Biomarkers of Cardioembolic Stroke in Thrombus Composition: Diagnostic Value of DNA and Fibrin Content

ABSTRACT

Background and Purpose

Cryptogenic strokes, accounting for 25%–40% of ischemic strokes, represent a major challenge in secondary prevention due to their uncertain etiology and high recurrence risk. Identifying biomarkers to reliably distinguish cardioembolic (CE) strokes among embolic strokes of undetermined source (ESUS) could help guide therapeutic decisions. Previous studies have indicated thrombus DNA content as a potential biomarker of CE stroke etiology, but direct quantification of fibrin, another key component, has not been adequately explored.

Methods

We analyzed thrombi collected from 186 ischemic stroke patients undergoing endovascular treatment between 2019 and 2023. Thrombi were processed using a quantitative method based on ex vivo tPA‐mediated fibrinolysis followed by mechanical homogenization. Stroke etiology was classified according to TOAST criteria: 40% cardioembolic, 24% non‐cardioembolic (large artery atherosclerosis or dissection), and 36% ESUS. Biomarker content was correlated with stroke etiology, and the diagnostic performance of DNA and fibrin (D‐dimer) content was evaluated.

Results

Cardioembolic thrombi contained significantly higher levels of DNA (median [IQR]: 325.3 [177–484] ng/mg) and D‐dimer (17.5 [9.1–23.8] μg/mg) compared to non‐cardioembolic thrombi (DNA: 128 [76.4–263] ng/mg; D‐dimer: 11.4 [6.8–13.2] μg/mg), with no significant differences observed in heme or GPVI content. The combined use of thrombus DNA and fibrin (D‐dimer) content provided good discrimination between CE and non‐CE thrombi, with an area under the ROC curve of 0.79 (95% CI, 0.70–0.87).

Conclusion

DNA and fibrin content in thrombi are promising biomarkers for identifying cardioembolic stroke etiology. Prospective studies should evaluate their use in selecting ESUS patients who may benefit from anticoagulant therapy.

1. Introduction

Cryptogenic strokes, accounting for 25%–40% of ischemic strokes (IS), present significant challenges for secondary prevention due to the uncertainty surrounding their etiology and the associated risk of recurrence [1, 2]. Identifying reliable biomarkers that reflect the underlying stroke mechanism is therefore a key objective, particularly for the selection of patients who might benefit from anticoagulation.

Previous work has shown that thrombus DNA content correlates with cardioembolic (CE) etiology when using a mechanical grinding method to quantify thrombus components [3]. This approach enables direct analysis of thrombus homogenates, offering improved accuracy and reproducibility over semi‐quantitative histological techniques [4]. However, it does not allow for quantification of fibrillar components such as fibrin.

To address this limitation, we produced thrombus lysates by ex vivo digestion with a tPA/plasminogen fibrinolytic cocktail, before mechanical homogenization. This study aimed to assess the performance of these biomarkers—particularly DNA content and ex vivo fibrinolysis‐generated D‐dimers—for discriminating CE from non‐CE stroke etiologies, and to explore their potential value in stratifying cryptogenic stroke patients.

2. Methods

2.1. Standard Protocol Approvals, Registrations, and Patient Consents

Thrombi and clinical data were collected in the same manner as in our previous study [3], following the exact same protocol for endovascular procedures, data acquisition (Endovascular Treatment in Ischemic Stroke registry, URL: https://www.clinicaltrials.gov; Unique identifier: NCT03776877), and stroke classification. All patients or their representatives were informed and could refuse participation. The study was approved by the local ethics committee (CPP Nord Ouest II, ID‐RCB number: 2017‐A01039‐44).

2.2. Preparation of Thrombus Lysates

Frozen thrombi from the Compoclot Bio‐bank were rehydrated with phosphate buffered saline (PBS) 1X (30 min, at 37°C). The rehydration buffer was used as the lysis solution, to which plasminogen (25 μg/mL, ref. TC41005, Technoclone) and tissue plasminogen activator (tPA, 1 μg/mL, Boehringer Ingelheim) were added (40 μL/mg of thrombus). Thrombi were incubated in this fibrinolytic solution for 3 h at 37°C. After the lysis phase, samples were subjected to sonication to ensure complete homogenization (20 s, Amplitude 30, Bioblock scientific Vibracell). Homogenates were then stored at −80°C.

2.3. Quantification of Red Blood Cells, Fibrin, DNA and Platelets

Red blood cells (RBC), DNA and platelet content were estimated as described previously [3]. Briefly, red blood cell content was estimated by measuring heme concentration in thrombus homogenates using a formic acid‐based colorimetric assay. DNA content was quantified using the Quant‐iT PicoGreen dsDNA assay (Life Technologies). Soluble GPVI levels, used as a marker of platelet content, were measured by immunoassay on a MesoScale Discovery platform. Plates were coated with anti‐GPVI antibodies, incubated with thrombus homogenates and a biotinylated detection antibody, followed by streptavidin‐Sulfo‐TAG labeling and signal quantification using a Quickplex reader. D‐dimer, generated by plasmin during lysis, was quantified after dilution of thrombus lysates in PBS 1X by an automated immunoturbidimetric method (INNOVANCE D‐Dimer, Siemens), adapted on CN‐6000 analyzer (Sysmex), and was thought to reflect the fibrin content of thrombi.

2.4. Statistical Analysis

Full statistical methodology is detailed in the Supporting Information Methods. In summary, thrombus biomarker levels were compared across the three stroke etiology groups using one‐way ANOVA with Bonferroni‐corrected post hoc comparisons. In our previously published cohort based on mechanical homogenization [3], D‐dimer levels were measured but not reported. For the present study, we retrieved these unpublished D‐dimer data to enable direct comparison with the current cohort. Similarly, previously reported values for DNA, heme, and GPVI were also used to compare the performance of each biomarker across the two processing methods. The ability of DNA and D‐dimer content to differentiate CE from non‐CE strokes was evaluated using ROC curve analysis, with thresholds derived from the Youden index and fixed sensitivity/specificity criteria. A multivariable logistic regression model combining DNA and D‐dimer was used to assess their joint diagnostic performance. Results were compared to a prior dataset using mechanical lysis, and standardized differences were calculated to assess baseline comparability. To assess the potential influence of prior IVT on D‐dimer performance, we conducted subgroup analyses stratified by IVT status.

3. Results

From February 2019 to December 2023, 186 thrombi from 186 patients with consecutive acute IS with LVO treated by EVT were collected and analyzed by optimized enzymatic lysis to quantify RBC (via Heme), fibrin (via D‐dimer), DNA, and platelets (via GPVI). Patient and treatment characteristics of the whole study sample are reported in Table S1 and are comparable to the characteristics of patients included in the previously reported analysis of thrombi after mechanical grinding [3]. Stroke causes were classified in 40% of cases as CE origin (vs. 57% for previously reported thrombi collection), 24% as non‐CE origin (vs. 13% for previously reported thrombi collection), and in 36% as ESUS (vs. 30% for previously reported thrombi collection). Patient and treatment characteristics according to stroke etiology are reported in Table 1.

TABLE 1.

Patients and treatment characteristics of thrombi collection analyzed by optimized enzymatic lysis according to suspected stroke etiology.

Characteristics	Suspected stroke etiology
	Cardioembolic	Non‐cardioembolic	ESUS
Number of patients (thrombi)	74	45	66
Demographics
Age, years, mean (SD)	74.1 (10.4)	66.0 (9.9)	61.9 (17.2)
Men	42/74 (56.8)	34/45 (75.6)	31/65 (47.7)
Medical history
Hypertension	43/74 (58.1)	25/45 (55.6)	31/66 (47.0)
Diabetes	21/73 (28.8)	6/45 (13.3)	11/65 (16.9)
Hypercholesterolemia	22/73 (30.1)	15/44 (34.1)	17/66 (25.8)
Current smoking	11/72 (15.3)	17/44 (38.6)	15/64 (23.4)
Coronary artery disease	0/72 (0.0)	2/45 (4.4)	2/66 (3.0)
Previous stroke or TIA	10/72 (13.9)	5/45 (11.1)	6/66 (9.1)
Previous antithrombotic medications	36/73 (49.3)	10/41 (24.4)	16/64 (25.0)
Current stroke event
NIHSS score, median (IQR)	15 (10–19) [72]	15 (10–19) [44]	17 (14–20) [65]
Pre‐stroke mRS ≥ 1	26/74 (35.1)	8/44 (18.2)	16/65 (24.6)
Site of occlusion
M1‐MCA	41/73 (56.2)	13/43 (30.2)	31/64 (48.4)
M2‐MCA	8/73 (11.0)	6/43 (14.0)	8/64 (12.5)
Intracranial ICA or tandem	15/73 (20.5)	5/43 (11.6)	13/64 (20.3)
Tandem	1/73 (1.4)	10/43 (23.3)	5/64 (7.8)
Extracranial ICA	1/73 (1.4)	5/43 (11.6)	0/64 (0.0)
Vertebro‐basilar	3/73 (4.1)	2/43 (4.7)	3/64 (4.7)
Others	4/73 (5.5)	2/43 (4.7)	4/64 (6.3)
Treatment characteristics
Intravenous rt‐PA	33/74 (44.6)	19/45 (42.2)	32/66 (48.5)
Onset to groin puncture time, min	289 (198–350) [71]	282 (232–368) [42]	251 (173–357) [63]
Thrombi contents
Heme, ng/mg, mean (SD)	220.1 (97.8)	246.1 (106.6)	195.6 (127.2)
GPVI, ng/mg, mean (SD)	17.8 (11.5)	20.5 (15.3)	19.0 (14.1)
DNA, ng/mg	325 (177–484)	128 (76.4–263)	194 (97–287)
D‐Dimers, g/mg	17.5 (9.1–23.8)	11.4 (6.8–13.2)	10.2 (6.0–18.5)

Note: Values expressed as no/total no. (%) or median (IQR) unless otherwise indicated. In case of missing data for quantitative characteristics, number of available cases are reported in brackets.

Abbreviations: GPVI, glycoprotein VI; ICA, internal carotid artery; IQR, interquartile range; MCA, middle cerebral artery; mRS, modified Rankin scale; NIHSS, National Institutes of Health Stroke Scale; rt‐PA, recombinant tissue plasminogen activator; SD, standard deviation; TIA, transient ischemic attack.

3.1. Thrombus Cellular Content and Stroke Etiology

As showed in the Figure 1, thrombus heme and GPVI contents did not differ significantly between the three stroke etiologies. A significant difference between stroke etiologies was found for both DNA and D‐Dimer contents. CE thrombi contained significantly more DNA (median [IQR]: 325 [177–484] ng/mg) than non‐CE thrombi (128 [76–263] ng/mg), and ESUS thrombi (194 [97–287] ng/mg, global p < 0.001). Similarly, thrombus D‐dimer content was also increased in CE thrombi (median [IQR]: 17.5 [9.1–23.8] μg/mg) by comparison to non‐CE thrombi (11.4 [6.7–13.8] μg/mg) and ESUS thrombi (10.2 [6.0–18.6] μg/mg, global p < 0.001).

FIGURE 1.

Distribution (Tukey's box plot) of histological features of thrombi according to suspected stroke etiology. Boxes show the 25th, 50th, and 75th, and whiskers indicates values outside the lower and upper quartile with a length equal to 1.5 interquartile range; diamond indicates the mean values. p‐values for global comparison (one‐way ANOVA) are reported after a log‐transformation for DNA and D‐dimer values; * indicated p‐values < 0.05 for post hoc pairwise comparison between cardioembolic stroke and each other stroke subgroups (after adjustment for multiple comparison using Bonferroni correction): For DNA outcome p‐values < 0.001 for all comparisons, for Ddimeres outcomes p‐value < 0.001 for non‐cardioembolic vs. Cardioembolic and p‐value = 0.004 for ESUS versus Cardioembolic.

3.2. Ability of Thrombi Content Features to Discriminate CE From Non‐CE

As shown in Table S2, there was a fair discrimination ability of DNA and D‐dimer thrombi content to differentiate thrombi of CE and non‐CE origins, with an AUC of 0.76 (95% CI, 0.66 to 0.85) and 0.71 (95% CI, 0.62 to 0.81), respectively. When we investigated the combination of DNA and D‐Dimer in a logistic regression model, the AUC was slightly increased (AUC = 0.79; 95% CI, 0.70 to 0.87). The specificity and sensitivity of DNA and D‐dimer thrombi contents for discriminating CE thrombi from non‐CE thrombi for various thresholds are reported in Table S2.

As shown in Figure 2, the discriminatory ability of thrombus DNA and D‐dimer content (alone or in combination) to differentiate thrombi of CE from non‐CE origins assessed by optimized enzymatic lysis used in the present cohort did not differ significantly from the discrimination ability of thrombus DNA and D‐dimer content assessed after mechanical grinding of thrombi used in the previous cohort published [3].

FIGURE 2.

Receiver operating characteristic (ROC) curves for the discrimination of CE versus non‐CE strokes using significant thrombus biomarkers (individual or combined). Results are shown for the two analytical approaches: Mechanical grinding (first cohort of 250 thrombi [3]) and optimized enzymatic lysis (current cohort). Note: The DNA‐related ROC curves from the mechanical grinding cohort were previously published, while the D‐dimer data from that same cohort had been collected but were not reported until now. The two datasets come from independent cohorts, and comparisons are presented to illustrate the consistency of biomarker performance across processing methods, not as direct within‐sample comparisons.

3.3. Subgroup Analysis Stratified by Intravenous tPA Treatment Before EVT

In the current cohort using optimized ex vivo fibrinolysis, D‐dimer levels demonstrated similar discriminatory ability between CE and non‐CE strokes regardless of IVT status (AUC = 0.73 with IVT vs. 0.70 without IVT, p = 0.73; p‐heterogeneity by ANOVA = 0.77; Figure S1, panel A). Conversely, in the previous cohort processed by mechanical grinding, D‐dimer performance was significantly lower in patients who had received IVT (AUC = 0.60 with IVT vs. 0.76 without IVT, p = 0.006; p‐heterogeneity by ANOVA = 0.017; Figure S1, panel B).

4. Discussion

This study validates thrombus DNA content as a robust biomarker of CE stroke etiology, confirming previous findings [3]. Using an optimized enzymatic lysis technique, we demonstrated that fibrin content, measured via D‐dimer, is also significantly higher in CE thrombi, making it an additional promising co‐biomarker for distinguishing stroke etiology.

Multiple studies have explored the relationship between thrombus histological composition and stroke etiology. Several large and recent cohorts have suggested that CE thrombi are typically richer in fibrin and platelet aggregates and poorer in RBCs [5, 6, 7]. However, findings remain inconsistent, with some studies reporting opposite trends [8], and others finding no significant association [9, 10]. Our results are broadly consistent with the recent meta‐analysis by Sujijantarat et al. [11], which found increased fibrin content in CE thrombi. Similarly, we observed significantly higher fibrin levels in CE compared to non‐CE thrombi, reinforcing the concept that cardiac thrombi are more fibrin rich. However, in contrast to the meta‐analysis—which reported lower RBC content in CE thrombi—we found no significant difference in heme concentration across etiologies. This discrepancy likely stems from methodological differences. Previous studies mostly used semi‐quantitative histology, which estimates component proportions by surface area, meaning an increase in one component (e.g., fibrin) necessarily implies a decrease in others (e.g., RBC). In contrast, our approach directly quantified each biomarker relative to thrombus weight, allowing independent measurement and potentially explaining these differing results. Moreover, antigen‐based quantification of these markers may be affected by proteolytic degradation during ex vivo fibrinolysis, potentially altering epitope recognition and leading to underestimated concentrations. These factors might explain why platelet content (GPVI) did not correlate with stroke etiology in our current cohort, unlike our earlier findings [3]. The existing literature is inconsistent, with most studies reporting no significant association between platelet content and stroke etiology [12, 13]. Whether platelet markers consistently vary by thrombus origin warrants further investigation. DNA quantification, performed by a non‐antigenic method capable of detecting both intact and fragmented DNA, likely provides more robust results. We found significantly higher DNA content in cardioembolic thrombi. DNA in thrombi primarily originates from neutrophils, which contribute to thrombosis through the formation of neutrophil extracellular traps (NETs). This observation is consistent with previous studies reporting increased NET expression in cardioembolic thrombi [14, 15]. However, DNA is not specific to neutrophils and may also derive from other white blood cells. Prior studies investigating total WBC content in cardioembolic thrombi have yielded conflicting results, with some reporting higher WBC levels [5, 6, 7] and others finding lower concentrations [16]. Finally, omics technologies are rapidly advancing our understanding of thrombus biology. In 2023, López‐Pedrera et al. [17] demonstrated that specific protein expression clusters within the thrombus proteome correlate with stroke severity, etiology, and clinical outcomes.

In our first cohort [3], using thrombus DNA content with a 90% specificity threshold, approximately half of the cases were classified as cardioembolic, yielding a high positive predictive value to guide anticoagulation strategies. In the current cohort, applying the same threshold, DNA alone identified 12% of ESUS cases as having a cardioembolic profile, while the combination of DNA and D‐dimer increased this proportion to 36%. These findings suggest that analyzing thrombus biomarkers, either individually or in combination, may help identify a clinically meaningful subgroup of ESUS patients with a likely cardioembolic mechanism, at a high specificity threshold. Given that ESUS accounts for nearly 50% of thrombectomy cases, this approach could significantly impact secondary prevention strategies. To date, several large interventional trials have attempted to stratify ESUS patients for anticoagulation based on indirect markers of atrial cardiopathy. However, trials such as ARCADIA [18] and a recent meta‐analysis [19] failed to demonstrate the superiority of anticoagulation in these subgroups. By relying on direct thrombus analysis rather than surrogate markers, our study offers a novel strategy to more precisely identify ESUS patients who may benefit from anticoagulation. If validated in multicenter retrospective cohorts, this approach could lay the groundwork for a prospective interventional trial targeting biomarker‐defined ESUS subgroups.

From a pathophysiological perspective, our findings highlight the key role of neutrophils in thrombus formation in cardiac cavities. DNA in thrombi largely originates from neutrophils [15], which promote thrombosis through NETosis. Neutrophil extracellular traps (NETs), released during inflammation or vascular injury [20], enhance coagulation by interacting with platelets and fibrin—revisiting the Virchow triad [21]. While this mechanism is well established in deep vein thrombosis [22], recent studies suggest it may also contribute to thrombus formation in the left atrium [23]. Moreover, flow is recognized as a key modulator of fibrin formation [24]. The differences in fibrin content observed in our cohort may reflect the distinct flow conditions under which CE and non‐CE thrombi form. Our findings, together with previous evidence linking NET formation to elevated stroke risk markers in atrial fibrillation [25, 26], underscore the pivotal role of neutrophils and NETs in thrombus formation within the cardiac cavities.

Several methodological limitations should be acknowledged. Although DNA serves as a biomarker, it does not differentiate between neutrophils, endothelial cells, or other immune cell types, which could contribute to variability. While the enzymatic lysis technique offers improved accuracy, its reproducibility across different laboratories and settings needs to be further validated. Additionally, while D‐dimer quantification serves as a proxy for fibrin content, it also reflects thrombus susceptibility to fibrinolysis. In our current protocol, ex vivo exposure to saturating concentrations of tPA and plasminogen standardizes D‐dimer release, partially mitigating variability due to thrombolytic resistance. This is supported by our subgroup analysis: in the current cohort, D‐dimer performance was unaffected by prior IVT, whereas in the previous cohort, where D‐dimers reflected in vivo fibrinolysis, discrimination between CE and non‐CE thrombi was reduced in IVT‐treated patients. This suggests that unstandardized in vivo fibrinolysis introduces heterogeneity that can confound biomarker interpretation. In addition, we and others have shown that thrombi may present structural barriers [27] of compact fibrin and platelets which limit enzyme penetration and impair lysis. Since thrombi were not sectioned before lysis, residual fibrin may persist in some cases. Thus, D‐dimer levels likely reflect both fibrin burden and lytic accessibility. Future studies incorporating systematic analysis of thrombus remnants, pre‐sectioning, and expanded fibrinolytic cocktails may enhance the robustness of D‐dimer as a quantitative marker.

Despite these limitations, the study confirms the value of thrombus DNA and fibrin content as biomarkers for CE strokes. This novel approach not only enhances our understanding of thrombus pathophysiology but also provides a foundation for improved patient stratification and treatment personalization in ESUS. Future multicenter studies and advanced omics technologies may further refine and expand the clinical utility of thrombus analysis.

Author Contributions

Mialitiana Solo Nomenjanahary: investigation, methodology, validation. Jéromine Fasille: conceptualization, investigation, writing – original draft. Dorothée Faille: investigation, methodology, validation. Capucine Habay: investigation. Bertrand Lapergue: investigation. Arturo Consoli: investigation. Laurine Bedoucha: investigation. Mikael Mazighi: supervision. Benoit Ho‐Tin‐Noé: resources, supervision, writing – review and editing. Michel Piotin: investigation. Raphael Blanc: investigation. Lucas Di Meglio: conceptualization, writing – review and editing. Julien Labreuche: formal analysis. Maeva Kyheng: formal analysis. Jean Philippe Desilles: supervision, project administration, writing – review and editing.

Funding

The authors have nothing to report.

Disclosure

Guarantor Statement: Lucas Di Meglio takes full responsibility for the article, including the accuracy and appropriateness of the reference list.

Ethics Statement

The study was approved by the local ethics committee (CPP Nord Ouest II, ID‐RCB number: 2017‐A01039‐44).

Consent

All patients or their representatives were informed and could refuse participation. Data acquisition followed the protocol: Endovascular Treatment in Ischemic Stroke registry, URL: https://www.clinicaltrials.gov; Unique identifier: NCT03776877.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: Diagnostic performance of thrombus D‐dimer content according to prior intravenous thrombolysis (IVT). (A) In the current cohort using optimized chemical lysis, the diagnostic accuracy of D‐dimer levels for differentiating cardioembolic from non‐cardioembolic stroke was similar in patients with and without prior IVT (AUC 0.73 vs. 0.70, p = 0.73; P‐heterogeneity by ANOVA = 0.77). (B) In the previous cohort using mechanical grinding, D‐dimer performance was significantly lower in patients who had received IVT (AUC 0.60 vs. 0.76, p = 0.006; P‐heterogeneity by ANOVA = 0.017). Note: While this cohort was previously described in Di Meglio et al., Stroke 2020, the D‐dimer data had not been published before and are presented here for the first time. These results were obtained from independent thrombus samples and processing protocols and should be interpreted as an indirect comparison.

Table S1: Patient and thrombus content characteristics according to thrombi collection, analyzed by optimized chemical lysis (February 2019 to Decembre 2023) and those analyzed by mechanical grinding (June 2016 to November 2018) [1].

Table S2: Accuracy of significant histological features of thrombi for differentiating cardioembolic and ESUS strokes.

Fasille J., Nomenjanahary M. S., Faille D., et al., “Biomarkers of Cardioembolic Stroke in Thrombus Composition: Diagnostic Value of DNA and Fibrin Content,” European Journal of Neurology 33, no. 2 (2026): e70484, 10.1111/ene.70484.

Data Availability Statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.