Introduction

Lung cancer presents as a heterogeneous malignant disease, characterized by a broad spectrum of clinicopathological features. Approximately 85% of lung cancer cases, classified by pathohistological type, are non-small cell lung cancer (NSCLC), while the remaining 15% are small cell lung cancer (SCLC)1. According to the Global Cancer Epidemiology Database GLOBOCAN 2020, which assesses morbidity and mortality rates across 185 countries, lung cancer stands as the second most prevalent and the deadliest form of cancer worldwide, accounting for approximately 2.2 million new cases and 1.8 million deaths annually2. Moreover, the incidence and hospitalization rates of lung cancer are also increasing. Therefore, this disease represents a serious threat to human health and has emerged as a major public health concern globally3,4,5.

Meanwhile, cardiovascular diseases encompass a range of conditions affecting the circulatory system. This includes ischemic heart disease, cardiomyopathy, heart failure, and arrhythmias. Various risk factors contribute to the development of cardiovascular disease, such as smoking, hypertension, obesity, and infections. These risk factors often aggregate and interact, leading to morbidity and mortality related to cardiovascular disease6. As a leading cause of death worldwide, cardiovascular disease threatens the health of approximately 126 million individuals worldwide, constituting approximately 1.72% of the world’s population. Annually, it claims an estimated 17.9 million lives7. Recent trends indicate a concerning shift toward younger individuals being affected by cardiovascular diseases, while the global prevalence of these conditions continues to rise8.

Despite appearing as unrelated pathologies, cancer and cardiovascular disease are intricately interconnected and often coexist within the same individual. Many individuals facing cancer find themselves vulnerable to cardiac complications, sometimes having to prioritize cancer treatment while heart tissues remain unaddressed. Conversely, cardiovascular disease can be exacerbated by cancer. In recent years, research into both cardiovascular diseases and tumors has delved deeper, revealing a significant correlation between the two. They not only share common risk factors but also influence and promote one another. This research highlights the importance of understanding the intertwined nature of cancer and cardiovascular disease to improve patient outcomes and develop more effective, targeted treatments.

Metabolites serve as the key functional units of biological systems, mediating the intricate interplay between genes, proteins, and environmental factors, thus offering valuable insights into biological processes and their underlying mechanisms. Mass spectrometry imaging (MSI) is a novel molecular imaging technique that combines mass spectrometry analysis and two-dimensional spatial imaging9. Widely employed, MSI provides crucial information regarding the composition and spatial distribution of molecules, including endogenous metabolites and exogenous drugs in tissues. Matrix-assisted laser desorption/ionization (MALDI) mass spectrometry, a fundamental technique within MSI, operates by forming co-crystals between matrix molecules capable of absorbing 337 nm or 355 nm ultraviolet laser and the substance under examination. These matrix molecules absorb laser energy to facilitate ionization of the substance, generating ions at each sampling point upon laser beam irradiation. Subsequently, the tissue sample moves through an XY two-dimensional platform, allowing for separation and detection of the ions by the mass spectrometer. By correlating mass spectra with the spatial position of the sample, MSI enables the visualization of metabolites ranging from 50 to 1300 Da across the tissue sample through software processing10. MSI is suitable for biomarker discovery, tumor research, drug development, and investigations into metabolic abnormalities. Its immense potential extends to clinical research as well as basic life science research, promising advancements in understanding disease mechanisms.

4D-proteomic technology is a cutting-edge generation of proteomics analysis technology, building upon the foundation of 3D proteomic technology, which considers retention time, mass-to-charge ratio (m/z), and ionic strength. The addition of a fourth dimension, ion mobility, which separates precursor ions based on their collision cross-sectional area (CCS), further enhances the analytical capabilities. 4D-DIA quantitative proteomic technology harnesses this innovative approach for protein characterization and quantification. It involves acquiring data on the retention times, mass-to-charge ratios, and ionic strengths, alongside ion mobility, for both parent ions and their fragment ions. This technology represents impressive progress in proteomics analysis, offering invaluable insights into the complex landscape of protein biology.

This study investigates the interconnection between heart and lung cancer as a foundation for exploring alterations in the differential expression of metabolites and proteins using spatial metabolomics and proteomics techniques. Through this focus, the study aimed to uncover potential evidence of association and metabolite alterations between heart and lung cancer. By delving into the relationship between these two conditions and identifying molecular alterations, the study seeks to contribute to the broader understanding of disease mechanisms and pave the way for innovative approaches to disease management and intervention.

Results

Evaluation of A549 lung carcinoma model construction

First, we evaluated tumor growth and progression in mice with A549. After implanting A549 cells, the tumor model exhibited robust construction, with tumor weight showing a steady increase, as shown in Supplementary Fig. 1A-D.

Analysis of mass spectrometry results in spatial metabolomics

Mean mass spectra

The imaging raw data were imported into SCiLS Lab software for processing, which included reading, smoothing, and root mean square normalization. Each tissue’s entire imaging region was selected separately, enabling the extraction of the average Mass spectra. The horizontal coordinate represented the mass-to-charge ratio, while the vertical coordinate represented the average peak intensity values in the region, normalized by the root mean square. The MSI region for each tissue sample in this project are shown in Supplementary Fig. 1E, F.

Spatial distribution of metabolites

All identified target peaks of heart tissues of nude mice in the control and model groups were imaged to visualize the distribution of the target substances on the tissues. Figures 1A-B illustrate the distribution of possible metabolic substances in heart tissues. From the analysis of the figure, it was obtained that different colors represent different relative intensities of the substances in the region, with 0 to 1 indicating a sequential increase in the content of the target substances.

Fig. 1
figure 1

(A,B) Spatial distribution map of metabolites in heart tissue (Pathological staining results are included); (C,D) are the spatial segmentation maps of heart tissue and spatial segmentation region distance thermograms of heart tissue. (E,F) t-statistics and spatial distribution maps of regional characteristic metabolites in heart tissue. In the figure, (A), (C), and (E) are the control group (normal nude mice), and (B), (D), and (F) are the model group (heart tissue in nude mice lung cancer model).

Full size image

Analysis of mass spectrometry imaging data in spatial metabolomics

Spatial segmentation analysis

Spatial segmentation analysis provides an overview of mass spectrometry imaging datasets, facilitating the rapid detection of salient features. This method assesses the mass spectral similarity within a given region, grouping together pixels with similar mass spectral information and assigning them specific colors11. Partitioning information is derived through spatially-aware nearest shrunken centroids clustering (SSC), using intensity data from all identified metabolites. Pixels within the same region exhibit a more consistent metabolic trend12,13. This analysis serves as a reference for subsequent region of interest (ROI) selection (Fig. 1C, D), different colours in the map represent different regions of spatial segmentation. The Euclidean distance between the centroids of different spatial segmentation regions reflects the degree of difference between clusters. A greater distance indicates a larger difference between two regions. Spatially contracted center of mass analysis calculates t-statistics for each target peak in the region of heart tissues. These t-statistics reflect the relationship between the center of mass in each region and the global mean mass spectrogram. A t-statistic greater than 0 indicates relatively high levels of the target substance in that region of heart tissues, while a t-statistic less than 0 indicates relatively low levels. Close to 0 suggest minimal contribution of the target substance to region delineation, while higher absolute t-statistics indicate greater contribution (Fig. 1E, F).

t-distributed stochastic neighbor embedding analysis

The t-distributed stochastic neighbor embedding (t-SNE) algorithm is a nonlinear dimensionality reduction method. In the original high-dimensional space, data point similarity is represented by a Gaussian joint distribution, while in the low-dimensional space, it is represented by a t-distribution14. t-SNE ensures that points close in the high dimensional space remain close in the low-dimensional space, and vice versa, thereby preserving the local features of the data15.

In our study, intensity data for each pixel point in the tissue sample plane, obtained from substance identification, were embedded in 3D space using t-SNE. The fraction of each pixel point on the three components was translated into RGB color coding by linearly varying the red, green, and blue intensities along the three t-SNE axes. The red, green, and blue intensities were adjusted linearly between 0 and 1, corresponding to the minimum and maximum values on the t-SNE axes, respectively (Fig. 2A-B)16.

Fig. 2
figure 2

(A,B) is t-SNE hyperspectral visualization and t-SNE 3D scatter plots in heart tissue; (C,D) is UMAP hyperspectral visualization and UMAP 3D scatter plot in heart tissue; (E,F) is spatial distribution map of metabolite co-localization analysis in heart tissue. (A), (C), and (E) are control groups, and (B), (D), and (F) are model groups.

Full size image

Uniform manifold approximation and projection analysis

Uniform manifold approximation and projection (UMAP) is a nonlinear dimensionality reduction technique rooted in the framework of Riemannian geometry and algebraic topology theory16,17. In contrast to t-SNE’s random initialization, UMAP uses spectral clustering and graph partitioning methods for nearest-neighbor graph initialization in low-dimensional space. The intensity data of all identified metabolites at each pixel point in the tissue sample plane were projected into 3D space using UMAP (Fig. 2C, D)18,19.

Metabolite co-localization analysis

Metabolite co-localization analysis seeks to identify metabolites with spatial distributions of content that align consistently with the target metabolite. This analysis uncovers connections between metabolites and reveals potential molecular interactions. Typically, Pearson’s correlation coefficient and Manders’ colocalization coefficient are used to quantify co-localization patterns (Fig. 2E, F)20.

Evaluation and analysis of metabolite data in spatial metabolomics

The composition of metabolites varies between samples, with different sample types containing distinct metabolite classes and proportions. Metabolite composition ratio analysis provides a comprehensive assessment of the distribution of major metabolites in the samples; each color represents a metabolite category, and the area of the color block indicates the proportion of the category, as shown in the ring diagram of metabolite category composition in Fig. 3A.

Fig. 3
figure 3

(A) Metabolite category composition ring; (B,C) PCA score plots of mass spectrometry data for each group of samples in 2D and 3D; (D) overall sample clustering plot; (E) inter-sample correlation plots.

Full size image

Before conducting the analysis of variance, PCA was performed on the grouped samples to compare variances, enabling the observation of variance magnitudes between subgroups and within samples of each group21,22. As depicted in Fig. 3B, C, a notable degree of dispersion and substantial variation between groups was observed. These findings suggest discernible differences in the metabolomes within the sample groups. Clustered heat map analysis was performed on all samples and the clustered heat maps were plotted using the R program script, as in Fig. 3D. Biological replicates between samples within groups could be observed by correlation analysis between samples. Also the higher the correlation coefficient of intra-group samples relative to inter-group samples, the more reliable the differential metabolites obtained, as in Fig. 3E.

Orthogonal partial least squares discriminant analysis

The OPLS-DA model validation is shown in Fig. 4A. Figure 4B shows the S-plot generated from OPLS-DA analysis. The horizontal axis represents the covariance between principal components and metabolites, while the vertical axis indicates the correlation coefficient between principal components and metabolites. Metabolites closer to the upper right and lower left corners signify greater significance in differences. Red dots indicate metabolites with a VIP value exceeding 1, while green dots represent metabolites with a VIP value of 1 or less.

Fig. 4
figure 4

(A) OPLS-DA validation; (B) OPLS-DA S-plot; (C) differential multiplicity histogram; (D) radar plot of differential metabolites; (E) volcano plot of differential metabolites; (F) sample level clustering tree.

Full size image

Screening for differential metabolites

In the absence of MS/MS fragmentation spectral validation (including known internal standard fragmentation patterns), metabolite identification results should be considered tentative or speculative.Metabolites with a VIP value greater than 2 were selected, as outlined in Table 1.Following qualitative and quantitative analyses of the detected metabolites, multiplicative changes in the quantitative information of the metabolites occurring in each subgroup were compared to specific sample subgroups. Figure 4C-E shows the results of the top 20 metabolites with multiplicity of difference in each group comparison and the top 10 metabolites with the largest absolute value of log2FC were selected for radar charting, and the relative content differences in the two groups of samples, respectively. Hierarchical cluster analysis was performed on the samples of different comparison groups, as shown in Fig. 4F samples clustered in the same cluster have higher similarity between samples.

Table 1 Spatial metabolomics differential metabolite basic information.
Full size table

Correlation analysis of differential metabolites

Different metabolites exhibit synergistic or mutually exclusive relationships with each other, and correlation analysis serves to measure the degree of metabolic closeness (metabolic proximities) between significantly different metabolites. This analysis is instrumental in further understanding the inter-regulatory relationships among metabolites during biological state changes, as shown in Fig. 5A, B. The correlation analysis of the differential metabolites identified according to the screening criteria was performed by Pearson’s correlation analysis method and the results are shown in Fig. 5C below.

Fig. 5
figure 5

(A,B) Differential metabolite chord diagrams and differential metabolite correlation network diagrams; (C) differential metabolite correlation heatmap; the names of differential metabolites are shown both horizontally and vertically. (D) Differential metabolite KEGG enrichment map. (E) Difference statistics of each subgroup; (F) heatmap of clustering of difference proteins; (G) heatmap of clustering of all difference proteins.

Full size image

Kyoto Encyclopedia of Genes and Genomes functional enrichment analysis of differential metabolites

Among the KEGG enrichment results, pathways such as metabolic pathways, choline metabolism in cancer, glycerophospholipid metabolism, and metabolism of xenobiotics by cytochrome P450 were identified, as shown in Fig. 5D.

Analysis of proteomics results

Identification and quantification of differential proteins

The screening criteria for identifying significant differential proteins in this study were as follows:

  1. (1)

    For repeated items: In two-group comparisons, proteins with a FC ≥ 1.5 or FC ≤ 0.6667 and a p-value ≤ 0.05 were defined as differential significant proteins. In comparisons involving more than two groups, significance was determined based solely on a p-value ≤ 0.05.

  2. (2)

    For non-duplicated items: In two-group comparisons, proteins with a FC ≥ 1.5 or FC ≤ 0.6667 were defined as differentially significant proteins.

The statistical results of differential proteins in all groups are presented in Table 2 and Fig. 5E. In order to facilitate the observation of the expression pattern of different differential proteins in different samples, the differential proteins were normalised by z-score and plotted in a clustered heat map, see Fig. 5F, G.

Table 2 Differential protein screening results.
Full size table

Gene Ontology enrichment analysis of differentially expressed proteins

Gene Ontology is an international standard classification system for gene function. It aims to establish a language and vocabulary standard that is applicable to a wide range of species, qualifies and describes the functions of genes and proteins, and can be updated as research progresses. GO is divided into three parts: molecular function, biological process, and cellular component. The number of differentially expressed proteins annotated to all secondary GO entries under the three primary classifications was counted, and the GO classification statistics of differentially expressed proteins are shown in Fig. 6A. The number of differentially expressed proteins annotated to all secondary GO entries under the three primary categories was counted, and the GO enrichment results of up- and down-regulated differentially expressed proteins are shown in Fig. 6B.

Fig. 6
figure 6

(A) Histogram of GO classification; (B) histogram of upward and downward adjustments of GO enrichment; (C) KEGG classification bar graph; (D) KEGG classification up- and down-regulation comparison bar graph; (E) KOG annotation bar graph, (F) KOG annotation up- and down-regulation comparison bar graph.

Full size image

Kyoto Encyclopedia of Genes and Genomes enrichment analysis of differentially expressed proteins

Pathway analysis enabled the identification of the most prominent biochemical metabolic pathways and signal transduction pathways in which proteins are involved. The identified proteins were compared with the KEGG database to obtain KEGG annotation results, and the number of differentially expressed proteins contained in each KEGG pathway was counted and plotted on a bar graph, as in Fig. 6C, D.

Eukaryotic orthologous groups enrichment analysis of differentially expressed proteins

Eukaryotes were annotated with KOG. This project was annotated with the KOG database, and the number of differentially expressed proteins contained in each KOG entry was counted and plotted in a bar graph, as shown in Figs. 6E, F.

Differently expressed protein interaction network

Using StringDB (http://string-db.org/) protein interaction database for protein protein interaction (PPI) analysis, if there is a corresponding species in the database, the protein sequence of the corresponding species will be extracted directly, if not, the protein sequence of the near source species will be extracted, the The protein sequences of differentially expressed proteins were compared with the extracted sequences by blast, and then the interactions of differentially expressed proteins were extracted according to the confidence score > 400 (medium confidence), and the results of the PPI interactions network analysis are shown in static Fig. 7A. In addition, a clustering tree was constructed based on the correlation of expression among proteins and modules were divided by WPCNA analysis. If certain proteins always have similar expression changes in a physiological process or different tissues, then the colours are similar, as shown in Fig. 7B. heatmap analysis of correlation between samples and modules is shown in Fig. 7C.

Fig. 7
figure 7

(A) Interaction network graph of differentially expressed proteins; (B) hierarchical clustering tree graph of the modules; (C) heatmap of correlation between the samples and the modules; (D) combined analysis of Sankey plots by proteomics and spatial metabolomics.

Full size image

Discussion

Lung cancer is one of the most common malignant tumors worldwide, with the highest incidence and mortality rate2. At the same time, cardiovascular disease (CVD) is one of the leading causes of death globally23. Cancer has replaced cardiovascular disease as the leading cause of death in developed countries, according to a new study in The Lancet. According to the study, cancer will become the "number one health killer" worldwide in the coming decades24. While traditionally viewed as distinct conditions, numerous studies have revealed a bidirectional relationship between them25. Both diseases share a reliance on energy metabolism for sustained cardiac contractility and tumor proliferation. Biomarkers such as atrial natriuretic peptide and CA-125 have been identified as predictors for both heart failure and malignancy26,27,28. Further, epidemiological data indicates a notably higher incidence of cardiovascular disease among cancer patients, with cardiovascular complications being a leading cause of mortality in this population. Conversely, individuals with cardiovascular issues appear to have a heightened risk of cancer diagnosis29,30,31,32,33. In addition, common pathogenic mechanisms and shared risk factors link these two conditions. Elevated levels of inflammatory cytokines, including C-reactive protein, TNF-α, interleukins, and macrophage migration inhibitory factor, have been observed during lung cancer development, contributing to cardiovascular diseases such as atrial fibrillation and heart failure34. In patients with malignant tumors, the physical compression of primary tumor lesions, ectopic mechanical pull, invasion of surrounding tissues by tumor cells, and secretion of various substances like 5-hydroxytryptamine and antidiuretic hormone can disrupt the normal function of the autonomic nervous system, particularly leading to sympathetic excitation and consequent electrophysiological changes in the atria, potentially triggering atrial fibrillation35. Additionally, lung-cancer associated intracellular oxidative stress and the accumulation of reactive oxygen species, coupled with insufficient oxygen supply, form a vicious cycle promoting vascular endothelial smooth muscle proliferation and plaque formation36.

A cohort study, using data from primary care practices, hospitals, and cancer registries linked to the UK Clinical Practice Research Datalink, found an elevated risk of certain cardiovascular diseases among cancer survivors when compared to the general population, after adjusting for shared risk factors. Notably, survivors of various cancers, including hematological neoplasms (non-Hodgkin’s lymphoma, leukemia, multiple myeloma), esophageal, lung, renal, and ovarian cancers, exhibited an increased risk of heart failure or cardiomyopathy37. This finding corroborates previous epidemiological studies highlighting the interconnection between tumors and cardiovascular diseases38. Elsewhere, de Boer et al.39 demonstrated that several known tumor markers hold prognostic significance in patients with heart failure, indicating disease severity and progression. Among these biomarkers, CA125, CYFRA 21–1, CEA, and CA19-9 exhibited independent prognostic value, predicting all-cause mortality. The study further demonstrated that several tumor biomarkers were strongly associated with disease severity in heart failure. The study suggests that the presence of signaling pathways and pathological signals, mirrored by tumor biomarkers, are implicated in heart failure and correlate with its severity and prognosis39.

With the advancement of medical understanding, there is a growing recognition of the simultaneous occurrence of cardiovascular disease and cancer. However, few studies have been conducted on their correlation, with most focusing on the relationship between the development of cardiovascular disease and chemotherapeutic agents used in oncology8. In cardiovascular studies, the revelation that heart disease triggers the secretion of numerous proteins has inspired the utilization of biomarkers for disease diagnosis and piqued scientific interest in risk stratification and targeted therapies for patients. Nevertheless, in reality most of these circulating factors are biologically active proteins that can impact surrounding organs, including tumors, such as TNF-α, IL-6, IL-1, and vascular endothelial growth factor40. Yet, it remains unclear whether the heart functions as a genuine endocrine organ.

This study explored differential metabolites and differential protein expression in the heart by establishing a nude mouse lung cancer model using spatial metabolomics and proteomics approaches. The results of mass spectrometry analysis in spatial metabolomics identified 2,892 substances, with 514 at the secondary level. Meanwhile, spatial imaging of metabolites was performed on the target peaks (Fig. 2E, F), revealing distinct patterns across different heart regions, notably between the control and model groups. The results of PCA and OPLS-DA revealed a distinct dispersion pattern between the control and model groups, indicating significant differences. 28 possible metabolites have been tentatively identified., including Gln-His-Val-Glu, LysoPC 22:6, LPC (20:2/0:0), and (1R,4aR,4bS,6aS,9R,10R,10aS)-1-(carboxymethyl)-10-hydroxy-2-(1-methoxy- 2-methyl-1-oxopropan-2-yl)-1,4a,4b,9,10-pentamethyl-3,4,5,6,7,8,9,10a,12,12a-decahydro-2H-chrysene-6a-carboxylic acid, were identified, mainly involving amino acids, metabolites, and glycerophospholipids. KEGG enrichment highlighted pathways such as metabolic pathways, choline metabolism in cancer, glycerophospholipid metabolism, and metabolism of xenobiotics by cytochrome P450. These findings suggest altered cardiac metabolism in nude mice following the establishment of a lung cancer environment, potentially related to amino acid and lipid metabolism, which may be linked to shared pathogenesis between the cardiovascular diseases and lung cancer. At the same time, in an immunocompetent mouse model of lung cancer, parameters of cardiovascular disease may manifest as cardiac dysfunction, as assessed by echocardiography, which reveals a significant reduction in cardiac parameters such as cardiac ejection fraction (EF) and shortening fraction (FS), suggesting a decrease in the heart’s pumping capacity41. Elevated inflammatory factors (e.g., TNF-α, IL-6) released in the tumor microenvironment may affect the heart through blood circulation, leading to myocardial inflammation and fibrosis41,42. Increased oxidative stress may damage cardiomyocytes by generating excess reactive oxygen species (ROS), further leading to cardiac dysfunction. In addition, lung cancer may further exacerbate cardiac inflammation and fibrosis by altering the distribution and function of immune cells, leading to increased infiltration of immune cells (Treg cells, monocytes) in the heart41,43,44. These changes may be associated with factors released directly from the tumor, systemic inflammatory responses, and metabolic disturbances.

The proteomics analysis identified 582 differential proteins, among which 19 were associated with enriched KEGG pathways such as the IL-17 signaling pathway, protein digestion and absorption, cell adhesion molecules, influenza A, toxoplasmosis, TNF signaling pathway, and NF-kappa B signaling pathway. Notable proteins included Mknk1, Trafd1, Dab2ip, Tab1, Ripk3, G3PDH, and Mapk15. RIPK3, a receptor interacting protein kinase, forms necrotic vesicles with RIPK1 and activates mixed lineage kinase domain-like protein (MLKL), which induces programmed cellular necrosis, characterized by calcium inward flow and plasma membrane damage. As such, RIPK3 plays an important role in both inflammation and tumorigenesis45,46. Mknk1, a member of the MAPK interacting kinase family, regulates various physiological and pathological processes such as cell growth, differentiation, stress, and inflammatory response47. These differential proteins play a key role in cardiovascular injury, underscoring the potential impact of establishing a lung cancer environment on the heart, leading to cardiovascular injury. Proteomics and spatial metabolomics together demonstrated that it was mostly associated with the TNF, cancer-related, and HIF-1 signaling pathways, providing the groundwork for determining possible targets for treatment (Fig. 7D).

The intricate relationship between cardiovascular disease and lung cancer emphasizes the need for vigilant monitoring of cardiac function and toxicity in patients of all ages, including those with lung cancer and other tumors48. Early detection and intervention for cardiovascular injuries before, during, and after treatment are crucial. Additionally, patients with lung cancer or related tumors may benefit from cardioprotective therapies alongside oncological treatments, including the adoption of heart-healthy behaviors such as regular exercise, following a healthy diet, and managing stress. For patients with long-term survival and a young age of disease onset, timely improvement of cardiac function is essential to prevent potential cardiovascular injury, ultimately enhancing survival rates, improving quality of life, and alleviating the economic burden on families.

In conclusion, this study highlights the inextricable connection between lung cancer and cardiovascular disease. The findings suggest that the establishment of a lung cancer environment in nude mice leads to significant changes in both metabolic and protein profiles within their heart tissue, reflecting the intricate crosstalk between cancer biology and cardiovascular physiology. This observation underscores the potential implications for future treatment and intervention strategies for both diseases, potentially facilitating the development of novel therapies aimed at mitigating cardiovascular complications in cancer patients. However, there remains a scarcity of large-scale clinical studies investigating the incidence of cardiovascular disease in lung cancer populations, and the underlying mechanism of action is still poorly understood both in China and internationally. Additionally, the absence of a longitudinal viewpoint to monitor the dynamic evolution of molecular characteristics and metabolic alterations throughout tumor growth or therapy is a significant drawback of our investigation. The study restricts our knowledge of how tumor alterations occur over time and the possible connection between treatment response and tumor biological activity. A longitudinal strategy should be used in future research to sample and analyze data at several periods in time in order to capture dynamic changes during tumor growth or therapy. Deeper understanding will be possible for the creation of individualized treatment plans thanks to this longitudinal viewpoint. Therefore, further research is warranted to provide additional evidence-based insights and deeper exploration regarding the mechanisms responsible for the observed alterations in cardiac metabolism and protein expression.

Materials and methods

Animal culture

Animal experiments were conducted in accordance with the guidance outlined by the Experimental Animal Ethics Committee of the Affiliated Hospital of Inner Mongolia Minzu University, under approval number NM-LL-2024-03-15-01. Five-week-old male athymic mouse were purchased from the Changsheng Experimental Animal Centre of Liaoning Province(SCXK(辽)2020-0001), China, and were subsequently housed in the specific pathogen-free grade experimental animal facility at Inner Mongolia University for Nationalities, maintaining environmental conditions at 23 ± 1℃, 55 ± 5% humidity, and 12-h light–dark cycles. Throughout the experimental period, the mice had ab libitum access to food, water, and activities. The experimental design and animal husbandry are in accordance with the ARRIVE guidelines.

Cell culture

A549 cells were obtained from the Cell Bank of the Zhongqiao Xinzhou (Shanghai, China) and were cultured in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 1% (v/v) antibiotic–antimycotic (100 U/mL penicillin and 100 U/mL streptomycin) and 10% (v/v) fetal bovine serum. The cells were maintained at 37℃ under a humidified atmosphere containing 5% carbon dioxide.

Tumor growth in vivo, animal modeling and grouping

Mice were acclimatized to the feeding environment for one week. A549 cells in the logarithmic growth phase, with 80% confluence, were collected, and the cell count was manually determined using a cell counter plate. Phosphate-buffered saline (PBS) was added to resuspend the cells, resulting in a cell suspension with a density of 1 × 106/100μL, which was then kept on ice for future use. The skin in the middle and posterior part of the right axilla of the mice was prepared, and the area was disinfected with 75% ethanol. Subsequently, 100 μL of the cell suspension was injected subcutaneously into the designated area using a 1 mL syringe. After injection, the needle was withdrawn slowly, and gentle pressure was applied with a cotton swab to the injection site for 30 s to prevent leakage of the cell suspension. Tumor volume was calculated, and modeling was successful when the average tumor volume of the model group reached 130mm3. The formula was applied: V (mm3) = Dd2/2 (D is the long diameter of the tumor, d is the short diameter of the tumor). Tumor size was measured every other day using digital vernier calipers49. Tumor size was measured every other day using digital vernier calipers. Following measurement, the mice were randomly divided into two groups, each comprising 6 mice. After analyzing the differences in body weight and tumor size between the groups, the mice were numbered and allocated to cages accordingly.

Spatial metabolomics experimental procedures

Sample preparation

After the final gavage, mice were anaesthetised with isoflurane and blood and hearts were taken. Frozen tissue samples were fixed in three drops of distilled water during the cutting stage. The tissues were then sectioned to a thickness of 12 μm using a Leica CM1950 cryostat (Leica Microsystems GmbH, Wetzlar, Germany) at − 20 °C. Following sectioning, the tissue sections were arranged in groups on electrically conductive slides coated with indium tin oxide (ITO). The slides were then dried in a vacuum desiccator for 30 min.

Matrix coating

Desiccated tissue sections mounted on ITO glass slides were sprayed using an HTX TM sprayer (Bruker Daltonics, Germany) with a solution of 15 mg/mL 2,5-dihydroxybenzoic acid (DHB) dissolved in 90% acetonitrile and 10% water. The sprayer temperature was maintained at 60 °C, with a flow rate of 0.12 mL/min and a pressure of 5 psi. Thirty passes of the matrix were applied to the slides, with 5 s of drying time between each pass.

Mass spectrometry imaging

MALDI timsTOF MSI experiments were performed using a prototype Bruker timsTOF flex MS system (Bruker Daltonics, Bremen, Germany), which was equipped with a 10 kHz smartbeam 3D laser. The laser power was initially set to 90% and maintained at this level throughout the entire experiment. Mass spectra were acquired in positive mode, covering a mass range from m/z 50 to 1300 Da. The imaging spatial resolution was set to 50 μm for tissue samples, and each spectrum was generated from 400 laser shots. MALDI mass spectra were normalized using the root mean square method, and the signal intensity in each image was shown as the normalized intensity. For further detailed structural confirmation of the identified metabolites, MS/MS fragmentations were performed on the timsTOF flex MS system in the MS/MS mode.

4D-DIA quantitative proteomics experimental procedure

Protein extraction

Samples were initially ground using liquid nitrogen, and the resulting powder was transferred to a 1.5 ml centrifuge tube. Subsequently, the samples underwent sonication three times on ice, using a high-intensity ultrasonic processor in a lysis buffer consisting of 8 M urea supplemented with 1 mM phenylmethylsulfonyl fluoride (PMSF) and 2 mM ethylenediaminetetraacetic acid (EDTA). Following sonication, any remaining debris was removed by centrifugation at 15,000 g at 4 °C for 10 min. Finally, the protein concentration was determined using a bicinchoninic acid (BCA) kit, following the manufacturer’s instructions.

Digestion and cleanup

Equal amount of proteins from each sample were subjected to tryptic digestion. To the supernatants, 8 M urea was added to a final volume of 200ul, followed by reduction with 10 mM dithiothreitol (DTT) for 45 min at 37 °C, and alkylation with 50 mM iodoacetamide (IAM) for 15 min in a dark room at room temperature. The mixture was then precipitated by adding four times the volume of chilled acetone and incubating at -20 °C for 2 h. After centrifugation, the protein precipitate was air-dried and resuspended in 200ul of 25 mM ammonium bicarbonate solution and 3ul of trypsin (Promega), and digested overnight at 37 °C. Following digestion, peptides were desalted using a C18 Cartridge, dried with a vacuum concentration meter, concentrated by vacuum centrifugation, and redissolved in 0.1% (v/v) formic acid.

Liquid chromatography-tandem mass spectrometry detection

Mobile phase A consisted of 0.1% formic acid aqueous solution, while phase B comprised a 0.1% formic acid acetonitrile solution (100% acetonitrile). Samples were uploaded by an autosampler onto an analytical column (IonOpticks, Australia, 25 cm × 75 μm, C18 packing 1.6 μm) for separation. The temperature of the analytical column was maintained at 50 °C using an integrated column oven. A volume of 200 ng of sample was loaded at a flow rate of 300 nL/min, and the separation gradient spanned 40 min. The liquid-phase gradient proceeded as follows: from 0 to 25 min, the proportion of liquid B increased from 2 to 22%; from 25 to 30 min, liquid B increased from 22 to 35%; from 3o to 35 min, liquid B increased from 35 to 80%; and from 35 to 40 min, liquid B was maintained at 80%.

Detection using the timsTOF Pro2 mass spectrometer involved initial separation by chromatography, followed by acquisition of mass spectrometry data in dda-PASEF mode for dia-PASEF acquisition method to establish a suitable acquisition window. For the analysis, an effective gradient of 40 min was employed, with positive ion detection mode. The parent ion scan range spanned from 100 to 1700 m/z, while ion mobility 1/K0 ranged from 0.85 to 1.3 V⋅s/cm2. Ion accumulation and release time was set to 100 ms, with nearly 100% ion utilization rate. The capillary voltage was maintained at 1500 V, drying gas rate at 3 L/min, and drying temperature at 180 °C. Parameters for the dda-PASEF acquisition mode included four MS/MS scans with a total cycle time of 0.53 s, a charge range of 0–5, and dynamic exclusion time of 0.4 min. Ion target intensity was set to 10,000, with an ion intensity threshold of 1500. Collision energy increased linearly with ion mobility, with 1/K0 at 0.85 Vs/cm2. The collision energy was set at 27 eV when m/z < 700, with quadrupole isolation width set to 2Th. When m/z > 800, the quadrupole isolation was set to 3Th. For the diaPASEF acquisition mode, parameters included a mass range of approximately 400–1200, mobility range of 0.85–1.3 V⋅s/cm2, mass width of 25 Da, and mobility range of 0.85–1.3 V⋅s/cm2. In this mode, the acquisition parameters involved a mass overlap of 0.1, 24 mass steps per cycle, and 2 mobility windows, resulting in a total of 48 acquisition windows. The average acquisition period was 1.17 s.

Mass spectrometry data analysis

Database searching is a complex computational process that necessitates the use of specialized mass spectrometry data analysis software for data parsing. In this study, DIA-NN (v1.8.1) was employed as the library search software for analyzing data-independent acquisition (DIA) mass spectrometry data. The library was searched using the Libraryfree method, with the following search parameters:

  • Database: Uniprot-proteome_UP000000589_Mus_musculus.fasta database, containing a total of 55,319 sequences.

  • Deep learning parameters were activated to predict a spectral library.

  • Match-between-runs was enabled to generate a spectral library using the DIA data, and this library was subsequently used to reanalyze the DIA data for protein quantification.

  • Both precursor ions and protein-level false discovery rates (FDRs) were filtered at 1%, ensuring high confidence in the identified proteins.

Protein quantification

Enhancing the quantification principle through library search software

Protein quantification through the DIA-NN software was achieved using the MaxLFQ algorithm, operating on the following principles:

1. In a study with N experiments (samples) and a protein with M quantitative peptides (Razor + Unique), the initial step involves calculating the logarithmic ratio of intensities for each peptide p ∈ [1, M] between samples i and j, as follows:

$$r_{i,j} = \log \frac{{I_{i} \left( p \right)}}{{I_{j} \left( p \right)}} = \log I_{i} \left( p \right) - \log I_{j} \left( p \right)$$

where Ii(p) denotes the signal intensity of peptide p in sample i. Logarithmic values are not calculated if the peptide ion is absent in either sample i or j.

2. The linear relationship of a protein between two samples is expressed by the median of the logarithmic ratio of intensities from corresponding specific peptide segments, calculated as follows:

$$x_{i} - x_{j} = Median\left( {r_{i,j} \left( p \right)} \right)$$

where xi denotes the log-transformed protein intensity.

3. Given N experiments (samples), Equation 2 is represented as a matrix: Ax = b, where Ax can be expressed as follows:

$$\begin{array}{*{20}c} {A_{i,j} = \left\{ {\begin{array}{*{20}c} { - 1\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,i \ne j} \\ {\sum\nolimits_{i = 1}^{N - 1} {1\left( {i,j} \right)\,\,\,\,i = j} } \\ \end{array} } \right.} \\ {x = \left[ {\begin{array}{*{20}c} {x_{1} } \\ \vdots \\ {x_{N} } \\ \end{array} } \right]} \\ {b_{i} = \left\{ {\begin{array}{*{20}c} {\sum\nolimits_{j = i + 1}^{N} {Median\left( {r_{i,j} \left( p \right)} \right)\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,i = 1} } \\ {\sum\nolimits_{j = i + 1}^{N} {Median\left( {r_{i,j} \left( p \right)} \right) - \sum\nolimits_{j = 1}^{i} {Median\left( {r_{j,i} \left( p \right)} \right)\,\,\,\,\,\,\,i > 1} } } \\ \end{array} } \right.} \\ \end{array}$$

where 1 (i, j) equals 1 when the peptide is present in both samples i or j, and 0 otherwise. Equation 2 is efficiently solved using Cholesky decomposition to determine the log-transformed protein intensity xi. Subsequently, the protein intensity in experiment i is represented as exi.

In projects with only 1 experiment (sample), where the MaxLFQ algorithm is not applicable, the software automatically resorts to the Top-N algorithm for quantitative analysis completion.

Standardization

After the library search software completed the protein quantification, the next step involved extracting the intensity of each protein in different samples as provided in the library search results. Subsequently, an in-sample normalization process was conducted using centroid transformation to obtain the relative quantification value (R) of the protein in different samples. The calculation formula is as follows:

$$R_{i,j} = {{I_{ij} } \mathord{\left/ {\vphantom {{I_{ij} } {Median\left( {I_{i} } \right)}}} \right. \kern-0pt} {Median\left( {I_{i} } \right)}}$$

where i denotes the sample and j denotes the protein.

Variance analysis

Following standardization, the differential quantitative analysis of proteins was required to identify proteins exhibiting differential expression among samples in each group during the differential grouping process. Commonly employed statistical methods for proteomics difference analysis include parametric and non-parametric tests, with the specific choice contingent upon the nature of the data. For experiments with biological replicates, when the differential grouping entails two groups of samples, the mean ratio of quantitative values for each protein across all biological replicates serves as the multiplicity of difference (fold change, FC). Subsequently, a t-test was performed using the quantitative values of each protein in the two groups of samples to calculate the corresponding p-value. In cases where the differential grouping involves more than two groups of samples, a t-test is conducted using the quantitative values of each protein in each group of samples, or alternatively, an analysis of variance (ANOVA) test is employed, with both methods generating relevant p-values. For experiments lacking biological replicates, only the FC can be calculated when the differential grouping involves two samples, and p-values cannot be calculated. If a p-value hypothetical test is required, the FDR is typically calculated using the Benjamini–Hochberg method.

Bioinformatic analysis

To thoroughly understand the functional properties of different proteins, we conducted comprehensive enrichment analyses on both the identified proteins and the differential proteins in each comparison group. The differential proteins in each group underwent enrichment analysis at four levels: Gene Ontology (GO) classification, Eukaryotic Orthologous Groups (KOG) functional classification, Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis, and protein structural domains. Enrichment analysis was performed using hypergeometric calculation to determine the significance of the enrichment p-value, with the aim being to identify whether the differential proteins exhibit a significant enrichment trend for specific functional categories relative to the background proteins (in this case, all identified proteins). Additionally, predictive analyses of subcellular locations and signal peptides were conducted to further elucidate the physiological functions in which the differential proteins are involved.

KEGG pathway enrichment analysis was performed based on the results of differential metabolites. The Rich Factor, representing the ratio of the number of differential metabolites in the corresponding pathway to the total number of metabolites annotated to the pathway, indicates the degree of enrichment, with a higher value indicating greater enrichment. The p-value is determined through the hypergeometric test, calculated using the following formula:

$$P = 1 - \sum\limits_{i = 0}^{m - 1} {\frac{{\left( {\begin{array}{*{20}c} M \\ i \\ \end{array} } \right)\left( {\begin{array}{*{20}c} {N - M} \\ {n - i} \\ \end{array} } \right)}}{{\left( {\begin{array}{*{20}c} N \\ n \\ \end{array} } \right)}}}$$

where N is the total number of proteins with GO annotation information in the background proteins, n is the number of differentially expressed proteins in N, M is the number of proteins in the background proteins annotated to a specific GO entry, and m is the number of differentially expressed proteins annotated to a specific GO entry.

Statistical analysis

Unsupervised principal component analysis (PCA) was performed using the prcomp function in R (www.r-project.org). Prior to analysis, the data underwent unit variance scaling. Hierarchical cluster analysis (HCA) results for both samples and metabolites were visualized as heatmaps with dendrograms. Pearson correlation coefficients (PCC) between samples were calculated using the cor function in R and displayed as heatmaps. Both HCA and PCC analyses were carried out using the ComplexHeatmap package in R. For HCA, normalized signal intensities of metabolites (unit variance scaling) were visualized as a color spectrum. In two-group comparisons, differential metabolites were determined based on variable importance in projection (VIP) (VIP > 1) and p-value (p-value < 0.05, Student’s t-test). VIP values were extracted from orthogonal partial least squares discriminant analysis (OPLS-DA) results, which also included score plots and permutation plots. The OPLS-DA model was constructed using the MetaboAnalystR package in R with log2-transformed and mean-centered data. To prevent overfitting, a permutation test (200 permutations) was performed. Identified metabolites were annotated using the KEGG Compound database (http://www.kegg.jp/kegg/compound/), and annotated metabolites were then mapped to the KEGG Pathway database (http://www.kegg.jp/kegg/pathway.html). Pathways with significantly regulated metabolites were subjected to metabolite sets enrichment analysis (MSEA) using the hypergeometric test’s p-values to determine significance.