant risk (Fig 1). The subgroups of patients with E-cad 0, 1, 2, or E-cad 3, UPA less than 3, and ErbB2 0 have a risk of 0 in this model. Of the remaining 57 patients with UPA 3, the risk of isolated brain metastases in 37%, similar to the risk of patients with small cell lung cancer [23]. The role of E-cad in the development of cerebral metastases has not been described. All patients with isolated brain metastases in this study demonstrated intact E-cad, suggesting maintenance of cell cell interaction and less malignant potential. It is possible that the preserved integrity of cellular adhesion was protective for the development of widespread metastases, but not isolated brain metastases. The development of isolated brain metastases may be related to a unique property of E-cad; however, it appears more likely that the expression of other molecular factors, such as UPA and ErbB2, have a significant role in determining the site of the metastatic disease. Patients with stage I NSCLC who eventually present with isolated cerebral metastases demonstrate differential expression of molecular markers, compared with patients who never develop metastatic disease and patients who develop distant metastases. Although this model does not identify all patients who develop brain metastases, the pattern of marker expression may identify a subgroup of patients in whom the benefits of PCI may outweigh the risks. The pattern of differential expression may in the future provide insight into the metastatic process, to explain why some patients develop isolated brain metastases and other develop widespread metastatic disease. This research was supported by an American College of Surgeons Faculty Research Fellowship. References 1. Greenlee RT, Murray T, Bolden S, Wingo PA. Cancer statistics CA 2000;50: Mountain CF. Revisions in the international system for staging lung caner. Chest 1997;111: Non-Small Cell Lung Cancer Collaborative Group. Chemotherapy in non-small cell lung cancer: a meta-analysis using updated data on individual patients from 52 randomized clinical trials. BMJ 1995;311: Harpole DH Jr, Herndon JE II, Young WG, Wolfe WG, Sabiston DC Jr. Stage I non-small cell lung cancer. Cancer 1995;76: Nesbitt JC, Putnam JB Jr, Walsh GL, Roth JA, Mountain CF. Survival in early-stage non-small cell lung cancer. Ann Thorac Surg 1995;60: Ichinose Y, Hara N, Ohta M, et al. Is T factor of the TNM staging system a predominant prognostic factor in pathologic stage I non-small cell lung cancer? J Thorac Cardiovasc Surg 1993;106: Macchiarini P, Fontanini G, Hardin MJ, et al. Blood vessel invasion by tumor cells predicts recurrence in completely resected T1N0M0 non-small cell lung cancer. J Thorac Cardiovasc Surg 1993;106: Takise A, Kodama A, Shimosato Y, Watanabe S, Suemasu K. Histopathologic prognostic factors in adenocarcinoma of the lung periphery less than 2 cm in diameter. Cancer 1988;61: D Amico TA, Massey M, Herndon JE II, Moore MB, Harpole DH. A biological risk model for stage I lung cancer: immunohistochemical analysis of 408 patients with the use of ten molecular markers. J Thorac Cardiovasc Surg 1999;117: D Amico TA, Aloia TA, Herndon JE, et al. Molecular biologic substaging in patients with stage I non-small cell lung cancer: risk stratification according to gender and histologic subtype. Ann Thorac Surg 2000;69: Harpole DH Jr, Richards WG, Herndon JE II, Sugarbaker DJ. Angiogenesis and molecular biologic substaging in patients with stage I non-small cell lung cancer. Ann Thorac Surg 1996;61: Harpole DH Jr, Herndon JE II, Wolfe WG, Iglehart JD, Marks JR. A prognostic model of recurrence and death in stage I

5 1148 D AMICO ET AL Ann Thorac Surg LUNG CANCER METASTASES 2001;72: non-small cell lung cancer utilizing presentation, histopathology, and oncoprotein expression. Cancer Res 1995;55: Kwiatkowski DJ, Harpole DHH Jr, Godleski J, et al. Molecular pathologic substaging in 244 stage I non-small cell lung cancer patients: clinical implications. J Clin Oncol 1998;16: Strauss GM, Kwiatowski DJ, Harpole DH, Lynch TJ, Skarin AT, Sugarbaker DJ. Molecular and pathologic markers in stage I non-small cell carcinoma of the lung. J Clin Oncol 1995;13: Walker-Daniels J, Coffman K, Azimi M, et al. Overexpression of the EphA2 tyrosine kinase in prostate cancer. Prostate 1999;41: Harrington KJ, Syrigos KN. The role of E-cadherin-catenin complex: more than an intracellular glue. Ann Surg Oncol 2000;7: Salden M, Splinter TA, Peters HA, et al. The urokinase-type plasminogen activator system in resected non-small cell lung cancer. Ann Oncol 2000;11: Pedersen H, Brunner N, Francis D, et al. Prognostic impact of urokinase, urokinase receptor, and type 1 plasminogen activator inhibitor in squamous and large cell lung cancer tissue. Cancer Res 1994;54: Aloia TA, Massey MF, Joshi MM, Volmer RT, Harpole DH, D Amico TA. Molecular biologic markers predict the development of cerebral metastases in patients with stage I non-small cell lung cancer. Surg Forum 1999;50: Zhang H, Singer B. Recursive partitioning in the health sciences. New York: Springer-Verlag, NY Burt M, Wronski M, Arbit E, Galicich JH. Resection of brain metastases from non-small cell lung carcinoma. Result of therapy. J Thorac Cardiovasc Surg 1992;103: Patchell RA, Tibbs PA, Walsh JW, et al. A randomized trial of surgery in the treatment of single metastases to the brain. N Engl J Med 1990;322: Auperin A, Arriagada R, Pignon JP, et al. Prophylactic cranial irradiation for patients with small cell lung cancer in complete remission. N Engl J Med 1999;341: Stuschke M, Eberhardt W, Pottgen C, et al. Prophylactic cranial irradiation in locally advanced non-small cell lung cancer after multimodality treatment: long-term follow-up and investigations of late neuropsychologic effects. J Clin Oncol 1999;17: DISCUSSION DR JOE B. PUTNAM, JR (Houston, TX): Doctor D Amico, thank you for the opportunity to review this paper and the slides in advance. Cigarette smoking has created a worldwide epidemic of lung cancer. This deadly disease will affect over 160,000 citizens in the United States and countless millions worldwide. The current treatment paradigms include surgery, chemotherapy, radiation, or combinations of these treatments. These may be used as solitary, neoadjuvant, or adjuvant techniques. The surgeon s limitations, as we all know, in treating lung cancer are all related in applying mechanical techniques to solving a biological problem. These anatomic or clinical techniques will remain inconsistent techniques for clinical staging. Doctor D Amico and colleagues, in this paper and through their focus on molecular markers, have begun the careful dissection of the spectrum of molecular events in the initiation and propagation of lung cancer. Patients with solitary brain metastases represent a unique clinical and biological population to study. In this paper, the authors examined the molecular characteristics of the primary tumor and hypothesize that their findings will allow physicians to predict brain metastases; however, primary tumors are composed of biologically and molecularly different populations. This heterogeneity poses several questions. Is there a difference between adenocarcinoma histology, squamous cell histology, and other non-small cell histologies in the development of brain metastases based upon your model? As primary tumors are heterogeneous, did you examine the brain metastases themselves in those patients who underwent resection of these cerebral metastases in an attempt to compare the molecular characteristics of the metastases with the primary? Thirdly, were you able to correlate serum markers with your histologic examinations? And finally, do you recommend at this time prophylactic cranial irradiation in patients with a particular molecular pattern? I certainly await your prospective study confirming these observations and applaud your leadership in further refining the molecular staging of lung cancer. Thank you. DR D AMICO: Thank you, Dr Putnam. Regarding the issue of histologic subtype and risk, we have published data regarding the molecular markers of the patients with adenocarcinoma versus squamous cell carcinoma; there are four molecular markers that are distinctly different in these tumors. We have done a similar analysis of men and women, and there are three markers between men and women that show that these tumors are distinctly different. So, there is no question that there is a heterogeneity both in histologic subtype, and in other demographic presentations, that demonstrate that these tumors are different and that the prognosis of the tumor can be directly related to the molecular pedigree. In terms of looking at brain metastases from the primary tumor to the brain, we have done that study in this population of patients. What we have demonstrated is that the molecular marker expression in the metastases is similar but not exactly the same as the primary tumor, and the factors that are overexpressed in the metastases include E-cadherin, angiogenesis (factor VIII), and p53. We are using this data to construct a paradigm of the metastatic process to understand how the interrelationship of these three molecular markers may lead to the metastatic process. But we found it interesting that the primary tumor and the metastases are not identical in all patients. In terms of serum markers, we have also identified three serum markers that are linked to early recurrence and to death, but we have not developed a serum marker analysis to distinguish between distant metastases and isolated cerebral metastases only, and we are in the process of publishing that data. And finally, our recommendation for isolated cerebral metastases prophylaxis using PCI would await a prospective study. I am aware that at several institutions PCI is offered to patients after chemotherapy for nonresectable patients, and in multiple studies, the risk in non-small cell cancer of developing cerebral metastases after chemotherapy is in the range of 20% to 40%, similar to our study. So these investigators have identified a high-risk group of patients in which PCI may be effective. I do not think that we are ready to take stage I patients without a prospective study and treat them with PCI, but hopefully, someday, we will move toward a more molecular understanding and treatment of tumors rather than surgery and expectant observation.