Liquid biopsy and tumor heterogeneity in metastatic solid tumors: the potentiality of blood samples

A number of studies in different diseases have investigated CTCs as a way to gain insights in tumor spatial heterogeneity (Table 1).

A different phenotype between primary tumor and CTCs can potentially predict poorer response to conventional anticancer therapy: for example, in metastatic BC (mBC), estrogen receptor (ER)-negative CTCs can be detected in patients diagnosed with ER-positive mBC [20] and HER2-positive CTC clones can be found in patients affected by HER2-negative mBC [21]. The presence of such CTC clones might anticipate failure to established therapies respectively targeting the estrogen axis and HER2.

Similarly, in metastatic castration-resistant PC (mCRPC) patients it is possible to identify CTCs with different levels of regulation of the androgen receptor (AR) signaling pathway [22]. RNA sequencing of individual CTCs showed heterogeneity in the expression of AR alterations, including splice variants, and in the expression of AR-independent pathways, such as glucocorticoid receptor and non-canonical Wnt signaling responsible for resistance to antiandrogen therapies [23].

Moreover, CTCs have shown a possible role as marker of spatial heterogeneity also in metastatic CRC (mCRC). The concordance of the KRAS status between primary tumors and CTCs varies between 50% [26] and 77% [24], supporting the existence of different clones within primary mCRC. Additionally, CTCs from left-side mCRC more frequently display a mesenchymal phenotype coherent with EMT, while CTCs from right-side mCRC show an apoptotic morphology [27].

EMT may be significantly involved in metastasization process of several cancer types; in fact, CTCs from patients affected by Hepatocellular Carcinoma (HCC) display epithelial phenotype at early stages, but they undergo mesenchymal transformation during spreading to metastatic sites [28].

Given the growing interest in the relationship between cancer and immune heterogeneity, immune checkpoint biomarkers have been analyzed on CTCs, in particular in metastatic NSCLC and mBC patients, who show high inter-individual heterogeneity of PD-L1 expression [29]. In particular, CTCs resulted to be more frequently PD-L1 positive in comparison with tissue samples in NSCLC, suggesting that CTCs may reflect spatial heterogeneity better than tissue biopsy [30] or alternatively that PD-L1 positive cells are more likely to acquire features consistent with CTCs.

Finally, a number of studies are exploring whether CTCs-derived information might guide treatment decisions. For example, quantification of the phenotypic heterogeneity of mCRPC CTCs [31] or their expression of nuclear AR splice variant 7 (AR-V7) [25] might help to guide the choice between AR signaling inhibitors and taxanes: lower CTC degree of heterogeneity is associated with better outcomes during AR Signaling Inhibitors (ARSI), while AR-V7 positive CTC predict with better outcomes during taxanes over ARSI.

Although not validated in clinical practice, both the CTCs count and their characterization are being explored as tools for monitoring the evolution of metastatic cancers as well as their sensitivity to anti-neoplastic drugs. For example, the reduction of the number of CTCs during treatment is associated with lower probability of disease progression, and longer progression-free and overall survivals in HER2-positive and HER2-negative mBC patients, but not in triple negative patients [32].

Several other examples of dynamic cellular changes that might be monitored over time with CTCs exists: for example, in mCRC patients it is possible to monitor the mutation status of KRAS in CTCs to anticipate changes in therapy [33]. Alternatively, next-generation sequencing (NGS) can be used in CTCs to assess multiple genes associated to resistance to therapies targeting the epidermal growth factor receptor (EGFR) [34]. Similarly, in EGFR-mutated NSCLC cancer progressing to anti-EGFR TKIs, a number of studies have evaluated the expression of resistance mutations and rearrangements [35], such the EGFR T790M secondary mutation [36] or MET amplification [37]. Finally, CTCs can be used to longitudinally assess the presence and intra-patient heterogeneity of PIK3CA mutations [38] which are associated to resistance to anti-HER2 therapies in mBC patients [39] (Table 2).

The TCR is a polymorphic receptor dictating antigen specificity of the T-cell mediated immunity. TCR sequencing has been used as a tool to measure the heterogeneity of the T cells infiltrating tumor samples and of the immunogenic neoantigen burden [55‐57]. In lung cancer patients, TCR spatial heterogeneity reflected the heterogeneity of the mutational landscape. In particular, the number of ubiquitous and regional TCRs correlated with the number of ubiquitous and regional non-synonymous mutations, respectively [58]. The role of TCR profiling in circulating lymphocytes is less clear. In fact, the TCR repertoires of tumor-infiltrating and matched peripheral lymphocytes only partially overlap [59, 60]. Nevertheless, tumor-specific T cells have been identified in the peripheral blood of cancer patients [61, 62]. Whether these repertoires might predict response to immunotherapy in metastatic cancer patients is still unclear. In melanoma patients, the peripheral TCR repertoire correlated with responses to checkpoint inhibitors [42‐44], and similar results were reported for pancreatic ductal adenocarcinoma and other solid tumors [41]. Similarly, the evaluation of peripheral TCR repertoire of PD-1 + CD8+ lymphocytes also showed promising results as a non-invasive approach for selecting metastatic NSCLC patients who could benefit from immune checkpoint blockade [51, 52, 63].

With the technological advancements brought by flow cytometry first and mass cytometry after, our capability to identify rare subsets of circulating/ leukocytes has grown exponentially [64]. The predictive role of specific leukocyte subsets in patients undergoing anti-PD-1 and anti-CTLA4 immunotherapy has been investigated mostly in metastatic melanoma and lung cancer patients.

The baseline frequency of CD14 + CD16-HLA-DRhi monocytes [45], CD69 + MIP-1β + NK cells [46], and PD-1 + CD56+ T cells [47] were reported to be predictors of clinical response in metastatic melanoma patients treated with anti-PD-1 immunotherapy. In the same setting, the post-treatment increase of the specific subset of central memory CD4+ T cells, harboring the CD27 + FAS-CD45RA-CCR7+ phenotype, was associated to prolonged clinical responses [48]. Moreover, the decrease in the percentage of dysfunctional PD-1 + CD38hi CD8+ cells was also correlated with immunotherapy benefit [49]. In advanced lung cancer patients, baseline percentage of HLA-DR monocytes and dendritic cells also correlated to the response to PD-1 inhibitor therapy [53]. In both lung cancer and melanoma patients, elevated baseline frequencies of CD4 + Foxp3- T cells expressing PD-1 and/or lack of their significant down-modulation after PD-1 blockade resulted in a higher risk of death [54].

Finally, after treatment with anti-CTLA4 therapy, the levels of circulating CD33 + CD11b + HLA-DR- myeloid derived suppressor cells correlated with survival [50] in melanoma patients, as well as distinct CD4+ and CD8+ memory T cell subsets [46].

As more evidences accumulates, it emerges that cellular immunoprofiles and changes in TCR repertoires could predict responses to anti-PD1 and anti-CTLA4 therapies. It is yet unclear whether these biomarkers might represent alternative or complementary analyses in addition to more other markers for immunotherapy response (e.g. immunohistochemical analysis of PD-L1 or tumor mutation burden).

Although profiling of circulating leukocytes is a very promising tool, the absence of standardized protocols and the requirement of very sophisticated and expensive technologies to perform such analyses limit its use in the clinical setting.

Several evidences reported that somatic alterations in commonly mutated genes can be identified in ctDNA and, in some cases, ctDNA allows to detect additional mutations not found by sequencing of a single metastatic lesion. In this regard, Chu et al. showed that the sequencing of ctDNA through NGS systems identifies mutations in ESR1, the gene encoding the ER, not detected in the corresponding biopsy of a metastatic lesion in mBC patients [69]. These results support the increasingly evidences of heterogeneity between different metastatic sites within the same patient. In addition, ESR1 mutations found by dPCR in ctDNA of mBC patients were associated with resistance to endocrine therapy suggesting that ctDNA represents a more sensitive strategy to monitor treatment efficacy [70]. dPCR sequencing of baseline plasma DNA from SoFEA and PALOMA-3 trials identified ESR1 mutations that predicted resistance to exemestane and sensitivity to fulvestrant, helping clinicians to choose the best treatment strategy for patients [71]. Similarly, alterations in phosphatidylinositol 3-kinase (PI3K) PIK3CA gene, the most commonly mutated oncogene in BC, have been identified in ctDNA [72]. Since PIK3CA mutational status can change upon disease recurrence, ctDNA analysis might provide an excellent tool to monitor sub-clonal changes in real-time. Indeed, in the PALOMA-3 study a drop in PIK3CA ctDNA levels after 15 days of therapy with fulvestrant and palbociclib, strongly predicted PFS [73]. Notably, ctDNA is currently gaining momentum not only in the endocrine therapy setting, but also in detecting HER2 mutations. In particular, ctDNA analysis of HER2^mut variant allele frequency demonstrated to be predictive of response to neratinib with sensitivity of 79% and a specificity of 100%, when compared to tumor tissue analysis [74]. Finally, in triple negative mBC patients, a ≥ 10% ctDNA fraction and the presence of copy number gain or amplification at specific loci was associated with significantly worse outcomes [75, 76].

Several studies reported the detection of EGFR activating mutations in ctDNA of patients with advanced NSCLC [77, 78]. The multicenter ASSESS study demonstrated that ctDNA is a feasible sample type for real-world EGFR mutation testing in metastatic NSCLC patients [79]. In 2015, EMA approved the use of ctDNA for EGFR mutation assessment. A year later, FDA approved COBAS EGFR Mutation Test v2 (Roche Molecular Systems, Inc.) for the detection of EGFR mutations in liquid biopsy, when tissue biopsy is not available, to identify metastatic NSCLC patients eligible for anti-EGFR treatment. Importantly, EGFR-T790M mutation, the most frequent alteration associated with TKI resistance, can be detected in ctDNA as an alternative to tumor DNA derived from a tissue sample [80]. Although the concordance is not full, ctDNA can replace the solid biopsy when the last is not feasible, it can be repeated several times and represents an excellent way for monitoring the treatment with anti-EGFR targeted therapies and for identifying novel resistance mutations [81]. Moreover, in metastatic EGFR mutated NSCLC patients, therapeutic response significantly correlate to the longitudinal quantitative changes in plasma ctDNA [82].

ctDNA has demonstrated a particular utility in monitoring treatment response and identifying mechanisms of resistance in metastatic CRC (mCRC). In particular, a high concordance between the mutational status of KRAS, NRAS and BRAF in the tissue and in ctDNA of CRC patients was found [83]. Interestingly, ctDNA analyses often detected KRAS mutations not detected in the surgical specimen [84]. There are evidences that ctDNA analysis could inform about clonal heterogeneity and subclonal changes in real time [85, 86]. In this regard, Siravegna et al. demonstrated that ctDNA detection allowed to track clonal evolution during therapies with anti-EGFR antibodies. In particular, time-course profiles of ctDNA of patients treated with cetuximab and panitumumab revealed that mutant RAS clones, which rise in blood during EGFR blockade, decline upon withdrawal of EGFR-specific antibodies. In this way, EGFR inhibitor can be rechallenged leading again to patients’ response [84]. In addition, longitudinal analysis of ctDNA showed that several mutations rapidly emerge during EGFR blockades, often before radiological relapse [87]. Recently, Kato S et al., reported that 79% of analyzed advanced CRC patients presented ctDNA genomic alterations in TP53 (51% of patients), KRAS (34%), APC (27%), BRAF (16%), PIK3CA (16%), and EGFR (15%) genes. The authors showed that ctDNA could be helpful for clinicians in the definition of the most appropriate treatment for patients; indeed patients who received the matched targeted-therapy showed better responses compared with patients who received unmatched therapies [88].

ctDNA has demonstrated to be a promising source of clinically relevant information also in PC. A recent study demonstrated that primary tissue and ctDNA share relevant somatic alterations, suggesting that either is suitable for molecular subtyping in de novo metastatic castration-sensitive PC (mCSPC) [89]. Other evidences demonstrated that mutations in the AR gene can be effectively detected in ctDNA of mCRPC patients and provide insights into treatment response and resistance [90, 91].

In conclusion, the determination of EGFR mutations in ctDNA to guide anti-EGFR treatment in NSCLC patients is the first and so far only approved ctDNA assay. The main limitation in this field is related to the low abundance of ctDNA, whose detection requires advanced molecular platforms (NGS or dPCR) still too expensive to be routinely used in the clinics. Nevertheless, the continuous development of cheaper high-throughput sequencing techniques will most likely consolidate the evaluation of ctDNA in liquid biopsies as a promising tool in the next future [66].

Among non-coding RNAs, miRNAs are the best promising non-invasive biomarker due to the development of NGS technologies that have enabled the sequencing of the complete miRNA profile. Indeed, several studies identified miRNA as potential biomarkers, useful to monitor cancer progression, patients’ outcomes and the development of chemo-resistance [100] (Table 5). However, although miRNAs have been regarded as promising biomarkers for many years, they have never entered the clinics.

MiRNAs are endogenous, single-stranded, non-coding small RNAs with length of about 22 nucleotide that exert several cellular biological functions inhibiting their target genes [111]. Several overexpressed miRNAs, called oncomiRs, are involved in tumor onset and metastasis, instead those that decreased in cancer patients were considered tumor suppressors [112].

A recent study reported that the expression levels of different specific miRNAs (miR-21, miR-23b, miR-200b, miR-200c) were found higher in metastatic compared to early BC patients. In addition, miR-23b and miR-190 were correlated with low PFS in de novo metastatic BC patients, whereas high levels of miR-200b were associated to decreased OS in the HER2-negative subgroup [101].

The ability of circulating miRNAs to discriminate patients with lymph nodes and/or distant metastasis has been also examined in advanced stages of CRC. In particular, upregulated serum levels of miR-103 were associated with lymph node metastasis and advanced stage of CRC [102]. Similarly, higher serum levels of miR-29a were found in metastatic CRC patients compared to non-metastatic patients. However, this miRNA did not demonstrate a sufficient accuracy in discriminating these groups of patients, showing a sensitivity of 75% and a specificity of 75% [103].. Another study showed instead that plasma expression of miR-203 and miR-141 are able to discriminate between advanced from early stage CRC patients with a good performance [104]. Finally, Tsukamoto et al. have recently reported the levels of miR-21 are elevated in serum exosomes, primary tumor tissues, and liver metastasis tissues from CRC patients. In particular, exosomal miR-21 showed a correlation with liver metastases and TNM stage and was associated with worse OS and disease free survival [105].

Several evidences showed a potential role of circulating miRNA in identifying NSCLC patients with aggressive advanced disease. Indeed, elevated expression levels of miR-222-3p, miR-23b-3p, miR-10b-3p, and miR-21-5p in serum exosomes were associated with poor OS in NSCLC patients. Among these miRNAs, exosomal miR-21-5p overexpression was also associated with presence of liver metastasis and TNM stage [107, 108].

Since cancer heterogeneity is ultimately associated to drug resistance following prolonged treatments, circulating miRNAs could represent potential biomarkers useful to predict drug resistance and select effective treatment strategies. One example is miR-155-5p, whose expression is directly correlated with chemo-resistance and poor prognosis in pancreatic ductal adenocarcinoma patients receiving gemcitabine [110]. Another example is represented by the evidence of lower levels of miR-1914-3p and miR-1915-3p in the plasma of chemo-resistant CRC patients compared to responders [106]. Finally, lower levels of miR-146-5p are found in serum exosomes of cisplatin-resistant NSCLC patients and they are associated to shorter PFS [109].

Circulating miRNAs could represent non-invasive biomarkers, useful not only for large screening programs for tumors with higher sensitivity, but also to dynamically monitor tumor progression and treatment response [100]. Unfortunately, the complexity of RNA biomarker assays and data interpretation contribute to their low success rate and slow down their use into the clinical practice.

Besides the well-known miRNAs, there are other species of non-coding RNAs that are currently emerging as candidate biomarkers assessable in blood-based liquid biopsies, including long non-coding RNAs (lncRNAs), circular RNAs (circRNAs) and PIWI-interacting RNAs (pi-RNAs).

LncRNAs activate or silence the expression of genes regulating chromatin state, mRNA splicing, transport and translation of RNAs, act as competing endogenous RNAs and influence protein modification [113‐115]. Intriguingly, among the target loci of lncRNA there are well-known genomic regions with oncogenic or tumor-suppressive functions, suggesting a key role of these RNAs in the control of cancer progression [116].

For example, the aberrant expression of the lncRNA HOTAIR (HOX antisense intergenic RNA) is associated to tumor proliferation, angiogenesis, progression, drug resistance and worse prognosis [117]. Numerous evidences have focused the attention on the potential role of HOTAIR as a circulating marker and therapeutic target is solid tumors [118]. In this regard, Li et al. demonstrated that high levels of HOTAIR were associated to tumor recurrence, radio-resistance and shorter OS in cervical cancer [119].

Similarly, in metastatic HCC patients, plasma levels of lncRNAs XLOC_014172 and LOC149086, allowed to discriminate, with high performance patients with metastatic disease from patients without secondary lesions [120]. Similarly, Wang et al. showed that serum lncRNA-p21 decreased in liver metastatic cancer patients, respect to non-metastatic HCC patients [121].

Several evidences showed the potential prognostic value of some lncRNAs also in CRC. Among these, higher exosomal levels of the oncogenic 91H were found in CRC patients’ serum and these levels were associated with metastatic disease and tumor recurrence [122]. Similarly, higher serum levels of exosomal CRNDE-h were observed in metastatic CRC patients compared to non-metastatic patients and were positively associated with poor OS [123]. Another lncRNA correlated with outcome in metastatic CRC is GNAT1–1, whose serum levels were decreased in patients with worst prognosis [124].

LncRNAs could represent important biomarkers also to monitor drug resistance that could occur in distinct tumor clones, helping clinicians in choosing patients’ best treatment. Increased lncRNA XIST levels, for instance, were found in serum of CRC patients nonresponding to 5-FU [125]. Similarly, reduced serum lncRNA MEG3 expression identified CRC patients resistant to oxaliplatin-based chemotherapy [126].

Circular RNAs (circRNA) are another class of non-coding RNAs that could be implicated with several human diseases, including cancer [127]. CircRNAs are closed circular one-stranded RNAs that act as competing endogenous RNA sponges modulating the activity of their targets (miRNAs, proteins and RNAs) [128]. Specific circRNAs have been detected in high quantities in blood of patients with different advanced solid tumors representing potential biomarkers useful to monitor tumor progression and response to treatments in metastatic stage [129]. In HCC patients, for instance, high serum expression of circ-ZEB1.33 was associated to tumor progression and was also related to patients’ OS [130]. Other circRNAs were associated to tumor stage and presence of metastases in gastric cancer (GC) patients. In particular, Hsa_circ_0000190 was found downregulated in GC patients’ plasma and was correlated to the presence of metastasis, tumor stage and CA19.9 levels [131]. Also in GC patients, decreased expression of plasma hsa_circ_0000745 was associated to TNM stage [132]. Recently, a key role of circRNA was found also in metastatic urothelial cancer patients, where high levels of circPRMT5 in serum exosomes correlated with the presence of metastasis and tumor stage [133].

PiRNAs regulate gene expression, ensure genome integrity of germline cells and control developmental timing. Aberrant piRNA expression was found in blood and is currently being associated to cancer progression and metastases [134]. Recently, a study demonstrated the positive correlation between piR-54,265 levels and tumor stage in CRC patients, with highest levels in the metastatic setting. Moreover, piR-54,265 lower serum levels were also predictive of chemotherapy response and potentially useful to select CRC patients who benefit from chemotherapy [135].