Increased levels of chromogranin A (CgA), a protein secreted by many neuroendocrine cells, have been detected in sera of patients with neuroendocrine tumors or renal, hepatic, or heart failure. In patients with heart failure, serum CgA correlates with tumor necrosis factor-α (TNF) and soluble TNF receptors (sTNF-Rs), with important prognostic implications. The prognostic value of CgA and sTNF-Rs was investigated in advanced nonsmall cell lung cancer (NSCLC), a histologically heterogeneous group of tumors that may undergo neuroendocrine differentiation.
CgA and sTNF-Rs were analyzed in the sera of 88 patients with NSCLC before chemotherapy by enzyme-linked immunoadsorbent assay (ELISA) and in tumors by immunohistochemistry.
Thirteen percent of patients had CgA values greater than the highest value observed in normal subjects (distribution range, 9–724 ng/mL and 28–196 ng/mL, respectively). Immunohistochemical studies showed no correlation between CgA expression in tumors and serum levels. Conversely, circulating CgA was associated with worse Eastern Cooperative Oncology Group (ECOG) performance status (PS) (P = .0005), more advanced stage (P = .042), and survival, with CgA being an independent prognostic factor of poor outcome (hazards ratio [HR] 1.31 for 100 ng/mL increase; 95% confidence interval [95% CI], 1.08–1.60 [P = .0071]). sTNF-R1 and sTNF-R2 were also associated with ECOG PS (P = .0001 and P = .02, respectively). sTNF-Rs was weakly correlated with circulating CgA (r = 0.39 for TNF-R1 and r = 0.40 for TNF-R2), suggesting a regulatory link between sTNF-Rs and CgA secretion.
Chromogranin A (CgA) is an acidic glycoprotein belonging to a family of regulated secretory proteins stored in the dense core granules of the adrenal medulla and of many other neuroendocrine cells and neurons.1–3 Detection of CgA in tumor tissue is widely used for histopathologic analysis of neuroendocrine tumors as well as for assessing neuroendocrine differentiation of nonneuroendocrine tumors.4–6 For example, focal expression of CgA antigen in tumor tissues has been used to assess neuroendocrine differentiation in breast, prostate and nonsmall cell lung cancer (NSCLC).7–12 Moreover, circulating CgA is a sensitive marker for diagnosis of various types of neuroendocrine tumors and an independent prognostic indicator of mortality in patients with carcinoid tumors.3 Elevated levels of circulating CgA have been detected in the blood of patients with hepatic failure or renal failure,3, 13 and in patients with atrophic gastritis (type 1) and those receiving treatment with proton pump inhibitors.14–16 In addition, we have recently shown that circulating CgA is markedly increased in patients with heart failure, depending on the severity of the diseases, and that it correlates with tumor necrosis factor-α (TNF) and soluble TNF receptors (sTNF-Rs) with important prognostic implications.17, 18 This suggests that CgA may be associated not only with neuroendocrine tumors, but also with other inflammatory diseases or disabling medical conditions characterizing organ failure.
In this study, we analyzed CgA and soluble TNF receptors in patients with NSCLC, a disabling disease with variable clinical behavior. Because it has been previously shown that chemotherapy with cisplatin and other chemotherapeutic agents is associated with an increase of plasma CgA,19, 20 we analyzed CgA in serum samples before chemotherapy. We demonstrated that the circulating levels of CgA are increased in a subpopulation of patients with NSCLC, despite a lack of detectable CgA in tumor tissue, and are correlated with Eastern Cooperative Oncology Group (ECOG) performance status (PS), extension of the disease, and, weakly, with sTNF-Rs. Furthermore, we demonstrated that CgA is an independent negative prognostic indicator of mortality.
MATERIALS AND METHODS
Patients with histologically proven, untreated, and locally advanced unresectable or metastatic NSCLC were eligible for participation in the study. Additional eligibility criteria included age >18 years, a white blood cell count ≥3.5 × 109/L, a platelet count ≥100 × 109/L, hemoglobin of ≥9 g/dL, an absolute granulocyte count >2.0 × 109/L, bilirubin <1.5-fold the upper limit of normal (ULN), prothrombin time or activated partial thromboplastin time <1.5-fold greater than controls, alanine aminotransferase or aspartate aminotransferase <3-fold the ULN (elevated to 5-fold in patients with known hepatic metastases), and a calculated creatinine clearance rate of >45 mL/min.
Patients were ineligible if they had evidence of prior/concurrent malignancy other than in situ carcinoma of the cervix or adequately treated basal cell carcinoma of the skin or other malignancies treated ≥5 years previously without evidence of recurrence.
Patients with clinical or conventional laboratory evidence suggestive of hypertension; rheumatoid arthritis; or heart, hepatic, or renal failure, were excluded from this study.
TNM classification was based on the new international staging system.21 The staging procedures included clinical examination, computed tomography (CT) scan of the chest and abdomen, or abdominal ultrasonography. Brain CT scan and bone scanning were performed according to symptom presentation or other clinical indications.
After blood collection patients were treated according to their stage, ECOG PS, and comorbidities with best supportive care, chemotherapy, or chemoradiotherapy. Follow-up was performed every 2 months with baseline staging procedures.
The study protocol was revised and accepted by the local ethics committee and appropriate written informed consent was obtained from each patient before entering the study. Recommendations of the Declaration of Helsinki for biomedical research involving human subjects were also followed.
Blood samples of age-matched healthy donors were collected as controls.
Blood Collection and Storage
Five mL of peripheral blood was collected before any treatment in a sterile testtube (without anticoagulants) and centrifuged at 3000g for 10 minutes at 4°C. Serum was stocked in 0.5-mL aliquots in cryovials and stored at −80°C until measurement was performed.
CgA, sTNF-R1, and sTNF-R2 Assays
CgA assay was performed by sandwich enzyme-linked immunoadsorbent assay (ELISA) based on an anti-CgA monoclonal antibody (B4E11) and a rabbit-polyclonal anti-CgA antiserum. Monoclonal antibody (MoAb) B4E11 is a mouse immunoglobulin (Ig) G1 that recognizes an epitope of CgA corresponding to residues 68–70 of human CgA.22 The assay was performed as previously described.18 sTNF-R1 and sTNF-R2 were measured by sandwich ELISA using commercial kits from R&D Systems (Abingdon, UK)
Immunohistochemical Analysis of CgA
Immunohistochemical analysis of CgA was performed on formalin-fixed and paraffin-embedded surgical specimens and bronchial biopsies (3–4 μm sections). Sections were deparaffinized in xylene and rehydrated through graded alcohol steps. Each slide was treated with 0.3% hydrogen peroxide in 95% methanol for 7 minutes to quench endogenous peroxidase activity and washed with water for 10 minutes. Antigen retrieval was performed by microwaving in citrate buffer (pH 6.0) (700 watts for 5 minutes for 4 cycles). After gradual cooling to room temperature, the slides were washed with phosphate-buffered saline (PBS) for 5 minutes and blocked with universal serum blocking reagent (Vectastain Universal Quick Kit; Vector Laboratories, Burlingame, Calif) for 10 minutes at room temperature. Detection of CgA was performed using the monoclonal antibody LK2H10 (BioGenex, San Ramon, Calif) diluted 1:100, followed by detection using the Vectastain Kit. CgA immunostaining was considered positive if cytoplasmic staining was present in >1% of tumor cells. All specimens were evaluated independently by 2 pathologists without knowledge of pathologic or clinical data concerning the patients under investigation. A minimum of 10 fields was screened (×40 objective lens).
Data are presented as the median value, interquartile range (Q1-Q3), and minimum and maximum values for continuous variable, absolute and relative frequencies for categoric variables. The distributions of the serum levels of CgA, TNF-R1, and TNF-R2 were compared between cases and controls by means of the Wilcoxon test. The same test (or the Kruskal-Wallis test in case of >2 categories) was applied to compare the same parameters in groups defined by clinical characteristics, such as stage and histologic type. A nonparametric test for trend was used to assess the presence of a monotonic trend of CgA, TNF-R1, and TNF-R2 serum levels according to PS.
The capability of CgA serum levels to discriminate between cases and controls was assessed by means of Receiver Operator Characteristics (ROC) curves. The relations between serum levels were assessed using Spearman correlation coefficients, adjusting for group (controls and NSCLC patients). Kaplan-Meier curves for overall survival were compared using the Mantel-Cox version of the log-rank test.
Overall survival was defined as the time from the date of pathologic diagnosis to death for any cause; patients known to still be alive were censored at the time of the analysis. All P-values are 2-sided and statistical significance was defined as P < .05. Analyses were performed using SAS software (version 8.20; SAS Institute Inc, Cary, NC).
Eighty-eight consecutive patients with untreated and locally advanced unresectable or metastatic NSCLC as well as 50 controls were included in the study. The median age was 62.5 years (range, 21–80 years) and 83% of the patients were males. ECOG PS was generally good (0 in 18%, 1 in 67%, and 2–1 in 15% of patients). All patients had advanced NSCLC (33% with stage IIIB disease and 67% with stage IV disease), with bone representing the main metastatic site. Histologic types were adenocarcinoma (25%), squamous carcinoma (31%), large cell carcinoma (11%), and NSCLC not otherwise specified (cytologic diagnosis) (33%).
Seven patients were treated with proton pump inhibitors. Patients with advanced disease (n = 60) were treated with cisplatin-based chemotherapy or single-agent (gemcitabine, vinorelbine, or taxanes) therapy. Eight patients received supportive care only due to the general conditions and/or comorbidities. Eighteen patients (30%) obtained a partial response, whereas stable and progressive disease rates were 32% and 38%, respectively, with no important differences noted in response between different chemotherapy regimens. Seven patients of 21 with locally advanced NSCLC without pleural effusion at the time of diagnosis or disease progression after 2 to 3 cycles of chemotherapy received radiotherapy (60–65 grays).
CgA Serum Levels
CgA levels were measured in sera samples obtained before treatment with chemotherapeutic agents. The distribution of serum CgA levels in patients is reported in Table 1. The median value of CgA in NSCLC patients and controls were 70.4 ng/mL (Q1–Q3: 37.9–114.6 ng/mL) and 77.4 ng/mL (Q1–Q3: 57.7–99.9 ng/mL), respectively, with no statistically significant difference noted (Wilcoxon test, Z = 0.88; P = .377). Moreover, CgA values were not differently distributed by histologic types: excluding patients with unspecified NSCLC, median values of CgA were 59.2 ng/mL (Q1–Q3: 35.2–85.6 ng/mL) in patients with adenocarcinoma, 80.0 ng/mL (Q1–Q3: 41–128.6 ng/mL) in patients with squamous carcinoma, and 82.1 ng/mL (Q1–Q3: 33.7–124 ng/mL) in patients with large cell carcinoma (Kruskal-Wallis test, χ2 = 1.53; 2 df [P = .465]).
Table 1. Distribution of Serum CgA Levels in the Study Population
Serum CgA levels significantly increased with worsening of ECOG PS: 37.7 ng/mL (Q1–Q3: 27.2–68.6 ng/mL) in patients with a PS of 0, 76.3 ng/mL (Q1–Q3: 43.6–119.2 ng/mL) in patients with a PS of 1, and 102.8 ng/mL (Q1–Q3: 55.8–259.4 ng/mL) in patients with a PS ≥ 2 (nonparametric test for trend, χ2 = 12.03; 1 df [P = .0005]).
CgA increased also depending on the extension of disease. The median CgA values of patients with stage IIIB and stage IV disease were 44.9 ng/mL (Q1–Q3 29.2–85.6 ng/mL) and 82.5 ng/mL (Q1–Q3: 47.1–119.2 ng/mL), respectively (Wilcoxon test, Z = −2.03 [P = .042]). It is noteworthy that the group with the highest values of CgA did not correspond to patients treated with proton pump inhibitors (n = 7 patients).
To examine the discriminating capability of CgA serum levels, we performed ROC analyses contrasting controls and patients affected by NSCLC. The serum CgA levels failed to differentiate between controls and NSCLC group (AUC = 0.457), as shown in Figure 1.
sTNF-R1 Serum Levels
The distribution of serum TNF-R1 levels in patients and controls is reported in Table 2. The values in NSCLC patients were significantly higher than in controls (Wilcoxon test, Z = −9.12 [P < .0001]). Median values were 0.72 ng/mL (Q1–Q3: 0.63–0.87 ng/mL) in the control group and 1.74 ng/mL (Q1–Q3: 1.33–2.52 ng/mL) in the patient group. In the NSCLC group, the values were not differently distributed according to histotype (Kruskal-Wallis test, χ2 = 1.72; 2 df [P = .422]) or stage (Wilcoxon test, Z = 2.12 [P = .145]), whereas they significantly increased with worsening of ECOG PS (nonparametric test for trend, χ2 = 14.37; 1 df [P = .0001]).
Table 2. Distribution of Serum sTNF-R1 and sTNF-R2 Levels in the Study Population
The distribution of serum TNF-R2 levels in patients and controls is reported in Table 2. The values in NSCLC patients were significantly higher than in controls (Wilcoxon test, Z = −4.11 [P < .0001]). The median values were 1.68 ng/mL (Q1–Q3: 1.46–1.92 ng/mL) in the control group and 2.17 ng/mL (Q1–Q3: 1.72–3.25 ng/mL) in the patient group. In the NSCLC group, values were not differently distributed according to histotype (Kruskal-Wallis test, χ2 = 4.14; 2 df [P = .126]) or stage (Wilcoxon test, Z = −1.61 [P = .107]), whereas they significantly increased with worsening of ECOG PS (nonparametric test for trend, χ2 = 5.38; 1 df [P = .020]).
Correlation Among Variables
Analysis of correlation among variables showed that sTNF-R1 and sTNF-R2 correlated closely (r = 0.82; 95% confidence interval [95% CI], 0.75–0.86). Both receptors weakly correlated with CgA levels (r = 0.39 [95% CI, 0.24–0.52] for TNF-R1 and r = 0.40 [95% CI, 0.25–0.53] for TNF-R2). For the other assessed variables such as gender, age, weight loss, and comorbidities (diabetes mellitus, arteriosclerosis, systemic arterial hypertension, deep vein thrombosis), no statistically significant association was detected.
CgA Expression in Tumor Tissues
The expression of CgA antigen in tumor tissue sections, obtained from bronchial biopsies or from surgical specimens of 10 patients with different CgA serum levels, was analyzed using the anti-CgA monoclonal antibody LK2H10. No correlation between serum levels and tissue expression was observed (Table 3). Of note, no expression of CgA was observed in a section corresponding to the patient with the highest serum level of CgA (724 ng/mL). Moreover, 2 patients with focal neuroendocrine differentiation, as suggested by the presence of 2% and 7%, respectively, CgA-positive cells in tissue sections, had low serum CgA levels (Table 3) (Fig. 2). These results suggest that tumor tissues are not the major source of circulating CgA in NSCLC patients.
Table 3. CgA Serum Levels and CgA Expression in Tumor Tissues
Commercial CgA-ELISA kit (available from Dakopatts, Copenhagen, Denmark).
At a median follow-up time of 44.3 months, 86 patients (97.7%) had died. The median survival of the entire group was 7.5 months (Q1–Q3: 3.1–16.4 months). Univariate analysis showed that CgA, TNF-R1, TNF-R2, ECOG PS, and stage were all associated with survival (Table 4). Given that all these parameters were associated with a higher probability of dying, they were included in a multivariate model. Multivariate analysis indicated that CgA is a relevant independent prognostic factor of a poor outcome (hazards ratio [HR], 1.31; 95% CI, 1.08–1.60 [P = .0071]) in addition to TNM and ECOG PS (Table 4).
Table 4. Univariate and Multivariate Analysis for Predictive Factors of Mortality in 88 Patients With NSCLC (Cox Proportional Hazards Model)
To describe the survival in relation to CgA and soluble TNF-Rs levels, Kaplan-Meier curves were drawn using the following cutoffs: CgA < 77 ng/mL (corresponding to the median value of the distribution in controls), CgA > 156 ng/mL (derived from ROC curves as the value with the best discriminative capacity), TNF-R1 ≥ 1.5 ng/mL, and TNF-R2 ≥ 2.0 ng/mL (from ROC curves) (Fig. 3). The median survival was 12.0 months in patients with CgA levels ≤ 77 ng/mL, 5.8 months in those with CgA > 77 and ≤ 156 ng/mL, and 2.1 months in those with CgA > 156 ng/mL. Considering the group with CgA ≤ 77ng/mL as a reference, both the group with a CgA level > 77 and ≤ 156 ng/mL (HR, 2.0; 95% CI, 1.24–3.31 [P = .0045]) and the group with CgA > 156 ng/mL (HR, 7.1; 95% CI, 3.5–14.2 [P < .0001]) demonstrated a higher probability of dying.
CgA Levels and Response to Chemotherapy
CgA levels were detected before chemotherapy. At follow-up, among the 60 patients treated with chemotherapeutic drugs we found 18 (30.0%) partial responses, 19 (31.7%) with stable disease and 23 (38.3%) with disease progression. When we compared the response of patients with CgA levels greater than the median of normal values (n = 23) with those of patients with lower levels (n = 37), we found 3 partial responses (13%), 7 cases of stable disease (30.4%), and 13 patients with disease progression (56.5%) in the first group, and 15 partial responses (40.5%), 12 cases of stable disease (32.4%), and 10 patients with disease progression (27%) in the second group, with no statistically significant difference (P = .38).
The results of the current study demonstrate that serum CgA and sTNF-Rs are increased in a subpopulation of NSCLC, independently from CgA expression in tumors, before administration of cisplatin or other chemotherapeutic drugs. Although the median values of CgA in normal subjects and in patients were similar (mean of 77 ng/mL and 70 ng/mL, respectively) the distribution ranges were different (28–196 ng/mL and 9–724 ng/mL, respectively). Thirteen percent of patients had CgA values greater than the highest value observed in normal subjects, whereas few patients had CgA levels lower than the lowest value observed in normal subjects, possibly due to reduced production or increased degradation of CgA. Increased levels of CgA have been observed also by other investigators in a subset of NSCLC patients.23, 24
The source of serum CgA and sTNF-Rs in patients with NSCLC and the pathophysiologic meaning of increased levels are unknown. Abnormal levels of circulating CgA have been detected in various nonneoplastic conditions, such as heart failure, renal failure, hepatic failure, and hypertension.3, 25 We exclude that high levels of CgA in NSCLC patients were related to these conditions, because these patients were excluded from our study. Furthermore, the majority of patients in the current study (>92%) were not receiving treatment with proton pump inhibitors, which are known to induce CgA release in circulation,14–16 except for a small number of subjects (n = 7) that in any case did not correspond to patients with the highest levels.
In patients with prostate cancer with neuroendocrine differentiation, a correlation between CgA-producing cells in tumor tissue sections and CgA serum levels has been observed.26 This suggests that the tumor itself may be a source of circulating CgA. In NSCLC, the experimental data so far available appears to argue against this hypothesis. Previous work showed that approximately 30% to 50% of patients with NSCLC have elevated serum levels of CgA compared with controls.23 However, other work performed by many groups demonstrated that only 0% to 19% of patients with NSCLC have focal expression of CgA in tumor tissues, as judged from the results of immunohistochemical studies.27–29 Another study, conducted in 90 patients with advanced NSCLC, showed that CgA is expressed in tumors even less frequently.30 It would appear, therefore, that patients with increased serum levels of CgA are more common than patients with tissue expression of this protein. Accordingly, when we analyzed the expression of CgA in tumor tissues sections of patients with different levels of circulating CgA, we observed CgA expression in 2 samples with relatively low levels of circulating CgA, but not in samples from patients with high serum levels. Of note, negative samples included those from patients with very high levels of circulating CgA. Thus, no correlation between tissue expression, as judged from mAb LK2H10 staining of tissue sections, and serum levels seems to occur in NSCLC. Interestingly, circulating CgA was associated with worse patient ECOG PS and more advanced NSCLC, suggesting that increased circulating levels of CgA are more likely related to worse general patient conditions than to neuroendocrine differentiation.
A previous study provided evidence to suggest that circulating CgA bears important information related to prognosis for NSCLC patients.24 Accordingly, we observed that patients with relatively high levels of CgA had worse survival. However, we also found that the prognostic value of CgA was independent of ECOG PS and stage.
The observation that serum CgA is an indicator of mortality in NSCLC independent from stage and ECOG PS raises the question as to whether CgA could contribute to tumor progression or is just a measurable surrogate marker of conditions associated with worse outcome. We have shown previously that local expression of CgA in mammary adenocarcinoma or lymphoma tumors delays, rather than promotes, the growth of tumors in mouse models.31 However, it is difficult to speculate on the effect of CgA on tumor progression in NSCLC patients on the basis of these results (obtained with tumor cells engineered to secrete CgA) because the cellular source of circulating CgA, its local concentration, and its posttranslational modifications are likely different in patients. Interestingly, CgA was associated with both sTNF-R1 and sTNF-R2. Although sTNF-Rs were not significant independent prognostic factors of poor survival, they were also significantly associated with ECOG PS. This raises again the question as to whether sTNF-Rs in concert with CgA play a pathophysiologic role in NSCLC. It is believed that endogenous TNF can regulate tumor physiology by affecting cancer cell proliferation and survival, inflammatory and immune responses, angiogenesis, vascular function, and stroma formation.32–34 Because of its complex activity, TNF may contribute to reduced or even to increased tumor growth,35, 36 likely depending on its site of production, its levels in tumor tissues, and its persistence, as well as the presence or absence of other cytokines in tumors. Soluble TNF-Rs can act as inhibitors of TNF interaction with membrane receptors, or conversely as molecules capable of prolonging its activity, eg, by stabilizing its bioactive homotrimeric structure.33, 37 Thus, it is very difficult to speculate on the overall effects of both sTNF-Rs and CgA on tumor progression in patients with NSCLC. However, it should be kept in mind that the effects of TNF and CgA are most likely not limited to tumors. Considering that 1 of the known effects of TNF is to affect vascular permeability and that CgA can inhibit TNF-induced vascular leakage in normal vessels,38 sTNF-Rs in concert with CgA could contribute to reducing some of the potentially dangerous effects of pathologic levels of circulating TNF on the vascular system in different organs. It is noteworthy that in the present study we observed a significant correlation between circulating sTNF-Rs and CgA, as we observed previously in patients with heart failure.17, 18 This supports the hypothesis that a regulatory link exists between CgA and sTNF-Rs secretion that may contribute to control the activity of TNF at the systemic level.
Based on these considerations, we hypothesize that circulating CgA in NSCLC is not just a marker of neuroendocrine differentiation, but it reflects stress-related systemic neuroendocrine activation associated with worsening conditions, also characterized by the release of potentially dangerous cytokines in circulation.
In conclusion, we found that serum sTNF-Rs and CgA are increased in a subpopulation of patients with NSCLC, despite a lack of detectable CgA in tumor tissue. CgA correlates with PS, stage, sTNF-R1 and sTNF-R2, and is an independent prognostic indicator of mortality. The lack of correlation with CgA expression in tissues and the correlation with sTNF-Rs suggests that CgA in certain cancer patients is more likely related to the worse general patient conditions, rather than to neuroendocrine differentiation.