Molecular Profiling of Malignant Pleural Effusions with Next Generation Sequencing (NGS): Evidence that Supports Its Role in Cancer Management

Malignant pleural effusions (MPEs) often develop in advanced cancer patients and confer significant morbidity and mortality. In this review, we evaluated whether molecular profiling of MPEs with next generation sequencing (NGS) could have a role in cancer management, focusing on lung cancer. We reviewed and compared the diagnostic performance of pleural fluid liquid biopsy with other types of samples. When applied in MPEs, NGS may have comparable performance with corresponding tissue biopsies, yield higher DNA amount, and detect more genetic aberrations than blood-derived liquid biopsies. NGS in MPEs may also be preferable to plasma liquid biopsy in advanced cancer patients with a MPE and a paucicellular or difficult to obtain tissue/fine-needle aspiration biopsy. Of interest, post-centrifuge supernatant NGS may exhibit superior results compared to cell pellet, cell block or other materials. NGS in MPEs can also guide clinicians in tailoring established therapies and identifying therapy resistance. Evidence is still premature regarding the role of NGS in MPEs from patients with cancers other than lung. We concluded that MPE processing could provide useful prognostic and theranostic information, besides its diagnostic role.


Introduction
Pleural effusion (PE), the pathologic accumulation of excess fluid inside the pleural cavity, is a complication that often appears in a wide spectrum of clinical settings. Prevalent underlying causes include non-neoplastic conditions (e.g., congestive heart failure and infections) and cancer, either metastatic or primary [1,2]. Lung and breast cancers make up the most common malignancies that spread into the pleural surfaces forming malignant pleural effusions (MPEs) [3,4]; the latter appear in around 16% of patients with M1b Stage IV non-small cell lung cancer (NSCLC) [5]. In contrast, mesothelioma, a primary pleural malignancy linked to asbestos exposure, is a relatively rare cause of MPE compared to metastases [3,4,6]. MPEs confer significant morbidity, mortality, and poor survival rates, while their management is primarily palliative, intending to ease symptoms and improve quality of life [3,7].
Laboratory examination of MPEs provides robust diagnostic, prognostic, and theranostic information [8]. As soon as a PE specimen arrives to the pathology laboratory for processing, it undergoes centrifugation, which divides the fluid into a cell pellet (cellular-rich material) at the

Pleural Fluid Liquid Biopsy Compared to Other Types of Samples
Among all sources of genetic material, tissue biopsy is considered the current standard for the molecular characterization of tumors and pre-analytical factors (such as the DNA/RNA concentration and quality) can heavily influence the results of the molecular analysis [25]. Of interest, Zhang et al. found that the DNA extracted from pleural fluid FFPE cell blocks had similar quality to its tissue counterpart, while freshly centrifuged pleural fluid preparations achieved even higher quality standards [26]. Yamamoto et al. also described comparable RNA concentrations between pleural fluid and tissue samples [27].
The aforementioned pre-analytical properties may account for the highly accurate results following MPE NGS analysis. Xiang et al. and Liu et al. reported concordance rates between pleural fluid and tissue NGS samples of 83.3% (50/60 mutations) and 86.7% (26/30 mutations), respectively [28,29]. Zhang et al. successfully detected EGFR mutations in 15/15 previously confirmed cases [30]. Song et al. identified EGFR mutations and ALK aberrations in the pleural fluid of 68/123 and 11/123 of the tested patients, respectively, while the EGFR wild-type was associated with a PD-L1 IHC score ≥ 50%. Of interest, pleural fluid and tissue samples had a concordance rate of 86.2% (25/29) for PD-L1 IHC expression with a 50% threshold; this directly translates to identifying patients that could benefit from pembrolizumab therapy [31,32]. In many studies, a large proportion of discordant mutations between pleural fluid and tissue were actually novel mutations not detected in the original tumor biopsy [26,28,29,33,34]. Like other types of liquid biopsy, these mutations may accurately reflect intratumoral (spatial and temporal genetic) heterogeneity, rather than being false positive results, thus providing a more complete picture of the tumor's mutational landscape [35]. In some cases, these mutations had direct implications for patient management, as they were potentially targetable or resistance-conferring to targeted molecular agents [26,33].
Liquid biopsy is predominantly performed via plasma analysis, while cerebrospinal fluid (CSF) and pleural fluid are gaining increasing popularity as alternative sources of genetic material. Plasma, specifically, offers the advantage of easy retrieval through a routine blood draw. On the other hand, pleural fluid collection requires thoracentesis, which is a procedure with more complications [36]. Tong et al. and Villatoro et al. reported that pleural fluid samples had higher cfDNA concentrations and mutation allele frequencies (MAFs) compared to plasma samples [24,37], while Liao et al. showed that pleural fluid NGS detected more unique mutations compared to other sample types, including plasma [34]. Zhang et al. found a concordance of 86.7% between the pleural fluid and plasma NGS among 15 patients harboring known EGFR mutations; two mutations were missed by plasma analysis while all mutations were successfully detected by pleural fluid analysis [30].
Overall, these data suggest that pleural fluid may be a more reliable source of genetic material than plasma, given its superior pre-analytic indices and mutation-detecting ability. Despite its slightly more invasive nature, pleural fluid liquid biopsy might be able to offer a more reliable alternative to the classic tissue biopsy, whilst avoiding many of the shortcomings of plasma liquid biopsy [35]. Pleural fluid also offers the additional advantage of cytologic examination. Fine needle aspiration (FNA) is another source of genetic material suitable for NGS that also provides this option. However, Zhang et al. showed that its DNA concentration and quality was lower compared to that of pleural fluid [26]. The main findings of the studies described in this section are summarized in Table 1.

Correlation of Pleural Fluid NGS with Cytomorphologic Findings and Tumor Cellularity
In a study of patients with various malignancies, Yang et al. noticed a correlation between the results of cytology and NGS. Specifically, at least one mutation was detected in every sample with confirmed (9/9) or suspicious (2/2) cytology for malignancy, while no mutations were found in benign effusions (0/4) [39]. The same authors found a significant correlation between tumor cellularity and the variant allele frequency [40]. Leichsenring et al. also noted many false negative results when tumor cellularity dropped below 10% [38]. These findings suggest that cytologic evaluation is crucial to guide molecular sequencing. Samples completely lacking malignant cells are less likely to detect mutations, while samples without sufficient tumor cellularity may be prone to false negative results due to inadequate allele frequencies. In these cases, physicians may opt for a sample of higher tumor cellularity instead, thus saving valuable resources from redundant molecular analyses while simultaneously obtaining a more accurate picture of the tumor's mutational profile. Interestingly, the MPE volume did not correlate with tumor cellularity and therefore low MPE volume should not deter NGS analysis, provided that tumor cellularity is sufficient [41].
Nonetheless, there are reports of adequate mutation detection even with unfavorable pre-analytic characteristics. Buttitta et al. showed that NGS was able to detect 70% (7/10) of tissue-confirmed EGFR mutations in matched pleural fluid samples of low tumor cellularity (<10%). In comparison, Sanger sequencing was only able to detect 20% (2/10) of mutations, showcasing the advantages of deep sequencing over conventional sequencing methods [42]. Liu et al. found that mutations were successfully detected in samples with even lower tumor cellularity (<5%) in 85.7% (6/7) of cases, thus making an argument for using pleural fluid specimens with a non-ideal pre-analytic profile [29]. However, false positive results may arise under those circumstances, as germline or clonal hematopoietic mutations may be mistaken for tumor mutations [39]. The same phenomenon has been described in plasma liquid biopsy [35]. As a solution, Yang et al. suggests that paired white blood cells should be simultaneously sequenced to exclude mutations of non-tumor origin [39]. The abovementioned information is summarized in Table 2. Table 2. Correlation of pleural fluid NGS with cytomorphologic findings and tumor cellularity.

First Author/Reference Pleural Fluid Material Summary of Findings
Yang et al. [39] Supernatants All malignant (9/9, five of which were from NSCLC) and suspicious (2/2, one of which was from NSCLC) pleural fluid samples revealed mutations.
No mutations were found in benign samples.
Yang et al. [40] Cell blocks A significant correlation between tumor cellularity and VAF was revealed.
In pleural fluid samples without malignant cells, only 20% (1/5) of mutations were detected by NGS and no mutations were detected by Sanger sequencing.
Carter et al. [41] Cell blocks No relationship was found between malignant pleural effusion volume and pleural fluid overall or tumor cellularity.

The Value of Supernatant-Derived cfDNA
Cell blocks are the most common source of genetic material when performing pleural fluid molecular analysis. Their preparation requires centrifugation of the sample, followed by retrieval and special processing of the sediment, while the supernatant is usually discarded [2,9,28]. However, recent evidence suggests that the cfDNA found in the supernatant can also be successfully used for molecular analysis with comparable, if not superior, results to cell blocks. Yang et al. showed that 100% (8/8) of the mutations found in FFPE cells blocks were also detectable in the matched supernatant of three NSCLC patients [39]. Xiang et al. found that pleural fluid supernatant NGS had higher concordance with tissue samples and yielded more known mutations compared to FFPE cell blocks [28]. Li et al. reported comparable mutational profiles and MAFs between supernatant and FFPE cell blocks for all genes included in the current diagnostic recommendations for NSCLC [43]. In addition, the MAFs from pleural fluid aspirate were even higher when compared to cell blocks in isolated cases [44]. Aside from high accuracy, supernatant analysis had a much shorter turnaround time compared to cell block preparation, which can last up to a week [28].
An alternative approach to cell block preparation is the direct extraction of genomic material from sedimentary tumor cells following centrifugation. In a study by Zhang et al., all known EGFR mutations (15/15) were successfully detected in both supernatant and sedimentary tumor cells, but the former achieved much higher MAFs [30]. Similarly, Tong et al. found that both sample types had comparable sensitivity (93% [27/29] with supernatant vs. 90% [26/29] with sediment-derived DNA) in detecting tissue-confirmed driver mutations. However, the supernatant displayed superiority in multiple other analytical indices, many of which had direct implications for treatment. These include achieving higher MAFs, tumor mutational burden, chromosomal instability, and sensitivity in locally metastatic patients, as well as higher detection of resistance-conferring (e.g., EGFR T790M) and unique mutations in advanced cancer patients, and more driver mutations in tissue-lacking patients. In addition, the supernatant retained its sensitivity in cases of cytologically-negative (86%) or hemorrhagic pleural fluid (72%), whereas the sensitivity of sedimentary tumor cells was significantly compromised (9% vs. 30% decrease in sensitivity, respectively) [37]. Overall, these findings suggest that supernatant analysis provides a more accurate depiction of the tumor mutational landscape and may be a more reliable source of genetic material when the sample tumor cellularity is either low or mixed by abundant non-neoplastic cellular elements.
Tumor exosomes are nucleic acid-containing microvesicles released into the tumor environment, thus providing a novel non-cellular source of genomic material for molecular analysis. Song et al. compared tumor exosome to supernatant cfDNA NGS analysis and found a concordance of 77.9% (243/312). Importantly, the concordance for ALK and EGFR mutations across 18 patients was 100%, showing that tumor exosomes derived from pleural fluid can be used to guide treatment with targeted agents. Another important observation was that the concordance of the two sample types increased to 94.1% (128/136) when copy number variations (CNVs) were excluded from the analysis [45]. Liao et al. has attributed the inherent weakness of cfDNA to detect CNVs to its fragmentation [34]. In addition, Tong et al. reported that supernatant-derived cfDNA revealed more CNVs (31%, 20/64) compared to sedimental tumor cell-derived DNA (14%, 9/64) [37]. Therefore, this shows that the sensitivity of supernatant in detecting CNVs via cfDNA is limited, thus alternative sources of genetic material could potentially be sought in this situation. The abovementioned information is summarized in Table 3.

Evaluation of Therapeutic Resistance, Response, and Management
cfDNA from (MPEs) can provide diagnostic information and reveal mutations in molecular pathways associated with therapy resistance. This information could direct the therapeutic management of NSCLC patients [38,46]. Common mutations associated with resistance to targeted therapy are EFGR T790M and ALK p.G1202R. Yang et al. and Villatoro et al. were able to isolate these mutations from MPEs of patients with disease progression after first-line treatment with tyrosine kinase inhibitors (TKIs). These results successfully guided therapeutic decisions and eventually led to a clinical benefit [24,39]. Zhang et al. discovered the EGFR T790M mutation in two NSCLC patients via MPE NGS, with one of them representing a novel mutation undetected in the matched tissue sample [26]. Goldberg et al. described a case of a patient progressing on third line TKIs, where pleural fluid NGS was utilized to detect a novel C797S mutation responsible for treatment failure [47]. Similarly, Li et al. reported a case for which MPE NGS was utilized to determine the cause of crizotinib non-response [48]. Other less common molecular biomarkers, including RET, HER2, and MET, can reveal additional therapeutic targets [12]. Wang et al. revealed MET amplifications in the pleural fluid of two patients who were subsequently treated with crizotinib based on this result [33]. Tong et al. found that there was no difference in the progression-free survival of 10 patients, when treatment decision with TKI was based on pleural fluid supernatant rather than tissue analysis [37].
Culturing cells from MPEs could be an alternative way for guiding therapeutic management. An example is choosing candidates for polyADP-ribose polymerase inhibitor (PARPi) therapy, which requires the presence of homologous recombination DNA repair (HRR)-defective cancer cells [49]. Patterson et al. successfully determined the HRR mutational status of patients using MPE-derived cell lines. Their sample included four NSCLC patients, three of whom had defective HRR, thus showing that selecting candidates for PARPi therapy using this alternative approach was feasible [50]. Using a similar approach, Roscilli et al. tested the in vitro response of MPE-derived cultured cell lines to different chemotherapy regimens and were able to successfully predict the patients' clinical response. This approach also allowed them to select regimens that had a synergistic effect, thus optimizing treatment selection. Interestingly, the same findings could not be reproduced in regard to TKIs, indicating that additional genetic or epigenetic factors may play a role in determining the response to these targeted molecular agents [51]. The main findings of these studies are summarized in Table 4. Table 4. NGS in MPE for the evaluation of therapeutic resistance, response, and management.

First Author/Reference Pleural Fluid Material Summary of Findings
Leichserning et al. [38] Cell blocks Clinically actionable mutations were detected that guided targeted therapy.
DiBardino et al. [46] Cell blocks, slides Mutations were detected in 4 out of the 5 patients whose pleural fluid samples were tested. These mutations changed management in two of these patients.

The Role of NGS in Malignant Pleural Effusions from Different Types of Cancers (Other than Lung)
There is a growing interest on the role of MPE analysis by NGS outside of lung cancer as well [38]. In two patients with breast cancer and colorectal cancer, Yang et al. found that their MPE supernatant NGS was concordant with their matched lymph node FNA cell block and surgical tissue NGS, respectively [39]. Shah et al. showed that MPE NGS in three metastatic ovarian cancer patients successfully revealed tissue-confirmed TP53 mutations. In two cases, TP53 MAFs were higher in pleural fluid compared to its matched tissues (FFPE and frozen), indicating higher tumor DNA fraction [53]. Similar to lung cancer, cases of novel mutations in MPE analysis have also been described with other malignancies. Using MPE-derived cultured cell lines from a patient with disseminated medulloblastoma, Xu et al. were able to detect a novel 17q deletion that was not previously detected in the tissue sample. After culturing cell lines from the tissue, the mutation was eventually confirmed, proving that it was also present in the tissue and that the FFPE tissue analysis results were false-negative [54]. This finding highlights the ability of pleural fluid to capture theumor heterogeneity more accurately and provides new insights for the applications of cell cultures in tumor molecular analysis. Finally, Zhou et al. showed that pleural fluid could potentially serve as an alternative to other liquid biopsy types, such as plasma and ascitic fluid, with comparable results in a case report of a gastric cancer patient [55]. A summary of these studies could be found in Table 5. Zhou et al. [55] Gastric Supernatants Concordant ATM (indel, frameshift, SNV, fusion), MET, and SMAD3 mutations were detected in pleural fluid, plasma, and ascites NGS from an advanced gastric cancer patient.

Discussion
MPEs from patients with advanced cancers are often processed in pathology laboratories for rendering diagnoses, assessing prognostic factors, and selecting patients for established precision therapies [8,10], Figure 1.

Discussion
MPEs from patients with advanced cancers are often processed in pathology laboratories for rendering diagnoses, assessing prognostic factors, and selecting patients for established precision therapies [8,10], Figure 1. As MPEs could be the first and only specimen received, especially when tissue biopsies are hypocellular or impossible to obtain, their efficient triage and processing is imperative to get maximum clinically relevant information [10,[20][21][22]. Besides the necessary morphologic evaluation and routine ancillary preparations, MPEs can provide robust theranostic information similar to tissue biopsies. Both cell pellets-including all preparations derived from it like direct smears, LBC slides, and FFPE cell blocks-and post-centrifuge supernatants can subsequently be processed for molecular testing [2,14,15]. This is critical for cases like advanced NSCLC, where testing for multiple biomarkers is recommended to tailor targeted oncologic therapies, according to the latest ASCO guidelines. NGS, a molecular technique that can detect multiple genomic aberrations in a single run, can successfully be applied in effusions [16][17][18][19]. Most published studies deal with advanced NSCLC; however, research has also shown promise in the application of NGS in MPEs caused by breast, colorectal, ovarian, gastric, and small cell lung cancers, as well as melanoma (Table 5) [24,38,39,53,55].
Recent evidence suggests that NGS in MPEs is highly concordant with NGS in correspondent tissue biopsies, whereas the former might even detect additional driver and resistance aberrations, highlighting intratumoral heterogeneity (Table 1) [26,28,33,37]. In contrast to tissue biopsies, isolating genetic material from effusions is not hampered from artifacts linked with formalin fixation (except for cell blocks), such as the crosslinking between nucleic acids and proteins [15,56].
A growing number of studies point to testing plasma liquid biopsies to guide clinical management in advanced cancer patients, especially when tissue biopsy is insufficient or impossible to obtain [24,26,37,39]. Similar to plasma yet with a superior performance, MPE liquid biopsy allows multiplex molecular testing to identify driver and resistance mutations, disease monitoring through serial collection, and optimal view of tumor heterogeneity in advanced cancer patients [8,19,57].
For any NGS analysis to be successful, standardization of pre-analytical factors is imperative for optimal results [2,16,19,25]. In cytologic specimens like MPEs, pre-analytical factors include sample adequacy, fixatives, preservatives, diverse cytopreparation methods, staining, tumor fraction, nucleic As MPEs could be the first and only specimen received, especially when tissue biopsies are hypocellular or impossible to obtain, their efficient triage and processing is imperative to get maximum clinically relevant information [10,[20][21][22]. Besides the necessary morphologic evaluation and routine ancillary preparations, MPEs can provide robust theranostic information similar to tissue biopsies. Both cell pellets-including all preparations derived from it like direct smears, LBC slides, and FFPE cell blocks-and post-centrifuge supernatants can subsequently be processed for molecular testing [2,14,15]. This is critical for cases like advanced NSCLC, where testing for multiple biomarkers is recommended to tailor targeted oncologic therapies, according to the latest ASCO guidelines. NGS, a molecular technique that can detect multiple genomic aberrations in a single run, can successfully be applied in effusions [16][17][18][19]. Most published studies deal with advanced NSCLC; however, research has also shown promise in the application of NGS in MPEs caused by breast, colorectal, ovarian, gastric, and small cell lung cancers, as well as melanoma (Table 5) [24,38,39,53,55].
Recent evidence suggests that NGS in MPEs is highly concordant with NGS in correspondent tissue biopsies, whereas the former might even detect additional driver and resistance aberrations, highlighting intratumoral heterogeneity (Table 1) [26,28,33,37]. In contrast to tissue biopsies, isolating genetic material from effusions is not hampered from artifacts linked with formalin fixation (except for cell blocks), such as the crosslinking between nucleic acids and proteins [15,56].
A growing number of studies point to testing plasma liquid biopsies to guide clinical management in advanced cancer patients, especially when tissue biopsy is insufficient or impossible to obtain [24,26,37,39]. Similar to plasma yet with a superior performance, MPE liquid biopsy allows multiplex molecular testing to identify driver and resistance mutations, disease monitoring through serial collection, and optimal view of tumor heterogeneity in advanced cancer patients [8,19,57].
For any NGS analysis to be successful, standardization of pre-analytical factors is imperative for optimal results [2,16,19,25]. In cytologic specimens like MPEs, pre-analytical factors include sample adequacy, fixatives, preservatives, diverse cytopreparation methods, staining, tumor fraction, nucleic acid extraction protocols, and DNA/RNA input [19,25]. NGS performance seems to deteriorate when cellularity is low, yet it is more effective than Sanger sequencing in hypocellular contexts [38,42]. Of interest, effusion-derived supernatant-a material normally discarded during fluid preparation -has provided robust molecular analysis results [9,15], exhibiting high concordance to tissue and superior performance to plasma liquid biopsy. Apart from MPEs, literature has also shown optimal results in supernatants derived from fine needle aspirations (FNAs) of diverse cancers [15,58,59]. Normally, we need to assess tumor cellularity before every molecular test, yet this is impossible to do when utilizing supernatants. However, the latter contain plenty of ctDNA derived from the high turnover of cancer cells, the active or passive release through necrosis and apoptosis, and potentially the disruption of cancer cells during centrifugation [9,15,60].
Apart from standardizing pre-analytic factors, implementing a complex, high-throughput procedure such as NGS in a clinical laboratory requires additional steps. After selecting the most suitable platform, gene panel, and enrichment method, each NGS assay needs to be validated, as it is imperative to establish its diagnostic performance (e.g., analytical sensitivity and specificity) [18]. Within this framework, laboratory personnel should optimize a workflow that includes library preparation, sequencing, and big data analysis [16,19]. Data mining begins with base-calling and continues with sequence alignment, variant detection (e.g., single nucleotide variants, indels, copy number variants, and gene rearrangements), and annotation, before reporting to the patient. To carry out this pipeline, it is necessary to closely collaborate with bioinformaticians, who apply sophisticated algorithms and filter background "noise" [16,18,19]. Of interest, a bigger database is created with each NGS run, allowing comparisons besides the discovery of novel genetic aberrations. However, reporting incidental findings of unknown clinical significance to the patient carries significant ethical and legal implications [18]. Except for the challenge to implement a complex procedure that generates such massive amount of data into laboratories that routinely perform low-throughput testing, management and storage of big data derived from NGS creates significant bioethics dilemmas related to the protection of patient privacy [61].

Conclusions
In conclusion, molecular analysis of MPEs with NGS is a powerful, high throughput modality able to identify targetable biomarkers, stratify for established targeted therapies and clinical trials, and pinpoint mechanisms of therapy resistance. Pleural fluid liquid biopsy is minimally invasive and often permits repeated follow-up, while it captures tumor heterogeneity more efficiently than tissue biopsy. Evidence suggests that NGS in MPEs has comparable performance with tissue and is more effective than plasma liquid biopsy, especially when post-centrifuge supernatants are used. In addition, NGS exhibits superior results to Sanger sequencing in hypocellular fluid specimens. Despite the aforementioned promising results, evidence is still premature and potentially misleading, as it mostly comes from small and, to a great extent, retrospective studies of low power or case reports. In this context, large studies in the form of randomized controlled trials would be of significant value. Furthermore, validation is still needed at both pre-analytical and analytical levels. Pathologists should triage MPEs efficiently, as multiple diagnostic, prognostic, and theranostic tests need to be performed from each sample. For instance, the cell pellet could be used for subsequent morphologic evaluation and routine ancillary techniques including HER2 or PD-L1 IHC, while the supernatant should not be discarded albeit saved for high-throughput molecular analysis, especially when the cellular material runs out. Lastly, when tissue biopsy or FNA are paucicellular or impossible to retrieve and a MPE is formed, molecular analysis of the latter can be favored over plasma liquid biopsy in advanced cancer patients.