105 patients diagnosed with small cell lung cancer (SCLC) were enrolled at the Department of Medical Oncology of Respiratory in Guangxi Medical University Cancer Hospital from January 2018 to April 2021. Ethics approval number: LW2022044.

SCLC patients were qualified for inclusion if they met the following criteria:1. They have pathologically diagnosed SCLC; 2. Obtained blood samples; 3. Response assessments were done with a computed tomography scan (CT scan) of the chest and abdomen every two treatment cycles according to Response Evaluation Criteria in Solid Tumors (RECIST version 1.1). Each patient’s response was classified into one of the following categories: responders, including cases of complete response (CR), partial response (PR), and stable disease (SD), and non-responders, including issues of disease progression (PD). The exclusion criteria were as follows: incomplete demographic statistical information such as age, sex, or race; incomplete clinicopathology information such as tumor size (defined as the most accurate measurement of a solid primary tumor), tumor-node-metastasis (TNM) stage, tumor grade; incomplete therapeutic information in chemotherapy and radiotherapy (defined as the method performed as part of the first course of treatment); missing survival status and follow-up data. According to the National Comprehensive Cancer Network (NCCN) combined with the Veterans Administration Lung Study Group (VALG) approach for SCLC staging, limited disease (LD)-SCLC is specified as stage I to III (T any, N any, M0), and extension disease (ED)-SCLC is defined as stage IV (T any, N any, M1a/b). All subjects involved had received treatment using etoposide plus platinum antineoplastic agent in the first place. For general methods, see Figure 1A.

In this retrospective study, the data on the peripheral blood of the patients diagnosed with SCLC were collected which included T lymphocytes, T helper cells, T suppressor cells, T helper cells/T suppressor cells, natural killer cells, B lymphocytes, carcinoembryonic antigen, carbohydrate antigen 19–9, thymidine kinase1, neuron-specific enolase and Ki-67. All blood samples were collected before subjects had received any treatment.

Heat shock protein 90α (Yantai Protgen Biotechnology Development Co., Ltd., Shandong, China): Collect blood from patients with EDTA-K2 anticoagulation tube in the morning and apply it in a centrifuge 3000RPM for 15 min. Incubate the kit in water at 37°C for 30 min. The liquid in the kit gets thoroughly mixed to a point where no air bubbles are visible. Add 4.75*10⁵ μL of deionized water to the concentrated washing solution and mix well. Add calibrator to 400 μL of analyte diluent to dissolve and mix. Plasma samples need to be diluted 20-fold with diluent. Take 50 μL each of the calibrator and the diluted plasma sample, add them to the calibration well and sample the microplate. Add 50 μL of heat shock protein 90α marker solution to the wells and mix gently. Subsequent incubation happens in warm water at 37°C for 60 min after being sealed. Shake off the reactant solution, add a 300 μL washing solution/washing machine, wash the plate six times, and dry on absorbent paper. Add 50 μL of chromogenic reagent A to the wells above, then add 50 μL of chromogenic reagent B and incubate with warm water at 37°C for 20 min after mixing. Add 50 μL of stopping solution to the above wells to stop color development. Read the OD value immediately at 450 nm/620 (630) nm wavelength. Calculate the content of HSP90α in plasma samples based on the software algorithm.

The least absolute shrinkage and selection operator (LASSO) regression model selected the relevant factors for response evaluation and prognosis of small cell lung cancer. Logistic regression analysis was then used to build a convincing model for predicting chemotherapy efficacy and prognostic outcomes for small cell lung cancer by incorporating features selected in the LASSO approach. Clinicopathological features and biomarkers with p-values less than 0.05 in peripheral blood were included in the model. All selected predictors were used to develop a nomogram model to predict small cell lung cancer.

NCI-H146 and A549 lung cancer cells were purchased from the Institute of Cell Biology, Chinese Academy of Sciences (Shanghai, China). Cells were grown in Roswell Park Memorial Institute 1,640 medium containing 10% fetal bovine serum and 1% penicillin/streptomycin at 37°C in a humidified atmosphere with 5% CO₂.

The data were expressed as median and quartile. The Spearman correlation analysis evaluated the association between eHSP90α and other indexes. Receiver operating characteristic (ROC) curves were used to identify optimal cutoff values of variables in overall survival, progression-free survival, and response to treatment. Grouped according to cutoff value, Least Absolute Shrinkage and Selection Operator (LASSO) was used to analyze eHSP90α and efficacy. R 4.03 software was used to draw Kaplan-Meier curves of OS and PFS. The univariate and multivariate Cox regression was performed to investigate the relationship between eHSP90α and OS and PFS. A p-value less than 0.05 was considered statistically significant. Statistical analyses were performed using the SPSS computer program (SPSS 22.0).

In total, 105 patients with SCLC were enrolled in our analysis. All patients were with an average age of 62.00 ± 8.55 years at the time of diagnosis. Of the 105 patients, 33 (31.43%) underwent radiotherapy and 94 (89.52%) underwent chemotherapy. 27 (25.7%) were classified with limited disease small cell lung cancer and 78 (74.3%) were extension disease small cell lung cancer. The other parameters clinicopathological data are shown in Table 1. Next, we analyzed the differences and progressive groups of all peripheral blood indicators in Table 1 Alive and DEAD groups in OS, progress and no progressive group in PFS and the differences between the CR \ PR \ SD and PD groups in Response Evaluation. We found that eHSP90α has different expressions in these groups, as shown in Figure 1B. The expression of other peripheral blood indicators in these groups, see Supplementary Figure S1.

NSE is currently a commonly used marker for SCLC, and eHSP90α is a new biomarker. The correlation between these two biomarkers and other clinical indicators is analyzed here. Correlation analysis revealed that eHSP90α was positively correlated to the diameter of the tumor and neuron-specific enolase (NSE) in SCLC patients. This analysis shows that eHSP90α was negatively related to T lymphocytes and T helper cells in SCLC patients. There is no significant relationship between eHSP90α and another clinical index in LD-SCLC patients. NSE and eHSP90α had the same trend in ED-SCLC patients ( Table 2 ). NSE was positively correlated with eHSP90α in total SCLC and ED patients, indicating that eHSP90α is likely to have excellent diagnostic and prognostic efficacy. We next use the scattered dot diagram to show the relationship between the two continuous variables eHSP90α and NSE. We found that NSE and eHSP90α in the OS group, PFS progress group and PD groups in the Response Evaluation have significant correlations. See Figure 1C. Other correlation analysis results show in Supplementary Figures S1,S2.

ROC curves were used to analyze the diagnostic effect of eHSP90α in dichotomous variables such as overall survival status (alive/dead), progression-free survival status (whether progress or recurrence) and response evaluation (RECIST version 1.1), see Table 3 eHSP90α had the highest AUC value (OS 0.791, PFS 0.662, response evaluation 0.685) in all three ROC curves. T helper cells/T suppressor cells had the highest specificity (0.984) in the OS ROC curve, and NSE had the highest specificity in the PFS ROC curve (AUC 0.889) and response evaluation ROC curve (AUC 0.903). T lymphocytes had the highest specificity (0.903) in the response evaluation ROC curve and the highest sensitivity (0.841) in the OS ROC curve. eHSP90α and T lymphocytes had the same highest sensitivity (0.792) in the PFS ROC curve. eHSP90α had the highest sensitivity (0.778) in the response evaluation ROC curve. eHSP90α had the highest Youden in the OS ROC curve (0.534) and response evaluation ROC curve (0.397), while T helper cells had the highest Youden (0.187) in the PFS ROC curve. Taken together, eHSP90α has better sensitivity and specificity than NSE, and similar diagnostic effectiveness as NSE in ROC curves of OS (Figure 2A), PFS (Figure 2B) and response evaluation (Figure 2C).

Compared with OS and PFS, response evaluation (RECIST version1.1) does not contain information on survival time. Therefore, eHSP90α was used to explore the diagnostic significance of response status. We used two indicators, including eHSP90α, to establish a nomogram to evaluate the response of SCLC (Figure 3A). The receiver operating characteristic curve of the model had a higher AUC (0.764) than eHSP90α (AUC = 0.564) (Figure 3B). First, based on 105 SCLC patients, we used LASSO and 10-fold cross-validation to screen out six indicators (Figures 3C,D), including diameter, staging, chemotherapy, radiotherapy, location, eHSP90α (Figure 3C lambda. min = 0.03050932). To further select independent variables for effective chemotherapy from LASSO results, multivariate logistic regression analysis among these above features is shown in Table 4. The model finally contains radiotherapy (OR = 2.1103, p = 0.002276), eHSP90α (OR = 2.2677, p = 0.000524). The model was verified by bootstrapping (Figure 3E, C-index = 0.764). In this study, the threshold is from 20 to 88% (probability of patients and doctors); the decision curve analysis of our model showed that the response evaluation nomogram would profit more than of threshold (Figure 3F).

The ROC curve of the PFS was performed to determine the cutoff points of the index. The result is shown in Table 5. The threshold values of eHSP90α for OS, and PFS were 61.2 and 48.7, respectively. To determine which parameters were the independent prognostic factors of SCLC, univariate and multivariate analyses were conducted to explore the relationship between multiple variables, OS and PFS. In the multivariate analysis, high eHSP90α was an independent prognostic factor of a poor evolution in SCLC (OS, p < 0.001 and PFS, p = 0.024), whereas receiving radiotherapy was a good prognostic factor in SCLC (OS, p = 0.010 and PFS, p = 0.016). High NSE and M1 stage were significantly associated with shorter OS (NSE, p < 0.001 and M1 stage, p = 0.009). High CA-199 and M1 stage predicts earlier progression (CA-199, p = 0.018 and M1 stage, p < 0.001). We constructed a nomogram with four factors, including eHSP90α, to predict overall survival in SCLC patients (Figure 4A). The receiver operating characteristic curve of this model shows that AUC is 0.788. A nomogram for predicting progression-free survival was established with two indexes (Figure 4B). AUC is 0.671 in the receiver operating characteristic curve of this model. We analyzed the distribution of risk levels, survival status, and survival time patterns in SCLC patients divided into two groups with 61.2 ng/mL as the eHSP90α cut-off value. Kaplan-Meier survival analysis revealed that SCLC patients with higher eHSP90α expression had worse OS than those with lower eHSP90α expression (Figure 5A). Patients with high eHSP90α expression levels died more and survived shorter, as seen by survival time patterns (Figure 5B). The distribution of the eHSP90α risk score distribution suggests that higher expression of eHSP90α implies higher risk (Figure 5C). We analyzed the progression of SCLC patients and divided them into two groups with an eHSP90α cut-off value of 48.7 ng/ml. Kaplan-Meier survival analysis showed that the low expression group was more likely to be free of progression (Figure 6A). We made a scatterplot to assess progression risk, suggesting that more progression and higher mortality rates appear in the high-expression group. In contrast, patients’ conditions in the low-expression group were mostly non-progressive (Figure 6B). By sorting the patients according to the risk value from low to high, we obtained the progression risk curve graph, which shows the high expression of eHSP90α corresponds to a greater risk of progression (Figure 6C). eHSP90α above was significantly associated with NSE in ED-SCLC patients, but not in LD-SCLC patients. The occurrence of distant metastases is the main cause of poor PFS and OS in SCLC. But only eHSP90α but not NSE was significantly associated with poorer PFS in SCLC. We performed based on distant metastasis status for NSE and eHSP90α as we made a prediction that NSE might be associated with distant metastasis (Supplementary Tables S1,S2). The results showed that NSE was an independent factor affecting the distant metastasis of SCLC.

To further explore the expression of eHSP90α in SCLC and NSCLC, we detected eHSP90α in different cell supernatants from six dishes (Supplementary Figure S3). The eHSP90α in SCLC cell supernatant is higher than in NSCLC (Figure 7). Consistent with our clinical results.