Nucleosomes, ProGRP, NSE, CYFRA 21-1, and CEA in Monitoring First-Line Chemotherapy of Small Cell Lung Cancer

Lung cancer is the leading cause of cancer mortality and shows high incidences in both men and women worldwide (1). About 15% to 20% of all lung cancer patients suffer from small cell lung cancer (SCLC), which is diagnosed in about half of the patients at stage of limited disease and extended disease (1, 2). In contrast to non-SCLC (NSCLC), SCLC often shows a multilocular growth pattern with an early tendency to metastasize in lymph nodes or distant organs and is frequently associated with paraneoplastic syndromes (2, 3). Further, SCLC mostly has a high sensitivity to chemotherapy and radiotherapy and reaches initial response rates of 80% to 90%. Nevertheless, in many patients treated successfully, tumor recurrence and formation of metastases in various organs have been observed in the further follow-up, which has led to the concept of preventive cerebral irradiation after the initial chemotherapeutic treatment (3, 4). When tumor progression occurs, several established cytotoxic and new biological drugs are available for the second- and third-line treatments (3–6). Although change of tumor volume often is visible by imaging techniques after several chemotherapeutic cycles, oncologic biomarkers may be helpful to earlier indicate the therapy response to optimize the individual management of the disease. Changing early the treatment strategy could then save time and costs and avoid unnecessary side effects.

In NSCLC, positron emission tomography alone and in combination with computed tomography (7–9) as well as algorithms of biomarker kinetics such as nucleosomes and cytokeratin-19 fragments (CYFRA 21-1; refs. 10–12) have already shown their potential to early indicate insufficient response to therapy sensitively and specifically. Both approaches, which mirror the cancer activity and metabolism, may also be applicable on patients with SCLC. In comparison with imaging techniques, which monitor only macroscopic alterations of the tumor mass, they also take into account the heterogeneity of the tumor tissue containing active, silent, apoptotic, and necrotic parts and respect the activity of minimal residual tumor disease (7–13). For the diagnosis of SCLC, neuron-specific enolase (NSE) and progastrin-releasing peptide (ProGRP) revealed high diagnostic sensitivity and specificity (14–17). Sensitivity was further increased when both markers were combined (15). In addition, carcinoembryonic antigen (CEA) and CYFRA 21-1 have shown to be elevated in many patients with SCLC, particularly in advanced stages (15, 17, 18). Concerning prognosis and therapy monitoring of SCLC, only few studies are available, which show some value for NSE, ProGRP, and CYFRA 21-1 (19–22) that has led to the inclusion of these markers for the use in SCLC in the recent recommendations of the National Academy of Clinical Biochemistry (23).

In the present study, we investigated serum concentrations of nucleosomes, ProGRP, NSE, CYFRA 21-1, and CEA of a homogeneous group of patients with newly diagnosed SCLC during first-line chemotherapy to analyze the power of those biomarkers for therapy monitoring and early estimation of therapy response.

Patients. In total, 128 consecutive patients with newly diagnosed SCLC under the care of the Asklepios Clinics Gauting were included in the present study. All patients were investigated initially by whole-body computed tomography, bone scan, and bronchoscopy. All patients received first-line chemotherapy without concomitant radiotherapy; most of them were treated with protocols containing carboplatin, etoposide, and vincristine and others with regimens of cisplatin and etoposide, cisplatin and topotecan, or ifosfamide, etoposide, and vindesine (Table 1).

In all patients, staging investigations were done before the start of the third cycle of chemotherapy consisting of clinical examination, whole-body computed tomography, and laboratory examinations. The response to therapy was classified according to the WHO classifications defining “partial remission” (R) as tumor reduction ≥50%, “progression” (P) as tumor increase ≥25% or appearance of new tumor manifestations, and “no change” (NC; stable disease) as tumor reduction <50% or increase <25% (24).

For evaluation 1, patients with progressive disease (P) were compared with those having remission and stable disease (R + NC); for evaluation 2, patients with remission (R) were compared with those having progressive and stable disease (P + NC).

Materials and methods. Blood samples were collected prospectively before application of the first three cycles of chemotherapy to determine the baseline values of the markers (BV1, BV2, and BV3). The samples for nucleosome determination were centrifuged at 3,000 × g for 15 min and treated with 10 mmol/L EDTA (pH 8) immediately after centrifugation. Subsequently, they were stored at −70°C and analyzed in batches containing all samples of a single patient. The details of the preanalytic handling are described in Holdenrieder et al. (25). Nucleosome fragments were determined by the Cell Death Detection-ELISA^plus of Roche Diagnostics, which was modified for its use in serum matrix (25).

Briefly, serum samples were placed into a microtiter plate; a buffer solution was added containing two monoclonal antibodies directed against DNA and histones (from the mouse clones M-CA-33 and H11-4), respectively. After incubation of this mixture with the serum sample for 2 h in which the nucleosomal complexes were specifically catched by the antibodies and the biotinylated anti-histone antibodies adhered to the microtiter plate, unbound reagents were removed and 2,2′-azino-di-3-ethylbenzthiazoline-sulfonate was added. This substrate reacted with the peroxidase-labeled anti-DNA antibodies and led to concentration-dependent color development, which was measured photometrically at 405 nm after 30 min. Nucleosome concentrations were calculated using a standard curve, which was established from plasma of healthy donors after artificially induced mixed lymphocyte reaction, with known amounts of nucleosomal DNA (25).

Correspondingly, baseline values of the oncologic biomarkers ProGRP (by ELISA; Advanced Life Science Institute; European distributor Immuno Biological Laboratories), NSE, CYFRA 21-1, and CEA (by electrochemiluminescence immunoassay on Elecsys 2010; Roche Diagnostics) were determined in serum samples before each therapeutic cycle (BV1, BV2, and BV3) at the day of sample collection without any storage procedures.

For quantification of ProGRP, a buffer solution and serum samples were placed into a microtiter plate precoated with antibodies directed against ProGRP [31-98] and incubated for 60 min. Subsequently, the plate was washed, peroxidase conjugate was added (incubation time 30 min), and after a further washing step substrate solution was added. Color development was stopped after 30 min by a stop solution and absorbance was measured photometrically at 495 nm. Finally, concentrations of ProGRP were calculated by use of a standard curve provided by the manufacturers.

Statistics. To test the association of therapy response at time of staging investigations before start of the third therapy cycle with overall survival of the patients, Kaplan-Meier curves and log-rank analyses were established for the various response groups.

Concerning the biochemical variables, the baseline values of all markers before the first, second, and third cycles (BV1, BV2, and BV3) and the percentual changes (BV1-2 and BV1-3) were considered for statistical analyses. In a first step, all biomarkers were evaluated on their power to univariately discriminate between (a) patients with progression (P) and nonprogression (R + NC; evaluation 1) as well as between (b) patients with remission (R) and nonremission (P + NC; evaluation 2) by Wilcoxon test. Clinical variables that were available in defined categories were tested by χ² test. To identify the best diagnostic biomarkers for insufficient therapeutic efficacy, receiver operating characteristic curves and corresponding areas under the curve (AUC) were calculated for all biochemical markers and for their combinations. To enable a fair comparison, all biomarker values were transformed to a normalized scale by dividing them by the median value of healthy donors and taking the logarithm of those values. This was done for (a) all markers being available before start of the third course of therapy at time of staging investigations and (b) all markers being available before start of the second course of therapy for the early estimation of therapy response. In addition, sensitivities for the detection of progression were calculated at 95% specificity for patients with response to therapy. The results of all marker combinations were further validated by the leave-one-out validation procedure. Finally, a multivariate model of all clinical and biochemical variables being available before start of the second course of therapy was calculated by logistic regression analyses using forward and backward selection.

P < 0.05 was considered statistically significant. All statistical analyses were done with software of SAS (version 8.2; SAS Institute).

The clinical characteristics of the patients are listed in Table 1. At staging investigations before start of the third cycle, 103 (81%) of the 128 patients had partial remission, 17 (13%) progression, and 8 (6%) no change of disease (Table 1). Therapy response was strongly associated with overall survival and was therefore considered as valid and reliable surrogate marker for the outcome of the therapy: Kaplan-Meier curves for overall survival showed highly significantly differences for patients with remission (R), no change (NC), and progression (P) at time of staging investigations before start of the third therapy cycle (P < 0.0001, log-rank test) with median survival times of 15.8 (R), 8.5 (NC), and 2.2 (P) months, respectively (Fig. 1A). When patients with remission and stable disease were compared with patients having progressive disease (evaluation 1), median survival times were 13.0 (R + NC) and 2.2 (P) months (P < 0.0001); when patients with remission were compared with patients having progressive and stable disease (evaluation 2), median survival times were 15.8 (R) and 4.4 (P + NC) months (P < 0.0001; Fig. 1B and C).

Concerning the biomarker concentrations in various response groups to chemotherapy, patients with remission often had considerably lower baseline values before the various treatment cycles (BV1, BV2, and BV3) and stronger decreases (BV1-2 and BV1-3) than patients with no change and even more than patients with progressive disease (Table 2; Fig. 2).

Although pretherapeutic concentrations of nucleosomes were comparable in both groups (BV1), they were significantly distinguished from each other by the baseline value before cycle 2 (BV2) and cycle 3 (BV3) of nucleosomes. Additionally, the courses of nucleosomes showed stronger decreases from cycles 1 to 2 (BV1-2) in responsive patients compared with nonresponsive ones (Table 2; Fig. 2). The fractions of patients with lower nucleosome values for BV2 compared with pretherapeutic values (BV1) in the various groups were 78% (R), 57% (NC), and 55% (P) and for BV3 were 84% (R), 100% (NC), and 60% (P), respectively. Interestingly, single values at BV2 and BV3 showed similar or greater differences between the response groups than kinetics did.

Similarly, ProGRP discriminated clearly between the groups concerning the baseline value before cycle 2 (BV2) and cycle 3 (BV3) as well as the kinetics from cycles 1 to 2 (BV1-2) and cycles 1 to 3 (BV1-3), whereas NSE showed differences only for BV3 but not for BV2 and the kinetics (BV1-2 and BV1-3).

In addition, CYFRA 21-1 and CEA discriminated significantly between response groups concerning BV1, BV2, and BV3 as well as CEA for the courses from cycles 1 to 3 (BV1-3; Table 2; Fig. 2). Regarding clinical factors, performance status (P < 0.001) and stage (limited versus extended disease; P = 0.027) showed already pretherapeutically predictive potential, but age, sex, and weight loss did not.

When focusing on biomarkers being available at time of staging investigations before the third treatment course (BV3; BV1-3), nucleosomes, CYFRA 21-1, NSE, and ProGRP indicated best the therapy response as single markers (Table 3). The combination of various marker combinations further increased the diagnostic sensitivity concerning the detection of progression (evaluation 1: P versus R + NC) and nonremission (evaluation 2: R versus P + NC), respectively (Table 3).

To test the relevance of the biomarkers for the early estimation of the therapy response, we then focused on the information being already available before start of the second course of the treatment. Among all relevant markers, the baseline values (BV2) of nucleosomes, ProGRP, and CYFRA 21-1 most efficiently identified those patients with insufficient therapy response reaching AUCs of 81.8%, 71.3%, and 74.9% in receiver operating characteristic curves of evaluation 2 and very similar results in evaluation 1, respectively (Table 4; Fig. 3). Combinations of nucleosomes with ProGRP (AUC 84.1%), CYFRA 21-1 (AUC 82.5%), and NSE (AUC 83.6%) further improved the diagnostic power particularly in the high specificity range and yielded sensitivities of 47.1%, 35.3%, and 35.3%, respectively, at 95% specificity concerning the differentiation of responsive from nonresponsive patients (evaluation 2; similar or even better results in evaluation 1; see also Table 4). Results of marker combinations were confirmed by leave-one-out cross-validation with only slightly lower AUCs in receiver operating characteristic curves (Tables 3 and 4).

When multivariate models were established including all clinical variables (performance score, weight loss, stage, age, and sex) and all biochemical variables being available before start of the second course of therapy, using logistic regression analyses in backward and forward selection for both evaluation modalities, the variables performance score (evaluation 1: P = 0.0029; evaluation 2: P = 0.0117) and nucleosomes BV2 (evaluation 1: P = 0.0220; evaluation 2: P = 0.0020) were found to be the only markers that independently indicated therapy response. The corresponding odds ratio estimates were 3.66 (95% confidence limits, 1.33-10.10) for each performance score step and 2.72 (95% confidence limits, 1.45-5.10) for each doubling of nucleosome values (evaluation 2).

Along with the development of new drugs for cancer diseases, there is a growing need for diagnostic tools to estimate the prognosis of the patient, to monitor the treatment course, and to early detect the response to therapy, which would help to optimize the disease management on an individual basis. In patients with lung cancer, several oncologic biomarkers, such as NSE and ProGRP in SCLC as well as CEA and CYFRA 21-1 in NSCLC, have shown considerable diagnostic and prognostic potential or have proved to be useful for the monitoring of systemic treatment (14–23). Concerning the early estimation of therapy response, which may be most helpful already after the application of only one course of cytotoxic therapy, the combination of the biomarkers nucleosomes and CYFRA 21-1 have shown to be valuable in patients with advanced NSCLC treated by chemotherapy (10, 11): in a prospective study including 311 patients with advanced NSCLC undergoing first-line chemotherapy, we recently found that both markers were able to detect insufficient response to therapy with 100% specificity in about one-third of all progressive patients when determined during the first treatment course. When specificity was lowered to 90%, sensitivity rose up to 55% (11). These data were confirmed by a second study on 161 patients with recurrent NSCLC receiving second-line chemotherapies (26). Because in SCLC only few information have been available on the relevance of biomarkers for the therapy monitoring thus far, we here analyzed serum concentrations of nucleosomes, ProGRP, NSE, CYFRA 21-1, and CEA of a homogeneous group of patients with newly diagnosed SCLC during first-line chemotherapy to test their power for therapy monitoring and the early estimation of therapy response.

Most importantly, therapy response at time of first staging investigations before start of the third therapy cycle was shown to be a valid and reliable surrogate marker for therapy outcome, as it strongly correlated with overall survival. The clinical characteristics performance score and stage, which are currently used for therapy stratification, were clearly associated with therapy response. Although the grade of significance was particularly high for the performance score, it was not able to appropriately identify patients who will not profit from chemotherapy alone by this criterion as some patients with performance score > 1 still achieved remission during chemotherapy.

Among biomarkers, only pretherapeutic baseline values (BV1) of CYFRA 21-1 and CEA were different in responsive and progressive patients. Before start of the second and third cycles of the treatment (BV2 and BV3), concentrations of nucleosomes, ProGRP, CYFRA 21-1, and CEA significantly discriminated between both response groups and also NSE (BV3) did. In consequence, the courses of nucleosomes and ProGRP showed significantly stronger decreases from cycle 1 to cycle 2 in patients with remission compared with those having progressive disease, whereas, for NSE and CYFRA 21-1, the level of significance was not reached.

Results of the small cell biomarkers ProGRP and NSE are in accordance with the hypothesis that marker levels would decrease if tumor mass and/or activity were efficiently reduced by chemotherapy. This decrease would be less pronounced in case of insufficient therapy. However, it is remarkable that the most significant difference between the response groups was already observed before start of the second therapy cycle. Notably, courses of single responsive patients revealed that, for example, the levels of ProGRP completely decreased from >500 pg/mL to the reference range within this short time frame. The fact of decreasing ProGRP and NSE levels during effective chemotherapy has generally been confirmed by the findings of other groups as well (19–22). However, those have not analyzed the courses in this systematic and close meshed way and therefore could not find that differences between responsive and progressive patients are present already after one cycle of chemotherapy.

Similarly, the reduction of cellular turnover during effective chemotherapy has been proposed to lead to decreasing values of cell death markers such as circulating nucleosomes. Comparable results have also been observed in patients with NSCLC and other epithelial tumors, suggesting that this may be less a tumor-specific but more a general phenomenon (10, 26–31). Whether the lower concentrations of nucleosomes in patients with therapy response are due to the reduced release from tumor and other blood cells or the more rapid removal from blood remains still speculative up to now. In line with other studies on various epithelial cancer types (32–36), our study confirms that CYFRA 21-1 not only has diagnostic, predictive, and prognostic value in NSCLC but can also be considered as panmarker for many tumor entities. Also in SCLC, CYFRA 21-1 showed characteristic changes during therapy and indicated therapy response pretherapeutically and at time before the second and third cycles. One explanation could be that CYFRA 21-1 is ubiquitously found in many body tissues and has been described as an apoptotic cell death marker in in vitro studies as well (32, 37). Therefore, its serum levels may rely on similar mechanisms of release from cells and removal from circulation as nucleosome levels do (38). CEA as a further panmarker is not related to any histologic subtype and showed differences in serum levels of responsive and progressive SCLC patients as well.

When patients were categorized into therapy response groups, those with remission and progression at time of staging investigations could be easily separated into responders and nonresponders. However, patients with no change of disease could dispose a tumor reduction not sufficient enough to be classified as “partial remission” or a tumor growth, which was not considered as “progression” yet. Correspondingly, survival curves of patients with “stable disease” ranged in between those with remission and progression. Depending on the clinical intention, it may be desirable to identify only those patients who suffer from progressive disease to early change their treatment plan. Alternatively, it would be helpful to find patients with the expected quick volume reduction of this chemosensitive tumor (remission) corresponding with the best prognosis to intensify the treatment in the other patients without remission. Therefore, we conducted two different evaluations of biochemical markers for both clinical questions. For most variables, the results in both evaluations were quite comparable with minor variations.

At time of staging investigations, nucleosomes, ProGRP, NSE, CYFRA 21-1, and CEA (BV3) correlated well with the response to therapy as single markers. Interestingly, the absolute concentrations of these biomarkers gave better information than their kinetics from cycles 1 to 3. However, biomarkers in this setting only would be useful if they identified nonresponsive patients with high accuracy for then they would potentially benefit most from a change or intensification of the treatment. In this setting, the single markers nucleosomes, CYFRA 21-1, NSE, and ProGRP as well as the combinations of these markers showed a very good diagnostic capacity indicated by high AUCs and high sensitivities in the high specificity range. This means that either combination of those markers indicated accurately the nonresponse to therapy and showed to be valuable for the monitoring of treatment efficacy.

For the early estimation of therapy response, it would be necessary to focus on the information already being available before start of the second course of treatment. Among all relevant markers, baseline values (BV2) of nucleosomes, ProGRP, and CYFRA 21-1 most efficiently identified those patients with insufficient therapy response. Combinations of nucleosomes with ProGRP, CYFRA 21-1, and NSE further increased the diagnostic power in terms of higher AUCs and higher sensitivities in the upper specificity range. Most notably, among all clinical and biochemical variables, which were included in multivariate regression analyses, only performance score and nucleosomes BV2 showed independent diagnostic value for indicating therapy response, whereas other variables did not.

As the number of patients in this study was too small to divide the group into an explorative biomarker finding group and a validation group, we used the leave-one-out validation procedure to validate our results. By doing this, our findings generally were confirmed with only slightly lower AUCs and sensitivities for most marker combinations.

As other studies have not respected biomarker determinations in defined intervals during the initial treatment phase for the follow-up evaluation in SCLC patients thus far, prospective validation studies will have to confirm the relevance of the markers presented here for the early estimation of therapy response and prognosis.

Nucleosomes are complexes of DNA and histones and are reported to be the main form of cell-free DNA in blood because they are better conserved in this constellation from digestion by plasmatic endonucleases (39–41). Although only few groups investigated the clinical relevance of circulating nucleosomes themselves, several studies focused on the relevance of quantitative and qualitative aspects of circulating nucleic acids in diagnosis, prognosis, and therapy monitoring of lung cancer disease (42–48), which yielded quite comparable findings than we did on circulating nucleosomes. Fournie et al. (43) as well as Gautschi et al. (46) have observed a correlation of DNA kinetics with efficacy of systemic therapy. Others have found a rapid increase of DNA concentration during the first days after application of systemic therapy followed by a rapid decrease we have reported on earlier with respect to nucleosomes (30, 49).

To our knowledge, this is the first study investigating in detail the initial changes of nucleosomes and lung cancer biomarkers for the early estimation of treatment response in SCLC patients during first-line chemotherapy. For this indication, our results clearly show the high relevance of circulating nucleosomes and either ProGRP, CYFRA 21-1, or NSE when determined before the start of the second therapy cycle.

No potential conflicts of interest were disclosed.