BMs are a common occurrence in advanced stages of primary solid cancers and often have a poor prognosis. The incidence of BMs has risen, possibly due to improved diagnostics and therapy methods that extend patient survival but also provide more opportunities for cancer cells to metastasize to the brain. Despite recent progress in BM treatment and imaging, the outlook for patients remains poor. Accurate and early diagnosis of the BM origin is crucial for tailoring adequate therapy to improve patients’ prognosis. Particularly for those with BMs that are poorly accessible to biopsy, diagnosis to differentiate primary tumor from BM and determine the origin can be arduous (21, 22). Therefore, it is crucial to identify new molecular biomarkers for precise diagnosis of BMs using less invasive liquid biopsies.
The goal of our study was to uncover specific miRNA patterns in tissue, CSF or blood plasma that can distinguish BMs in patients, compare these biological materials with respect to their suitability for use in the early diagnosis and find a diagnostic approach to accurately diagnose BMs with less invasive procedures than tissue biopsy. By doing so, treatment plans could be made in advance, improving patients' quality of life. Based on the fact that tissue miRNAs were previously found to be useful in classifying BMs, we hypothesized that CSF, as a biological fluid unique to the central nervous system (CNS), would be an even better option for miRNA detection since miRNAs have been proven to be stable in this body fluid (10). Additionally, CSF washes only the CNS, and, unlike blood plasma, is in direct contact with the tumor microenvironment. It should, therefore, contain fewer non-specific miRNAs compared to blood plasma or serum, making it less vulnerable to contamination and more tissue specific (8).
Based on our results, tissue miRNAs in patients with BMs could discriminate the individual analyzed BMs with a sensitivity of 50.0–100.0% and specificity of 71.4–100.0% (Table 1). The lowest sensitivity was observed in BML, which corresponds with high heterogeneity of histological types (23). These findings and identified miRNA profiles are consistent with the results of Roskova et al. (10). However, biopsy and subsequent histopathological examination is not always indicated due to the fragility of cancer patients and the high invasiveness of the surgical procedure or the BM localization. For these reasons, using small RNA sequencing, we analyzed and compared 2 types of liquid biopsies from BM patients, which are less invasive than needle or conventional biopsy. MiRNAs possess favorable biochemical properties that make them ideal candidates as they are easily accessible indicators from a technical perspective. These small transcripts exhibit high stability and have a prolonged half-life in biological samples, thereby eliminating the need for specialized handling. Moreover, miRNA analysis can be applied to readily available samples and quantified using standard techniques that are already employed in clinical laboratories, such as quantitative PCR, at a relatively low cost with high sensitivity and specificity. Nevertheless, the identification of new circulating miRNA biomarkers is challenging due to multiple factors.
The profiling of circulating miRNAs in peripheral blood has attracted significant attention as potential diagnostic and prognostic biomarkers for various diseases. However, it is important to note that the presence of non-specific miRNAs in peripheral blood can lead to a bias in data interpretation and other undesirable consequences. The two most prevalent miRNAs in human peripheral blood are miR-16-5p and miR-486-5p, both of which belong to the erythroid-specific group. Comparison of the potential use of serum instead of plasma for translational studies was performed by Dufourd et al., however, non-specific miRNAs were present at similar levels in both materials, and, in addition, plasma performed better with respect to overall sequencing data yield (24). Regarding the purity of the different types of biopsies analyzed in our present study with respect to contaminating miRNAs, the plasma samples were highly contaminated compared to CSF and tissue (Fi g. 1). Their high abundance in small RNA sequencing libraries can have negative impact, such as reducing the diversity and complexity of the sequencing libraries, leading to difficulty in detecting and accurately measuring less abundant miRNAs (25, 26). Furthermore, we observed more similar expression profiles among highly expressed miRNAs in tissue and miRNA levels in CSF samples compared to plasma samples. This result supports our hypothesis of the possible use of patients' CSF for the diagnosis of BMs. Overall, based on these data, we can consider CSF as a biological material less contaminated by non-specific miRNAs. This could be presumably caused by the fact that CSF surrounds only CNS, whereas peripheral blood circulates throughout the body and thus presumably contains more non-specific molecules. On the other hand, we observed overall higher levels of possible contaminants in plasma and CSF samples compared to tissue samples and neither miRNA level profiles in CSF nor plasma did not show significant specificity in samples from patients with brain metastases.
In CSF of BM patients, we found miRNAs with different levels characterized by a low diagnostic potential (Fig. 3a). In plasma, identified miRNAs showed no potential to serve as diagnostic biomarkers (Fig. 4a). In previous study, we demonstrated that miRNA profiles in CSF can distinguish between different brain tumors (8), however, we could not confirm this finding within the BMs. Nevertheless, we were able to determine the specificity and sensitivity with which the identified miRNAs can classify BM patients into diagnostic groups, whereas this could not be determined for plasma indicating further complications for potential use of this body fluid in clinical practice. To conclude, our data show that although we are able to discriminate individual BM types with limited specificity and sensitivity within a larger number of miRNAs based on their CSF level profile, this would probably not be possible within a potential diagnostic panel with a limited number of miRNAs. The use of plasma based on our data is not an option at all, as miRNA profiles showed no diagnostic potential even when using a large number of molecules. Nonetheless, these findings would need further confirmation or refutation in a larger cohort of patients.
We also investigated how the miRNA levels in plasma of BM patients originating in colorectal cancer may be influenced by the primary tumor or whether they are specifically indicative of the BM. It is evident from Fig. 5 that much more similarity was observed between the miRNA profiles corresponding to colorectal cancer and the BMs derived from this cancer compared to healthy controls. These observations would preclude the use of small RNA sequencing of plasma from patients with suspected BM as a diagnostic tool since it could not distinguish between primary tumor and the metastasis, nevertheless, these data may be limited to a certain extent by a possible batch effect between sequencing runs. Although it was not possible to make the same comparison for CSF in this study, as we did not have available samples adequate for this analysis, we suggest this to be explored in future research.
Recent advances in next-generation sequencing (NGS) technology have revolutionized the field of miRNA research, allowing for the identification of thousands of miRNAs simultaneously. NGS can provide high-throughput and comprehensive profiling of miRNAs in biological samples, including plasma and CSF. Moreover, the use of bioinformatics tools for data analysis and interpretation can aid in the identification of miRNA biomarkers with high sensitivity and specificity for BMs. Furthermore, the identification of circulating miRNAs with prognostic and predictive value could improve personalized treatment strategies for BM patients. However, as far as the miRNA profiling workflow is concerned, there is no standardized protocol so far, which complicates the achievement of final outputs and causes variability in the findings. This is particularly a problem for circulating miRNAs from liquid biopsies, which are present generally in low levels in body fluids. Low RNA input can lead to an increased proportion of adapter dimer and non-miRNA reads, while decreasing the number of reads that are mapped to miRNAs. This outcome requires a higher raw sequencing depth to compensate for the loss of miRNA reads. Moreover, a greater amount of contaminating RNA molecules from exogenous RNAs is frequently observed in low input samples, resulting in the detection of numerous non-target reads. In a recent study, Wong et al. demonstrated significant differences in the detection of individual miRNAs and their representation in sequencing data depending on the kit used for RNA extraction and subsequent preparation of cDNA libraries. Based on their data, it can be concluded that a consensus of the scientific community and the establishment of a standard protocol for miRNA profiling and subsequent bioinformatic evaluation of the obtained data is needed before routine use of miRNAs as biomarkers in clinical practice (27). Factors that affect their secretion into body fluids may also be problematic for the use of miRNAs. These include dynamic changes in their levels within a day or disease phase (28), but also factors related to the patients themselves such as gender, age, diet, smoking/non-smoking or physical activity (29–31). In the context of research aimed at analyzing and profiling miRNAs, patients should then correspond not only by diagnosis but also by the aforementioned factors, which is not always easy to do, especially in less common diseases such as BMs, which are associated with a very limited number of rare samples overall. The advantage is the relatively high level of miRNA stability even with long-term freezing, evidenced by the study of Balzano et al., who observed that miRNAs were detectable in the sample even within 14 years. However, their levels decreased with time, so it is advisable to use only fresh samples frozen for a maximum of 1 year for biomarker studies, which is often not achievable in practice and, again, may cause bias within the available data (32).
Since high-throughput sequencing technologies have become ubiquitous, the predicted number of miRNAs has skyrocketed. However, the validation of the results from exploratory phase of biomarker studies using qPCR poses many pitfalls as far as miRNAs are concerned. The lack of a validated intrinsic miRNA control in CSF/plasma remains a point of contention in the analysis of circulating miRNAs. Although 5S, U6, and other snoRNAs are commonly used as intrinsic controls for miRNA analysis in tissues or cells, their instability in serum/plasma from nuclear or cytosolic compartments renders them unsuitable for this purpose (33, 34). This absence of a validated intrinsic control has impeded research into the use of circulating miRNAs as biomarkers. Nevertheless, certain endogenous miRNAs, such as miR-15b, miR-16, and miR-24, have been proposed as suitable intrinsic controls for circulating miRNA analysis (35). Finally, miRNA isoforms (isomiRs) must also be considered, as they may not always be distinguishable from each other by conventional methods and may cause inconsistencies both in the evaluation of sequencing data and especially in the validation by qPCR (36). Various qPCR-based miRNA profiling platforms are being developed to address some of the associated bottlenecks (37), but to date there is no standardized procedure for validating circulating miRNAs in CSF or plasma.
Our findings indicate that the diagnostic value of miRNAs in liquid biopsies is low and remains to be further studied. Although aberrant levels of some circulating miRNAs are clearly associated with different BMs, given the relatively small patient cohort size and the limited number of patient groups included in our study, we suggest that further research is necessary to validate or refute these results. In particular, the study should be extended to a larger cohort with a higher number of patients and, possibly, more diverse groups of BMs. Such an investigation would provide a more comprehensive assessment of the diagnostic potential of circulating miRNAs in CSF and plasma for BMs and may lead to the identification of novel biomarkers that could improve diagnosis and, ultimately, patient outcomes. Although the use of liquid biopsies, especially CSF, would provide many advantages for the diagnosis of patients not only with BMs, such as lower invasiveness, the possibility of serial sampling and monitoring the dynamics of the disease course or earlier and faster diagnosis, it is necessary to standardize the methodology and solve some technical pitfalls associated with the analysis of circulating miRNAs. The potential of combining small RNA analysis of liquid biopsies with other diagnostic tools, such as imaging and clinical assessments, should also be investigated to improve the accuracy of BM diagnosis.