https://doi.org/10.1007/s00392-025-02625-4
1IKDT - Institut Kardiale Diagnostik und Therapie GmbH Berlin, Deutschland; 2Deutsches Herzzentrum der Charite (DHZC) Klinik für Kardiologie, Angiologie und Intensivmedizin | CBF Berlin, Deutschland
Background
Cardiac sarcoidosis (CS) is an inflammatory disease of the heart which is present in around a quarter of patients with systemic sarcoidosis. Diagnosis of CS remains challenging due to unspecific symptoms of the condition.
Even indications for CS can be obtained by non-invasive measurements, it can only be reliably identified by histological examination of endomyocardial biopsies (EMBs).
However, due to the focal nature of CS, diagnosis via histological examination still contains a degree of uncertainty due to a possible sampling error.
This inaccuracy raises the need for improved diagnostic methods.
Here, we investigate gene expression changes characteristic for CS with the goal of developing a gene expression profile (GEP) to clearly diagnose the presence of CS.
Even indications for CS can be obtained by non-invasive measurements, it can only be reliably identified by histological examination of endomyocardial biopsies (EMBs).
However, due to the focal nature of CS, diagnosis via histological examination still contains a degree of uncertainty due to a possible sampling error.
This inaccuracy raises the need for improved diagnostic methods.
Here, we investigate gene expression changes characteristic for CS with the goal of developing a gene expression profile (GEP) to clearly diagnose the presence of CS.
Methods
Expression of 112 inflammatory genes were analyzed in EMBs from N = 20 patients with CS and compared to N = 57 patients with inflammatory cardiomyopathy (DCMi) and N = 56 patients with giant cell myocarditis (GCMC).
We identified the top 50 genes with the highest discriminatory power between DCMi/GCMC and CS via analysis of the area under the receiver operator curve (AUC).
Top genes were used to develop GEPs based on a decision-tree approach and a linear model with varying numbers of predictors.
We identified the top 50 genes with the highest discriminatory power between DCMi/GCMC and CS via analysis of the area under the receiver operator curve (AUC).
Top genes were used to develop GEPs based on a decision-tree approach and a linear model with varying numbers of predictors.
Results
We aim to find the smallest possible gene sets which allow for a highly accurate diagnosis of CS.
Individual marker genes such as CXCL9 (AUC = 0.85), CCL17 (AUC = 0.83) and GSN (AUC = 0.82) were highly accurate to discriminate CS from DCMi controls. Other genes such as CYBB (AUC = 0.82), TLR2 (AUC = 0.81) and TIMP1 (AUC = 0.81) were suitable to discriminate between CS and GCMC.
Different GEPs using 3 to 10 genes were developed on the basis of decision tree models and linear regression and achieve AUCs of up to 0.94 (Sensitivity: 90.9%, Specificity: 84.8%) in a test data set.
Decision trees require 7 genes to resolve all patients in test data both in relation to DCMi and GCMC.
Individual marker genes such as CXCL9 (AUC = 0.85), CCL17 (AUC = 0.83) and GSN (AUC = 0.82) were highly accurate to discriminate CS from DCMi controls. Other genes such as CYBB (AUC = 0.82), TLR2 (AUC = 0.81) and TIMP1 (AUC = 0.81) were suitable to discriminate between CS and GCMC.
Different GEPs using 3 to 10 genes were developed on the basis of decision tree models and linear regression and achieve AUCs of up to 0.94 (Sensitivity: 90.9%, Specificity: 84.8%) in a test data set.
Decision trees require 7 genes to resolve all patients in test data both in relation to DCMi and GCMC.
Conclusion
The GEPs presented here enables the diagnosis of CS on a molecular level with a high diagnostic accuracy and can thus circumvent the sampling error in EMB analyses between inflammatory cardiomyopathy and sarcoidosis.
This is highly relevant as it significantly improves diagnostics and the resulting therapeutic decisions.
This is highly relevant as it significantly improves diagnostics and the resulting therapeutic decisions.