The value of plasma levels of heat shock protein 90α for predicting the treatment efficacy and prognosis of lung cancer

Abstract

Objective

This study aimed to evaluate the value of heat shock protein 90α (HSP90α) in predicting treatment response and progression-free survival (PFS) among patients with lung cancer.

Methods

We retrospectively analyzed 585 patients with lung cancer treated at Jilin Cancer Hospital between January 2014 and December 2023. Patients were categorized by pretreatment HSP90α levels into high- and low-level groups and by posttreatment changes after four cycles into increasing and decreasing groups. We assessed the association between baseline HSP90α levels and clinical characteristics, its correlation with treatment efficacy and PFS, and its combined predictive ability with classical tumor biomarkers (CEA, NSE, and SCC).

Results

High pretreatment HSP90α levels were associated with older age and specific pathological types and served as an independent risk factor for shorter PFS. Patients with low baseline HSP90α levels exhibited a better short-term treatment response and longer PFS than did those with high baseline HSP90α levels. Increasing HSP90α after treatment correlated with an improved response and prolonged PFS. HSP90α levels were positively correlated with CEA, NSE, and SCC levels. Notably, the combination of HSP90α and CEA expression significantly enhanced PFS prediction in patients with lung adenocarcinoma.

Conclusions

HSP90α is a significant predictive biomarker for treatment efficacy and prognosis in lung cancer patients, particularly when it is combined with traditional markers such as CEA.

Background

Lung cancer is a malignant tumor that poses a serious threat to human life and health [1]. The latest cancer statistics report released by the International Agency for Research on Cancer of the World Health Organization revealed that there were 2.48 million newly diagnosed lung cancer cases and 1.82 million lung cancer-related deaths worldwide in 2022, ranking first in both categories among all malignant tumors [2]. Most lung cancer patients are in the advanced stage of the disease at the time of diagnosis, which is characterized by a high tumor burden and poor prognosis [3]. Imaging examinations, including computed tomography (CT), magnetic resonance imaging (MRI) and color Doppler ultrasound, are the main methods used for the clinical diagnosis and treatment of lung cancer. In most cases, imaging examinations can accurately identify tumor lesions and are effective for diagnosing lung cancer and evaluating treatment effects [4]. However, imaging examinations also have certain shortcomings and limitations, including unavoidable ionizing radiation, relatively high examination cost, relatively long examination time, potential contrast agent allergy possibility and an inability to predict disease development [5]. A tumor biomarker is a necessary auxiliary tool that has been widely used in the clinical diagnosis and treatment of a variety of malignant tumors. In particular, serum tumor biomarkers are the most commonly used biomarkers. The commonly used serum tumor biomarkers for lung cancer include carcinoembryonic antigen (CEA), neuron-specific enolase (NSE), CYFA21-1 and squamous cell carcinoma antigen (SCC). These classical tumor biomarkers of lung cancer have certain specificity and sensitivity and can be used to evaluate clinical therapeutic efficacy and predict disease development to a certain extent, but they may be affected by diet, drugs, infection and other factors [6]. Searching for lung cancer biomarkers with higher specificity and sensitivity or performing joint detection of multiple lung cancer tumor biomarkers can better reflect the clinical therapeutic efficacy and more accurately predict the prognosis of the disease; this approach has become a hot topic in recent lung cancer research [7, 8]. Serum HSP90α is a newly discovered tumor biomarker. In recent years, many scientific papers have investigated the value of HSP90α for predicting the diagnosis and treatment of a variety of malignant tumors.

Heat shock proteins are a class of highly conserved immune-derived proteins that are widely found in mammals and microorganisms [9]. HSP90α is an important member of the heat shock protein family [10, 11]. Research on HSP90α is relatively new, as its gene sequence was reported only in 1989. In 1992, researchers reported that HSP90α can be secreted by tumor cells. In 2009, the team of Professor Luo Yongzhang, a well-known scholar at Tsinghua University, first reported the regulatory mechanism of tumor cell-specific secretion of HSP90α and revealed the molecular differences between extracellularly secreted HSP90α and intracellularly secreted HSP90α [12]. Studies have shown that extracellularly secreted HSP90α can promote the invasion and metastasis of a variety of malignant tumors [13]. Blood levels of HSP90α are positively correlated with the degree of malignancy and tumor development [14]. These findings indicate that HSP90α may be a tumor biomarker with great potential. To date, many clinical studies on the use of HSP90α in the diagnosis and treatment of a variety of malignant tumors have been published; these studies have preliminarily confirmed that HSP90α has high specificity and high sensitivity in a variety of malignant tumor diseases and can serve as a strong predictor of the short-term treatment effect and long-term survival of patients with a variety of malignant tumors [15–18]. However, our research team systematically reviewed the research reports in this field and reported that these research reports have certain limitations: older age, smaller sample size and more limited research. To fill this research gap in this field, we carried out the current study to investigate the serum HSP90α levels of patients with lung cancer. The aim of this study was to explore the value of serum HSP90α levels for predicting the treatment and prognosis of lung cancer and to provide evidence for the clinical application of HSP90α in lung cancer treatment.

Methods

Participants

The study included lung cancer patients who were admitted to the Department of Integrated Traditional Chinese and Western Medicine at Jilin Cancer Hospital from January 1, 2014, to December 31, 2023. The inclusion criteria were as follows: ① patients diagnosed with primary lung cancer through pathology or cytology; ② patients with stage III-IV disease who had not received any previous antitumor treatment (including surgery, chemotherapy, immunotherapy, targeted therapy, or traditional Chinese medicine treatment); ③ patients who had undergone at least four consecutive cycles of the same treatment plan (including medication and dosage) for antitumor therapy on the basis of authoritative domestic and international diagnostic and treatment guidelines at the time of treatment; ④ patients with complete clinical diagnosis and treatment data (including laboratory and examination results before and after 4 cycles of treatment); and ⑤ patients with traceable contact information and reliable prognostic information. The exclusion criteria were as follows: ① patients who presented with other primary malignant tumors; ② patients with severe infections, severe cardiovascular and cerebrovascular diseases, severe liver and kidney dysfunction, or autoimmune diseases; or ③ patients whose treatment was delayed for more than 2 weeks. The patients or their relatives were informed about the nature of this study and provided informed consent. This study was approved by the ethics review committee of Jilin Cancer Hospital (Ethics Batch Number: 202208-006-01).

Methods

In this correlation study, we collected the following personal information and clinical diagnosis and treatment information from the enrolled patients: age, sex, pathological classification, disease stage, initial treatment method, progression-free survival, and serum levels of HSP90α, CEA, NSE, CYFA21-1, and SCC (squamous cell carcinoma only) before and after 4 cycles of initial treatment. The serum levels of CEA, NSE, CYFA21-1, and SCC were measured via chemiluminescence. Serum HSP90α levels were measured via enzyme-linked immunosorbent assay (ELISA).

Patients were divided into a low-HSP90α group and a high-HSP90α group on the basis of a cutoff value of 150.48 ng/mL. The normal ranges for the biomarkers are as follows: CEA, 0–5 ng/mL; NSE, 0–20 ng/mL; CYFA21-1, 0–3.3 ng/mL; and SCC, 0–1.5 ng/mL. On the basis of the normal ranges of the tumor biomarkers, low-level and high-level groups of corresponding tumor biomarkers were defined. Changes in the levels of various tumor biomarkers before initial treatment and after 4 cycles of treatment were recorded and compared, and after 4 cycles of treatment, patients were divided into a group that exhibited higher biomarker levels (the increased group) and a group that exhibited lower biomarker levels (the decreased group).

The correlations of baseline characteristics (age, sex, pathological classification, disease stage, initial treatment method, and levels of commonly used lung cancer tumor biomarkers) with pretreatment HSP90α levels were calculated. Additionally, the associations of these baseline characteristics with progression-free survival (PFS) were determined via univariate and multivariate Cox regression analyses. To explore the correlation between pretreatment HSP90α levels and the curative effect of lung cancer treatment, differences in the objective response rate (ORR) and disease control rate (DCR) were compared between the high-HSP90α and low-HSP90α groups before treatment and after 4 cycles of treatment. To investigate the correlation between the increase and decrease in HSP90α after treatment and the therapeutic efficacy among patients with lung cancer, differences in the ORR and DCR between the increased and decreased groups after 4 cycles of treatment were compared. In accordance with the Response Evaluation Criteria for Solid Tumors (RESCIST) 1.1, treatment efficacy was evaluated according to changes in imaging findings before and after 4 cycles of treatment. Efficacy was evaluated on the basis of the following parameters: complete response (CR), partial response (PR), stable disease (SD) and progressive disease (PD). The ORR and DCR were calculated as follows: ORR= (CR + PR)/total number of patients and DCR= (CR + PR + SD)/total number of patients. To explore the correlations of pretreatment HSP90α levels with increases or decreases in this biomarker after 4 cycles of treatment and the PFS of patients with lung cancer, we generated survival curves, performed a statistical analysis, and analyzed the pathological classifications via subgroup analyses. Follow-up was conducted via telephone calls to the patients, family members, or guardians in the medical records to monitor the PFS of patients. PFS was defined as the time between the date of pathological diagnosis and the observation of disease progression or the occurrence of death for any reason. Follow-up for PFS ended on July 1, 2024. To investigate the effect of HSP90α combined with classical lung cancer tumor biomarkers on PFS among lung cancer patients, we analyzed whether there was a positive correlation between the expression of HSP90α and that of CEA, NSE, and SCC. On the basis of different pathological subtypes of lung cancer, we explored the ability of the combination of classical tumor biomarkers and HSP90α expression levels to predict PFS.

Statistical analysis

For multivariate analyses, P values and hazard ratios (HRs) were calculated via Cox regression with the survival package. For all survival analyses, P values were calculated via the log-rank test, and HRs were calculated via univariate Cox regression. For the analyses of the clinical response data (ORR and DCR), P values and odds ratios (ORs) were calculated via Fisher’s exact test (two-sided). All survival plots and forest plots for the Cox proportional hazards model were plotted via the ggplot2 (v3.4.3) and survminer (v0.4.9) packages. The optimal cutoff point for HSP90α was determined via the maximally selected rank statistics from the maxstat (v0.7–25) package. All survival and correlation analyses were performed using R Statistical Computing Environment v.4.3.1 (http://www.r-project.org).

Results

Patient characteristics

We systematically reviewed the data of lung cancer patients admitted to the Department of Integrated Traditional Chinese and Western Medicine at Jilin Cancer Hospital from January 1, 2014, to December 31, 2023. After the inclusion and exclusion criteria were met, a total of 585 lung cancer patients were included in the study, and their pretreatment information was collected and statistically analyzed. The average age of these patients was 63.24 ± 8.70 years. Among these patients, 240 were female (41.03%), and 345 were male (58.97%). In terms of pathological classification, the majority of the patients had adenocarcinoma (317 patients, 54.19%), followed by squamous cell carcinoma (155 patients, 26.50%), small cell carcinoma (82 patients, 14.02%), and other pathological classifications (31 patients, 5.30%). In terms of disease stage, 75 patients had stage IIIA disease, accounting for 12.82%; 96 patients had stage IIIB disease, accounting for 16.41%; 22 patients had stage IIIC disease, accounting for 3.76%; 166 patients had stage IVA disease, accounting for 28.38%; and 226 patients had stage IVB disease, accounting for 38.63%. In terms of initial treatment, 243 patients received chemotherapy, accounting for 41.54%; 157 patients received single-drug targeted therapy, accounting for 26.84%; 126 patients received chemotherapy combined with immunotherapy, accounting for 21.54%; and other treatment methods, such as chemotherapy combined with targeted therapy and monotherapy immunotherapy, were administered to a relatively small proportion of the included patients. In terms of tumor biomarker levels before initial treatment, 275 individuals were in the low-CEA group, accounting for 47.01%, whereas 310 individuals were in the high-CEA group, accounting for 53.99%; 420 individuals were in the low-NSE group, accounting for 71.79%, while 165 individuals were in the high-NSE group, accounting for 28.21%; 237 individuals were in the low-CYFA21-1 group, accounting for 40.51%, while 348 individuals were in the high-CYFA21-1 group, accounting for 59.49%; 295 individuals were in the low-SCC group, accounting for 77.02%; and 88 individuals were in the high-SCC group, accounting for 22.98% (SCC detection is not applicable to all patients, including all squamous cell carcinoma patients and some other pathological subtype patients); and 500 individuals were in the low-HSP90α group, accounting for 85.47%, while 85 individuals were in the high-HSP90α group, accounting for 14.53%. A total of 294 people exhibited higher levels of HSP90α after 4 cycles of treatment, while 291 patients exhibited lower levels.

There were significant differences in age, pathological type, and NSE level between the high-HSP90α group and the low-HSP90α group before initial treatment (P < 0.05), but there were no significant differences in disease stage, CEA level, or CYFA21-1 level (P > 0.05). The baseline characteristics of the patients and their correlations with pretreatment HSP90α levels are detailed in Table 1. Because only patients with squamous cell carcinoma of the lung were tested for SCC, the sample size of patients with SCC was small. Considering the reliability of the data and other aspects, we excluded this parameter from further analysis.

Table 1.

Baseline characteristics of patients and their correlation with HSP90α levels (n = 585)

	All (N = 585)	High HSP90α (N = 85)	Low HSP90α (N = 500)	p value
Age	63.24 ± 8.70	61.12 ± 7.70	63.60 ± 8.81	0.008*
Gender				1.000
Male	345 (58.97%)	50 (58.82%)	295 (59.00%)
Female	240 (41.03%)	35 (41.18%)	205 (41.00%)
Pathological classification				0.006*
Adenocarcinoma	317 (54.19%)	36 (42.35%)	281 (56.20%)
Squamous cell carcinoma	155 (26.50%)	21 (24.71%)	134 (26.80%)
Small cell carcinoma	82 (14.02%)	19 (22.35%)	63 (12.60%)
Other	31 (5.30%)	9 (10.59%)	22 (4.40%)
Disease staging				0.845
Stage ⅢA	75 (12.82%)	11 (12.94%)	64 (12.80%)
Stage ⅢB	96 (16.41%)	16 (18.82%)	80 (16.00%)
Stage ⅢC	22 (3.76%)	3 (3.53%)	19 (3.80%)
Stage ⅣA	166 (28.38%)	20 (23.53%)	146 (29.20%)
Stage ⅣB	226 (38.63%)	35 (41.18%)	191 (38.20%)
Initial treatment				0.071
Targeted therapy	157 (26.84%)	16 (18.82%)	141 (28.20%)
Targeted therapy + Immunotherapy	2 (0.34%)	1 (1.18%)	1 (0.20%)
Chemotherapy	243 (41.54%)	46 (54.12%)	197 (39.40%)
Targeted therapy + Chemotherapy	42 (7.18%)	3 (3.53%)	39 (7.80%)
Targeted therapy + Immunotherapy + Chemotherapy	5 (0.85%)	0 (0.00%)	5 (1.00%)
Chemotherapy + Immunotherapy	126 (21.54%)	17 (20.00%)	109 (21.80%)
Immunotherapy	10 (1.71%)	2 (2.35%)	8 (1.60%)
CEA				0.564
<5 ng/ml	275 (47.01%)	37 (43.53%)	238 (47.60%)
≥5 ng/ml	310 (52.99%)	48 (56.47%)	262 (52.40%)
NSE				<0.001*
<20 ng/ml	420 (71.79%)	47 (55.29%)	373 (74.60%)
≥20 ng/ml	165 (28.21%)	38 (44.71%)	127 (25.40%)
CYFA21-1				0.156
<3.3 ng/ml	237 (40.51%)	28 (32.94%)	209 (41.80%)
≥3.3 ng/ml	348 (59.49%)	57 (67.06%)	291 (58.20%)

*P < 0.05

Univariate and multivariate Cox regression of PFS in patients with lung cancer

Univariate Cox regression analysis revealed that female sex, squamous cell carcinoma, small cell carcinoma, high pretreatment levels of NSE, high pretreatment levels of CYFA21-1, and high pretreatment levels of HSP90α were significantly correlated with PFS in patients with lung cancer (P < 0.05). Multivariate Cox regression analysis revealed that after excluding the influence of other variables, female sex, squamous cell carcinoma status, small cell carcinoma status, a high pretreatment level of CYFA21-1, and a high pretreatment level of HSP90α were significantly correlated with PFS in patients with lung cancer (P < 0.05) and were independent risk factors for PFS. The probability of disease progression in the group with high pretreatment HSP90α levels was 1.17 times greater than that in the group with low pretreatment HSP90α levels (HR = 1.47; 95% CI, 1.13–1.90; P = 0.004). Table 2 shows the results of the comparative analysis of the univariate and multivariate survival rates, and Fig. 1 shows the multivariate Cox forest diagram.

Table 2.

Comparison and analysis of the univariate/multivariate survival rates of patients (n = 585)

	Univariate		Multivariate
Term	HR	p value	HR	p value
Female sex	0.73(0.61–0.88)	0.001*	0.76(0.63–0.93)	0.007*
Pathological classification of squamous cell carcinoma	1.46(1.17–1.82)	0.001*	1.43(1.11–1.84)	0.006*
Pathological classification of small cell carcinoma	1.72(1.32–2.26)	0.000*	1.68(1.23–2.29)	0.001*
Pathological classification others	1.43(0.96–2.13)	0.078	1.35(0.90–2.04)	0.151
Stage IIIB	1.25(0.87–1.79)	0.223	1.29(0.90–1.86)	0.167
Stage IIIC	1.56(0.89–2.74)	0.119	1.55(0.88–2.72)	0.130
Stage IVA	1.07(0.78–1.48)	0.666	1.28(0.91–1.80)	0.158
Stage IVB	1.12(0.82–1.52)	0.476	1.32(0.95–1.82)	0.100
CEA ≥ 5 ng/ml	1.01(0.84–1.21)	0.924	1.08(0.88–1.33)	0.443
NSE ≥ 20 ng/ml	1.38(1.13–1.69)	0.002*	1.12(0.89–1.40)	0.331
CYFA211 ≥ 3.3 ng/ml	1.40(1.16–1.69)	0.001*	1.30(1.07–1.59)	0.009*
HSP90α ≥ 150.48 ng/ml	1.60(1.24–2.06)	0.000*	1.47(1.13–1.90)	0.004*

 *P < 0.05

Fig. 1.

Forest plot of PFS among lung cancer patients derived from multivariate Cox regression. Note: *P < 0.05. LUAD: Lung adenocarcinoma. LUSC: Lung squamous cell carcinoma. SCLC: Small cell lung carcinoma. The Figure shows the factors with a significant effect on PFS. The vertical line at 1 represents "no effect." Factors with data shifted entirely to the right of this line are associated with a higher risk of progression, whereas those to the left are associated with a lower risk. If a factor exhibits a horizontal line that crosses the centerline, the effect of that factor is uncertain in this analysis.

Correlation between the HSP90α level and short-term therapeutic efficacy in patients with lung cancer

After 4 cycles of treatment, the ORR and DCR of the high-HSP90α group were 30.40% and 89.20%, respectively. The ORR and DCR were 24.71% and 83.53%, respectively, in the low-HSP90α group. Compared with those of the low-level group, the data of the high-level group were better, but the difference was not statistically significant (P > 0.05). After 4 cycles of treatment, the ORR of the group with increased HSP90α expression was 34.01%, which was significantly greater than that of the group with decreased HSP90α expression after treatment (25.08%, P < 0.05). The DCR of the group with increased HSP90α expression was 91.15%, which was significantly greater than that of the group with decreased HSP90α expression (85.56%) after treatment (P < 0.05). The results of these analyses are shown in Table 3.

Table 3.

Correlations between HSP90α levels and short-term efficacy before and after treatment (n = 585)

		ORR	p value	DCR	p value
Before treatment(%)	Low-HSP90α group(<150.48 ng/ml)	30.40	0.307	89.20	0.143
	High-HSP90α group(≥150.48 ng/ml)	24.71		83.53
After treatment(%)	Hsp90α increased group	34.01	0.019*	91.15	0.039*
	Hsp90α decreased group	25.08		85.56

*P < 0.05

Correlation between the HSP90α level and PFS in patients with lung cancer

In the entire study population, the median PFS in the low-HSP90α group (500 people) was 9.63 months (8.6–11.13), which was significantly longer than that in the high-HSP90α group (85 people) of 6.1 months (4.9–8.3) (P < 0.05). In the lung adenocarcinoma subgroup, the median PFS of the low-HSP90α group (281 people) was 12.6 months (10.7–14.4), which was significantly longer than that of the high-HSP90α group (36 people) (7.3 months; 5.57–11.53) (P < 0.05). In the subgroup of patients with SCLC, the median PFS of the low-HSP90α group (63 people) was 5.93 months (5.23–8), which was significantly longer than that of the high-HSP90α group (19 people) of 4.73 months (3.9-na) (P < 0.05). In the lung squamous cell carcinoma subgroup, the median PFS of the low-HSP90α subgroup (63 people) was 7.13 months (6.17–9.53), and the median PFS of the high-HSP90α subgroup (19 people) was 6.6 months (3.77–17.5); this difference was not statistically significant (P > 0.05). The details of these analyses are shown in Fig. 2.

Fig. 2.

K‒M curves for the PFS of lung cancer patients in the HSP90α groups before initial treatment. A K‒M curve for PFS in the high- and low-HSP90α groups among all lung cancer patients. B K‒M curves for PFS in the high- and low-HSP90α groups of patients with adenocarcinoma. C K‒M curves for PFS in the high- and low-HSP90α groups of patients with small-cell carcinoma. D K‒M curves for PFS in the high- and low-HSP90α groups of patients with squamous cell carcinoma. The lines represent the percentage of patients whose cancer did not worsen over time. A line that remains higher for longer periods indicates a group with better outcomes.

In the entire study population, the median PFS of patients with increased HSP90α levels after treatment (294 people) was 9.63 months (7.8–11.87), which was significantly longer than that of patients with decreased HSP90α levels after treatment (291 people) (8.3 months; P < 0.05). We performed subgroup analyses of pathological classification and therapeutic drugs to further analyze the correlation between serum heat shock protein 90α levels and progression-free survival in patients with lung cancer. The first analysis was based on pathological classification. In the lung adenocarcinoma subgroup, the median PFS of patients with increased HSP90α levels after treatment (168 people) was 13.07 months (10.73–15.17), and the median PFS of patients with decreased HSP90α levels after treatment (149 people) was 10.7 months (9.57–12.73); this difference was not statistically significant (P > 0.05). In the subgroup of patients with SCLC, the median PFS of patients with increased HSP90α levels after treatment (31 people) was 5.8 months (5.07–8.4), and the median PFS of patients with decreased HSP90α levels after treatment (52 people) was 5.7 months (4.77–8.3); this difference was not statistically significant (P > 0.05). In the lung squamous cell carcinoma subgroup, the median PFS of patients with increased HSP90α levels after treatment (81 people) was 7.4 months (6–11.13), and the median PFS of patients with decreased HSP90α levels after treatment (74 people) was 6.63 months (5.13–9.83); this difference was not statistically significant (P > 0.05). See Fig. 3 for details. Subsequently, an analysis based on the drug treatment was performed. In the chemotherapy subgroup, the median PFS of patients with elevated serum heat shock protein 90α levels (112 patients) was 8.00 months (5.85–10.15), and the median PFS of patients with reduced serum heat shock protein 90α levels (131 patients) was 5.77 months (4.81–6.73); this difference was statistically significant (P < 0.05). In the targeted therapy subgroup, the median PFS of patients with elevated serum heat shock protein 90α levels (87 patients) was 13.77 months (10.49–17.05), and that of patients with reduced serum heat shock protein 90α levels (70 patients) was 12.73 months (9.30–16.16); this difference was not statistically significant (P > 0.05). In the chemotherapy combined with immunotherapy subgroup, the median PFS of patients with elevated serum heat shock protein 90α levels (60 patients) was 6.67 months (4.59–8.75), and that of patients with reduced serum heat shock protein 90α levels (65 patients) was 7.40 months (4.16–10.64); this difference was not statistically significant (P > 0.05). See Fig. 3 for details.

Fig. 3.

K‒M curves of the changes in PFS in the HSP90α groups among lung cancer patients after 4 cycles of treatment. A K‒M curves for PFS in the increased and decreased HSP90α groups among all lung carcinoma patients. B K‒M curves for PFS in the increased and decreased HSP90α groups among patients with adenocarcinoma.C K‒M curves for PFS in the increased and decreased HSP90α groups among patients with small cell carcinoma. D K‒M curves for PFS in the increased and decreased HSP90α groups among patients with squamous cell carcinoma. The lines represent the percentage of patients whose cancer did not worsen over time. A line that remains higher for longer periods indicates a group with better outcomes

Prediction of PFS in patients with lung cancer via combined detection of HSP90α and other tumor biomarkers

Correlation analysis between the expression levels of HSP90α and CEA revealed that the expression level of CEA in the serum of patients with lung adenocarcinoma was positively correlated with the expression level of HSP90α (r = 0.27, P < 0.001). On the basis of the HSP90α and CEA levels, 317 patients with lung adenocarcinoma were divided into a high-HSP90α + high-CEA group (Hhigh + Chigh), a high-HSP90α + low-CEA group (Hhigh + Clow), a low-HSP90α + high-CEA group (Hlow + Chigh), and a low-HSP90α + low-CEA group (Hlow + Clow). PFS was 11.53 months (9.03–16.5) in the Hhigh + Chigh group, 10.7 months (9.13–15.67) in the Hhigh + Clow group, 9.7 months (7.33–13.27) in the Hlow + Chigh group and 14.4 months (11.33–19.2) in the Hlow + Clow group. There was a significant difference among the four groups (P < 0.05), and the PFS of the Hlow + Clow group was the longest (Fig.4).

Fig. 4.

K‒M curves of the changes in PFS in the HSP90α groups among lung cancer patients after 4 cycles of treatment. A K–M curves of PFS in patients with lung cancer in the groups with increasing and decreasing HSP90α expression who were treated with chemotherapy only B K–M curves of PFS in patients with lung cancer in the groups with increasing and decreasing HSP90α expression who were treated with targeted therapy C K–M curves of PFS in patients with lung cancer in the groups with increasing and decreasing HSP90α expression who were treated with chemotherapy combined with immunotherapy. The lines represent the percentage of patients whose cancer did not worsen over time. A line that remains higher for longer periods indicates a group with better outcomes

Correlation analysis of the expression levels of HSP90α and NSE revealed that the expression level of NSE in the serum of patients with SCLC was positively correlated with the level of HSP90α (r = 0.31, P < 0.05). On the basis of the HSP90α and NSE levels, 82 patients with SCLC were divided into a high-HSP90α + high-NSE group (Hhigh + Nhigh), a high-HSP90α + low-NSE group (Hhigh + Nlow), a low-HSP90α + high-NSE group (Hlow + Nhigh), and a low-HSP90α + low-NSE group (Hlow + Nlow). PFS was 4.5 months (3.9–6.13) in the Hhigh + Nhigh group, 7.07 months (6.23-NA) in the Hhigh + Nlow group, 6 months (5.7–23.1) in the Hlow + Nhigh group, and 5.47 months (4.3–15.9) in the Hlow + Nlow group. There were no significant differences among the four groups (P > 0.05).

Correlation analysis of the expression levels of HSP90α and SCC revealed that the expression level of SCC in the serum of patients with lung squamous cell carcinoma was positively correlated with the expression level of HSP90α (r = 0.3, P < 0.001). On the basis of the HSP90α and SCC levels, 155 patients with lung squamous cell carcinoma were divided into a high-HSP90α + high-SCC group (Hhigh + Shigh; 39 patients), a high-HSP90α + low-SCC group (Hhigh + Slow; 38 patients), a low-HSP90α + high-SCC group (Hlow + Shigh; 36 patients) and a low-HSP90α + low-SCC group (Hlow + Slow; 42 patients). PFS was 6.6 months (5.13–11.17) in the Hhigh + Shigh group, 10.23 months (5.77–18.2) in the Hhigh + Slow group, 5.1 months (3.67–9.53) in the Hlow + Shigh group, and 7.5 months (6.17–13.4) in the Hlow + Slow group. There were no significant differences among the four groups (P > 0.05). The predictive effects of the combination of HSP90α and classical tumor biomarkers on the PFS of patients with lung cancer are shown in Fig. 5.

Fig. 5.

Correlations between the expression levels of HSP90α and classical tumor biomarkers of lung cancer and the ability of combined detection to predict the PFS of lung cancer patients. A Correlation analysis between the expression levels of HSP90α and CEA in the serum of patients with adenocarcinoma. B K–M curves and log-rank tests for PFS in patients with different levels of HSP90α and adenocarcinoma. C Analysis of the correlation between the serum expression levels of HSP90α and NSE in patients with small cell carcinoma. D K–M curves and log-rank tests for PFS in patients with different levels of HSP90α and small cell carcinoma. E Correlation analysis between the expression levels of HSP90α and SCC in the serum of patients with squamous cell carcinoma. F K–M curves and log-rank tests for PFS in patients with different levels of HSP90α and squamous cell carcinoma. The relationship between the two tumor markers is shown in the figure. The overall pattern suggests whether they tend to rise and fall together (positive correlation), or one rises as the other falls (negative correlation). The lines represent the percentage of patients whose cancer did not worsen over time. A line that remains higher for longer periods indicates a group with better outcomes

Discussion

Serum HSP90α is an emerging tumor biomarker that has received increased attention from researchers since its discovery [19]. Multiple diseases, including lung cancer, liver cancer, breast cancer, colorectal cancer, gastric cancer, and malignant melanoma, are associated with HSP90α [15]. Recent research has shown that HSP90α may be a potential target for immune-refractory tumors [20].

Our study has the following characteristics. First, in terms of the inclusion and exclusion criteria of patients, considering that HSP90α of treated patients may be affected by a variety of factors, we focused on the initial treatment of stage III-IV lung cancer patients, which can maximally exclude the influence of many confounding factors on the level of HSP90α. It is important to note that our study excluded patients with other primary malignancies, severe infections, or significant comorbidities. While this was necessary to control for confounding and ensure a clear assessment of the intervention, it might limit the generalizability of our findings. Consequently, the results should be applied with caution to patients with complex medical backgrounds, and future research should aim to validate these findings in more heterogeneous, real-world populations. Second, our study design involved outcome indicators based on clinical reality, including short-term clinical efficacy and long-term PFS benefit. This study focused on the ORR, DCR and PFS, and subgroup analyses were performed on the basis of HSP90α levels and the pathological classification and treatment of lung cancer. Finally, on the basis of the pathological classification of lung cancer, we discussed the clinical application prospects of the combined detection of HSP90α and classical tumor biomarkers of lung cancer.

To explore the significance of HSP90α in lung cancer screening and diagnosis, we collected, summarized, and statistically analyzed the general data of the enrolled patients. Considering that there are great differences in the treatment methods used for the pathological classification of some lung cancers, we innovatively collected the initial treatment methods of patients, which provided us with more ideas for the analysis of subsequent research results. Our collection of these data provides a better entry point for our subsequent subgroup analyses of the correlation between serum heat shock protein 90α levels and the progression-free survival of patients with lung cancer. Through this part of the study, we found that age, pathological classification and NSE level were significantly correlated with the initial level of HSP90α, but the initial level of HSP90α was not significantly correlated with the disease stage at the initial treatment or with the CEA and CYFA21-1 levels. The level of HSP90α in newly diagnosed patients is associated with many factors. Our results may provide some future research directions. First, metabolism, which is directly related to age, may affect HSP90α levels. Second, the lung cancer patients in our study were tested for a variety of classical lung cancer tumor biomarkers, and the secretion of HSP90α may differ across different pathological types of lung cancer, which may further affect the correlation between the levels of classical lung cancer tumor markers and HSP90α levels in different pathological types of lung cancer. Finally, in the actual clinical diagnosis and treatment process, the disease stage of patients with advanced primary lung cancer is not necessarily related to the degree of tumor burden [21]. For example, lung cancer patients with axillary lymph node metastasis are judged to have stage IV disease, whereas patients with large local tumors without distant metastasis are judged to have stage III disease. We believe that this distinction may explain why disease stage does not directly affect HSP90α levels.

We subsequently analyzed factors related to the survival benefit of patients. HSP90α is among the factors of greatest concern. To avoid other interfering factors as much as possible, the enrolled patients were all newly treated patients with lung malignancies. After standard first-line treatment, the PFS of these patients was more reliable, and OS was related to follow-up treatment, compliance, adverse reactions, drug selection, and other aspects of patients. We did not analyze OS.

Short-term efficacy is the indicator of greatest concern among clinicians and lung cancer patients, and it is also the most commonly used outcome indicator in clinical tumor research. In this study, to explore the correlations of initial HSP90α levels and changes in HSP90α with short-term treatment efficacy, we divided all the patients into a high-HSP90α group and a low-HSP90α group. Furthermore, we divided patients into an increased group (i.e., those who exhibited increased HSP90α levels after treatment) and a decreased group (i.e., those who exhibited decreases in HSP90α levels after treatment). By exploring the short-term efficacy of each treatment separately, we found that patients in the low-HSP90α group had a higher ORR and DCR compared with those in the high-HSP90α group. Although the difference was not statistically significant, these findings also explained, to a certain extent, that higher pretreatment HSP90α levels were associated with more challenging disease control, which was consistent with the results of previous research [22, 23]. Interestingly, our study revealed that after 4 cycles of treatment, the ORR and DCR of the increased group were significantly improved compared with those of the decreased group. We reviewed the data of patients in the groups with increased and decreased HSP90α levels after treatment. First, compared with that in the group with decreased HSP90α levels, the proportion of patients with small cell lung cancer was lower in the group with increased HSP90α levels after treatment (9.52% vs. 18.56%), whereas the proportion of patients with lung adenocarcinoma was greater in the group with increased HSP90α levels (60.20% vs. 48.11%). Different pathological types are associated with different treatment methods, which may affect the short-term efficacy of treatment. This heterogeneity may explain some of the results of the current study. Second, to better analyze the short-term curative effect, we examined the HSP90α levels after four cycles of regular treatment; potential confounding factors could have affected the changes in HSP90α levels. There were 294 patients in the group with increased HSP90α levels after treatment, of whom only 47 patients exceeded the cutoff value for a high HSP90α level, accounting for 15.99% of the patients. This smaller patient proportion did not fully reflect the sensitivity of lung cancer to HSP90α. Finally, although previous basic research has described the intracellular secretion of HSP90α into the blood, we believe that tumor suppression after antitumor treatment may lead to increased secretion of HSP90α.

Owing to the rapid development of targeted therapy, immunotherapy and other emerging drugs, the OS of lung cancer patients has greatly improved [24–26]. In our study, the HSP90α levels of newly treated patients with stage III–IV lung cancer were recorded. Compared with long-term OS, PFS may be more suitable for exploring the correlation with HSP90α levels. In terms of the PFS of the whole population, the survival benefit of the low-HSP90α group was significantly greater than that of the high-HSP90α group. Moreover, the subgroup analyses of adenocarcinoma and small cell carcinoma revealed significant advantages. These findings indicated that the initial HSP90α level could predict the PFS of patients with lung cancer to a certain extent, which might be related to the tumor burden. Studies have shown that the higher the tumor burden is, the higher the HSP90α level. In the actual clinical treatment process, disease control is challenging for patients with lung cancer and high tumor burdens, which impacts PFS [27]. Additionally, the advantage of the low-HSP90α group in terms of PFS may also be related to the initial treatment advantage of lung adenocarcinoma and small cell lung cancer. For adenocarcinoma, an increasing number of targeted therapeutics have been developed and launched, resulting in more treatment options, which is more obvious for adenocarcinomas with low tumor burdens. Although small cell carcinoma is highly malignant and has a short tumor cell doubling time, it is more sensitive to chemotherapy and radiotherapy. In recent years, immunotherapy has been shown to be associated with an increasing number of significant benefits. Compared with a high tumor burden, a low tumor burden is obviously more likely to lead to longer PFS. In this part of the study, although we concluded that patients with low pretreatment levels of HSP90α had a PFS advantage compared with patients with high pretreatment levels, the opposite results were found in the follow-up study. After 4 cycles of treatment, the PFS of the increased group was significantly better than that of the decreased group. However, in the subgroup analyses based on pathological classification, adenocarcinoma, squamous cell carcinoma, and small cell carcinoma did not significantly differ in terms of PFS. According to our recent results, the increasing group showed PFS benefits after treatment, which warrants further in-depth consideration. After careful analysis, we considered that the PFS advantage in the whole population might be correlated with treatment efficacy. Recent efficacy has largely reduced the tumor burden, which may further affect the difference in PFS. However, no significant differences were detected between pathological subtypes. In the subgroup analysis of therapeutic drugs, we only observed a significant difference in the improvement in PFS benefit in the chemotherapy alone subgroup of the increasing group compared with the decreasing group. On the one hand, these results may be related to the DNA damage, protein misfolding, and other stress states in tumor cells caused by chemotherapy drugs, thereby altering the tumor microenvironment and indirectly leading to the upregulation of Hsp90α expression. On the other hand, after chemotherapy drugs cause tumor cell apoptosis or necrosis, Hsp90α in tumor cells is released into the blood, resulting in an increase in serum HSP90α levels for a period [28].

Our research revealed a significant correlation between HSP90α levels and PFS, which was consistent with the results of previously published mechanistic research. Researchers have reported that Hsp90α drives tumor progression by translocating to the plasma membrane via the EGF-PLCγ1-PKCγ signaling axis. This membrane-localized Hsp90α is essential for promoting cell motility and metastasis, as its neutralization abrogates these behaviors of advanced malignancies [29]. Another study reported that Hsp90α specifically drives tumor migration, invasion, and metastasis via a unique extracellular function determined by a conserved dual-lysine motif (K270/K277), which is absent in Hsp90β. Targeting this motif with a neutralizing antibody inhibits tumor progression, establishing extracellular Hsp90α as a pivotal and specific promoter of malignancy [30]. Mechanistic research in this field is not yet fully mature, and elucidating the exact underlying molecular mechanism remains a key focus of future research.

Many researchers are enthusiastic about the exploration of new tumor markers of lung cancer. Some researchers have reported that when emerging tumor biomarkers are considered, combining classic and emerging lung cancer tumor biomarkers may achieve results that are twice those of current outcomes with half the effort [31]. Therefore, in a follow-up study, we analyzed the combination of HSP90α and CEA and the combination of NSE and SCC to explore their ability to predict PFS. Considering the specificity of classical lung cancer tumor biomarkers in various pathological subtypes, we also divided the biomarkers into different pathological subtypes for research: for adenocarcinoma, CEA was used for correlation analysis [32] for squamous cell carcinoma, SCC was used for correlation analysis [33] and for small cell carcinoma, NSE was used for correlation analysis [34]. This study revealed that the initial HSP90α level was significantly and positively correlated with the corresponding classical biomarkers of various pathological types of lung cancer. The results revealed that in each pathological classification, only the correlation analysis of adenocarcinoma samples revealed significant differences in terms of joint detection, thus indicating that the initial HSP90α and CEA levels were low and that the PFS advantage of these patients was the most obvious. However, in the correlation analysis of SCC and NSE, no significant difference in PFS was found. This phenomenon might be related to differences in the secretion of tumor biomarkers across different pathological types. In both squamous cell carcinoma and small cell carcinoma, the combined detection of HSP90α and classical tumor biomarkers of lung cancer did not yield significant advantages. CEA, which is a specific acidic glycoprotein of human embryonic antigen, is the most widely used tumor biomarker in the diagnosis and treatment of malignant tumors. It has high sensitivity for detecting lung adenocarcinoma, but its specificity is moderate. The combined application of HSP90α and CEA can provide new ideas for the clinical treatment of lung adenocarcinoma and compensate for the deficiency of CEA in the treatment of lung adenocarcinoma to a certain extent [34]. The combination of the emerging tumor marker HSP90α with established biomarkers such as CEA and NSE significantly enhances the predictive value in clinical practice by leveraging their distinct biological pathways. HSP90α offers unique functional insight by reflecting cellular invasion and metastatic processes. This multifaceted approach captures a broader spectrum of tumor activity, leading to a diagnostically superior panel. Evidence has confirmed that the integrated use of these markers achieves a significantly greater area under the curve, sensitivity, and specificity for lung cancer detection than does any single biomarker used in isolation. This synergistic effect is particularly valuable for improving early detection rates and reducing false-negative outcomes in clinically challenging cases, thereby providing a more powerful tool for patient stratification and diagnostic confidence [31].

Although this study initially explored the therapeutic efficacy and prognosis of patients with lung cancer, several limitations were unavoidable. First, this study was a single-center, retrospective study, and most of the data were obtained from existing records. Recall bias in the collection of PFS data from patients’ families or guardians may also exist. Our results may provide ideas and directions for follow-up multicenter, large-sample and high-quality research. Second, the baseline information of the patients included in this study was limited, and some relevant variables were missing. For example, gene mutation and PD-L1 expression were not included as necessary variables in this study. Because information bias may exist, this study did not fully explore the role of serum heat shock protein 90α levels in the diagnosis and treatment of lung cancer. Our research can provide more references for researchers to improve the integrity of information collection in follow-up studies. Third, in vitro or in vivo data to support the exploration of relevant mechanisms are lacking. At the beginning of the design of this study, basic research was not included. After statistical analysis of our data, very convincing results were still difficult to obtain through discussion and subgroup analyses. Our research team will conduct in-depth basic research studies in the future. Building on our findings, future efforts will include first to elucidate the mechanism of extracellular Hsp90α, then validate it in diverse cohorts, and ultimately translate it into clinical practice. This roadmap from bench to bedside is outlined in Fig. 6.

Fig. 6.

A proposed research roadmap for future studies on extracellular Hsp90α in lung cancer, outlining the progression from mechanistic elucidation and clinical validation to clinical translation and application.

Conclusions

On the basis of the present results, we believe that HSP90α is a candidate biomarker with certain potential in the clinical diagnosis and treatment of lung cancer and merits further clinical verification and mechanistic exploration.

Abbreviations

CEA

Carcinoembryonic antigen

CR

Complete response

CT

Computed tomography

DCR

Disease control rate

ELISA

Enzyme-linked immunosorbent assay

HRs

Hazard ratios

HSP90α

Heat shock protein 90α

LUAD

Lung adenocarcinoma

LUSC

Lung squamous cell carcinoma

MRI

Magnetic resonance imaging

NSE

Neuron-specific enolase

ORR

Objective response rate

ORs

Odds ratios

PD

Progressive disease

PFS

Progression-free survival

PR

Partial response

RECIST

Response Evaluation Criteria for Solid Tumors

SCC

Squamous cell carcinoma antigen

SCLC

Small cell lung carcinoma

SD

Stable disease

Author contributions

B.W. contributed to conceptualization and writing—the original draft; Y.M. and Y.Z. contributed to the investigation; J.L. contributed to the statistical analyses; H.L. and N.J. designed and planned the work.

Funding

This study was supported by the Jilin Province Health Science and Technology Capacity Improvement Project (No. 2022LC023).

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request.

Declarations

Ethical approval and consent to participate

This study complies with the Declaration of Helsinki and was approved by the Ethics Review Committee of Jilin Cancer Hospital (Ethics Batch Number: 202208-006-01). Informed consent was obtained from all the individual participants included in the study.

Consent for publication

Informed consent was obtained from all the patients or their families.

Competing interests

The authors declare no competing interests.