1. Introduction
Colorectal cancer (CRC) is a leading cause of cancer-related morbidity and mortality worldwide [1]. While the global incidence continues to rise, disproportionately so in low- and middle-income countries, advances in CRC screening and multidisciplinary therapy have significantly improved mortality rates, especially among patients living in highly developed nations where mortality rates have declined in recent years [2]. Despite these improvements, CRC remains the second leading cause of cancer-related mortality in the United States [3].
Advances in molecular sequencing technology (e.g., next-generation sequencing (NGS)) and computational data analytics have greatly improved our understanding of CRC pathophysiology, revealing genomic variants responsible for CRC carcinogenesis. These molecular biomarkers, combined with the emerging fields of radiomics and artificial intelligence, provide clinicians with potentially actionable information. Current treatment algorithms for patients with CRLM incorporate certain biomarkers, including RAS, BRAF, and mismatch repair (MMR) status. Other biomarkers require validation on their clinical utility and are actively being investigated in clinical trials.
In 1988, Vogelstein et al. published a seminal report characterizing mutations related to KRAS, APC, and TP53 at various stages of CRC carcinogenesis, hypothesizing that invasive carcinoma develops from adenomatous polyps via the sequential acquisition of somatic mutations in multiple genes
[4]. Since this foundational work, the development/invention of NGS technology has dramatically advanced the field of oncogenesis with techniques such as whole-genome sequencing, whole-exome sequencing, and targeted sequencing that have revealed a host of genomic alterations
[5].
Figure 1 illustrates the interconnected relationships between multiple signaling cascades critical in CRC oncogenesis, highlighting two frequent mutation genes, the tumor suppressor gene TP53 (pathway A) and proto-oncogene PI3K (pathway B).
Figure 1. Overview of interlinked cellular signaling pathways involved in the proliferation and progression of colorectal cancer. In pathway A, TP53 normally inhibits activated RAS through lethal (Let) 7a23. However, Let-7a is not able to regulate activated RAS if TP53 is mutated. In pathway B, overactive phosphoinositide 3-kinase (PI3K), an oncogene, inhibits glycogen synthase kinase (GSK) 3β24, leading to β-catenin accumulation. EGF, epidermal growth factor; HER, human epidermal growth factor receptor; RTK, receptor tyrosine kinase; MEK, mitogen-activated protein kinase; ErK, extracellular signal-regulated kinase; Mt, mitochondria; IGF-1, insulin-like growth factor 1; IGF-1R/IR, IGF-1 receptor/insulin receptor; PIP2, phosphatidylinositol 4,5-bisphosphate; PTEN, phosphatase and tensin homologue; PIP3, phosphatidylinositol (3,4,5)-trisphosphate; PDK-1, phosphoinositide-dependent protein kinase 1; mTORC1/2, mammalian target of rapamycin complex 1/2; P70s6k, P70s6 kinase; APC, adenomatous polyposis coli. Used with permission
[6].
The CRC-liver metastatic cascade is a complex process wherein a subset of CRC cells acquires the capacity to evade the primary tumor, migrate through the extracellular matrix and neighboring tissue, intravasate, survive transit through the circulation, extravasate, and ultimately colonize the liver
[7]. The molecular alterations required to complete this process create a biologically aggressive and phenotypically distinct disease entity.
Genomic alterations do not always correspond to predictable changes in biological activity
[8]. Proteomics provides clarity to the molecular processes occurring in the gap between gene expression and disease phenotype. Recent proteomic profiling studies revealed unique protein and post-translational modifications present in metastatic CRC tumors, especially in proteins associated with the extracellular matrix, energy metabolism, and immune-cell-related migration
[9][10][11]. In the following section, significant genomic, proteomic, and post-translational mutations in CRLM development are discussed.
2. Genomic Biomarkers
Genome-wide sequencing studies have demonstrated actionable mutations in one-third of patients with CRC metastasis (
Figure 2)
[12][13]. The current National Comprehensive Cancer Network (NCCN) guidelines recommend testing for RAS (KRAS and NRAS) and BRAF mutations and HER2 amplifications, individually or as part of an NGS panel, for all patients with metastatic CRC and universal MMR or microsatellite instability (MSI) testing in all newly diagnosed patients with CRC
[14].
Figure 2. Frequent gene alterations based on wide-genome sequencing of 372 metastatic colorectal cancer (CRC) patients. Note: TSGs: tumor suppressor genes. Open Access Citation: Testa U, Castelli G, Pelosi E. Genetic Alterations of Metastatic Colorectal Cancer.
Biomedicines. 2020; 8(10):414.
https://doi.org/10.3390/biomedicines8100414 [15].
Alterations in the RAS proto-oncogene family, most notably KRAS, NRAS, and HRAS, result in unregulated cell proliferation via gain-of-function activity in the MAPK pathway. RAS mutations are found in up to 52% of patients with CRLM and have a high concordance between the primary tumor and CRLM
[16][17]. Patients with RAS-mutated CRLM have significantly worse recurrence-free and overall survival versus individuals with wild-type RAS
[18][19]. Due to the upregulation of the MAPK pathway, patients with RAS mutations quickly develop resistance to epidermal growth factor receptor (EGFR) antibody therapy
[20][21]. As a result, anti-EGFR antibody agents, cetuximab and panitumumab, are only recommended in the treatment of KRAS and NRAS wild-type tumors
[14]. Furthermore, RAS-mutated CRLM demonstrate more migratory or invasive biology, causing local tumor progression and a higher incidence of micrometastasis compared with wild-type RAS CRLM
[22][23][24]. Consequently, these tumors have a narrower median negative margin (4 mm vs. 7 mm) and double the rate of positive surgical margins after hepatectomy
[25]. Similar results are seen for smaller RAS-mutated CRLM treated with ablation therapy
[6]. The appropriate negative margin distance for patients with CRLM remains a controversial topic
[26]. Where patients with wild-type RAS CRLM benefited from negative surgical margins, patients with RAS-mutated CRLM had similarly poor outcomes between the R0 and R1 resection groups
[27][28]. A recent analysis of 1843 patients with CRLM who underwent curative-intent surgery used AI-based analytics to determine the optimal surgical margin in KRAS-variant CRLM
[29]. The AI model suggested an optimal margin width of 7 mm for KRAS-variant CRLM. Most of the associated prolongation of survival was seen with a 1 mm margin, with the extension from 1 mm to 7 mm contributing a smaller proportion of the improvement in survival. Ultimately, patients with RAS-mutated tumors should demonstrate disease stability on systemic therapy, with the absence of other poor prognostic factors, prior to attempting curative-intent surgical resection.
BRAF, another protein in the MAPK pathway, has emerged as a very poor prognostic indicator
[30]. BRAF mutations occur in 5% of patients with CRLM. BRAF
V600E contains a substitution of valine for glutamic acid at codon 600 and is responsible for over 90% of these mutations
[31]. Patients with BRAF-mutated CRC rarely present with isolated liver metastases. Furthermore, even in the small proportion of patients who present with resectable disease, median recurrence-free survival and overall survival after curative-intent hepatectomy is half compared to wild-type BRAF patients
[32]. Similar to RAS-mutated CRC, patients with a BRAF mutation do not respond to anti-EGFR therapy unless administered with a BRAF inhibitor (i.e., encorafenib)
[33]. Although data are limited to small studies, patients with non- BRAF
V600E mutations may have better outcomes compared with even wild-type BRAF patients
[34].
HER2 amplification is a targetable variant in the MAPK pathway found in 2–3% of metastatic CRCs
[35]. While well-studied in breast cancer, the low prevalence of HER2 amplification in CRC cases limits confident statements about the prognostic effect of HER2 amplification. Regarding targeted therapies, phase 2 trials support the use of a dual HER2 blockade in heavily pretreated, HER2-amplified metastatic CRC
[36][37]. However, anti-HER2 therapy is only indicated in HER2-amplified tumors that are also RAS and BRAF wild-type.
Approximately 5% of patients with metastatic CRC harbor deficient DNA mismatch repair (dMMR), leading to MSI
[6]. Alterations in the MMR system arise through germline mutations (i.e., Lynch syndrome) or sporadically from promoter hypermethylation and silencing of the MMR gene MLH1
[38]. Sporadic mutations are highly associated with BRAF
V600E mutations, present in one-third of dMMR patients, and have a worse prognosis versus patients with a germline dMMR
[39]. In early-stage CRC, dMMR is associated with reduced metastatic potential and a favorable prognosis; however, dMMR status in metastatic CRC results in a worse prognosis compared with proficient MMR tumors
[40]. Metastatic dMMR/MSI-H CRC responds well to immune checkpoint inhibitor (ICI) therapy, doubling the progression-free survival time compared with traditional chemotherapy (16.5 months vs. 8.2 months; HR 0.60,
p = 0.0002)
[41]. In addition, patients with metastatic CRC can have a high tumor mutation burden leading to MSI independent of the dMMR status. Patients with a high tumor mutational burden may also benefit from ICI therapy
[42].
Unfortunately, most of the genomic variants identified in metastatic CRC lack a targeted therapy, including four of the five most mutated genes: APC, TP53, PIK3CA, and SMAD4. However, knowing the status of these mutations does provide prognostic information. Concomitant TP53 and RAS mutations have negative prognostic effects worse than either mutation alone. Specifically, patients with CRLM containing both mutations have significantly worse recurrence-free survival and overall survival
[43]. A similar pattern emerges for patients with tumors containing both an APC and PIK3CA mutation. Together, these mutations have a synergistic effect on chemoresistance and, as a result, a worse prognosis
[44][45]. Not surprisingly, tumors containing both a tumor suppressor mutation (TP53, APC) and an oncogene (KRAS, PIK3CA) are more biologically aggressive, leading to earlier disease recurrence and mortality.
3. Proteomic Biomarkers
Early proteomic studies on CRLM revealed diverse proteomic profiles demonstrating significant metastatic CRC tumor heterogeneity
[46]. Current investigations remain in the preclinical exploratory phase, but these studies have identified promising targets for future research. In particular, studies comparing the proteome biology of primary CRC and CRLM have revealed distinct protein dysregulation profiles
[9][10][47]. Fahrner et al. compared the matched proteomes of seven patients with CRLM and noted upregulated proteins in liver metastasis involving metabolic processes such as gluconeogenesis (pyruvate carboxylase) and fructose metabolism (fructose-bisphosphate aldolase B (ALDOB), fructose-1,6-bisphosphatase (1), as well as proteins linked to the complement cascade, indicating an active immune response)
[9]. ALDOB upregulation indicates a poor prognosis and has been demonstrated to promote tumor progression and CRLM by facilitating the epithelial-mesenchymal transition
[48]. Conversely, comparatively downregulated proteins are involved with cellular structural integrity/cell junction assembly (desmin, synemin, and filamin-C). In a separate study of eight patients with CRLM, several extracellular matrix components were significantly upregulated in CRLM compared with the primary tumor, most notably THBS1, which facilitates CRLM by promoting epithelial–mesenchymal transition
[10]. Furthermore, THBS1 upregulation was associated with shorter overall survival. CD11b and ITGA2 are two additional proteins with a role in promoting the epithelial–mesenchymal transition that are overexpressed in CRLM
[49]. In addition, Ku et al. demonstrated that the filamin A-interacting protein 1-like (FILIP1L) and plasminogen were dysregulated in CRLM compared with primary CRC
[47]. FILIP1L overexpression inhibits WNT signaling and decreases metastatic behavior in CRC
[50]. Ku et al. reported the downregulation of FILIP1L in CRLM samples. In contrast, plasminogen has been noted to be overexpressed in CRLM samples. Overexpression of the plasminogen activating system in CRLM has been associated with worse overall and cancer-specific survival
[51].
The extracellular matrix (ECM) provides physical scaffolding for cells and plays a vital role in biochemical signaling
[52]. Alterations in the composition of the ECM are linked to pathological conditions, including carcinogenesis and metastasis. The ECM in cancer patients is often disorganized and characterized by the upregulation of many components (e.g., collagen) compared with normal tissue
[53]. Naba et al. performed the first proteomic study of the ECM composition in patients with CRLM, demonstrating that the ECM composition of CRLM resembled the ECM of primary CRC more so than a normal liver
[54]. Furthermore, seven proteins were uniquely associated with metastases, including TIMP1 (Tissue Inhibitor of Metalloproteinase-1), a protein that induces pro-invasive ECM remodeling when upregulated and has been associated with poor progression-free survival in patients with CRLM
[55]. In addition, TIMP1 may have utility as a non-invasive biomarker for preoperative risk stratification in patients with CRLM
[55].
Protein post-translational modifications (PTMs) such as crosslinking, acetylation, ubiquitination, methylation, glycosylation, and citrullination can occur on histone and nonhistone proteins. These modifications often contribute to protein degradation to maintain normal physiological homeostasis; however, the deregulation of PTMs supports carcinogenesis and the development of metastasis. Shen et al. constructed a complete atlas of differentially expressed acetylated proteins in primary CRC and paired CRLM
[56]. The authors identified 71 acetylated sites on 55 proteins in CRLM. These proteins were found primarily in the cytoplasm and associated with a broad range of biological processes, including metabolic pathways and carbon metabolism. Of the acetylated nonhistones, TPM2 and K152Ac were the most upregulated, and ADH1B K331Ac was the most downregulated. In another PTM study, Yuzhalin et al. demonstrated that peptidylarginine deiminase 4 (PAD4)-dependent protein citrullination plays an integral role in the development of CRLM
[57]. Citrullination is characterized by a post-translational conversion of arginine residues to citrulline. In the study, CRLM exhibited significantly higher levels of PAD4 and citrullination compared to both healthy livers and the primary CRC. Furthermore, citrullination may facilitate CRLM by promoting the epithelial–mesenchymal transition since citrullination of collagen type 1 in the ECM promoted greater adhesion, decreased migration of CRC cells, and increased expression of epithelial markers.
Recently, proteomics has played a central role in the multiomic evaluation of CRLM. Multiomics provides a more comprehensive evaluation of biological processes by integrating data generated from multiple omic analyses (e.g., proteomic, transcriptomic, and genomic)
[58]. A recent, large multiomic study integrated genomics, proteomics, and phosphoproteomics to provide a global evaluation of 146 patients with CRC, including 43 patients with CRLM
[59]. CRLM were genomically similar to primary CRC but exhibited significant heterogeneity at the proteomic and post-translational level. In addition, in vivo drug testing suggested that phosphoproteomic profiling may more accurately predict drug response to kinase inhibitors than the presence of genomic mutations. Multiomic studies have also demonstrated a correlation with survival
[60][61]. Two separate studies by Ma et al. reported that the mutated peptide number had prognostic value and that somatic variants demonstrated corresponding dysregulated protein abundance and biologic function. Relevant variants identified in CRLM included UQCR5, FDFT1, MYH9, and CCT6A.
4. Liquid Biopsy and ctDNA
A core tumor biopsy represents the current gold standard for obtaining tissue samples for molecular analyses. Core biopsies can be fraught with technical complications and fail to accurately capture complete tumor spatial heterogeneity and tumor evolution.
[62][63]. Liquid biopsy collects body fluid (e.g., blood, urine, saliva) for molecular evaluation and may overcome many of the limitations associated with core biopsy
[64]. Patients with solid tumors release isolated tumor DNA fragments into the blood after tumor lysis or apoptosis, called circulating tumor DNA (ctDNA). ctDNA is distinct from normal circulating-free DNA (cfDNA) and contains specific pathologic genetic variants
[65]. CRCs tend to shed high amounts of ctDNA relative to other solid tumors, making them an ideal candidate for further investigation
[66].
Investigation into the clinical utility of ctDNA as a biomarker for patients with CRC is ongoing. Surgical resection of CRLM remains the standard of care, even for patients who have a clinical complete response (cCR) to neoadjuvant therapy, as available biomarkers cannot accurately identify patients who have achieved a pCR
[67]. Studies specifically evaluating ctDNA in the neoadjuvant setting among patients with CRLM are limited; however, in retrospective studies of patients with stage II-III CRC, the ctDNA status following neoadjuvant therapy was associated with the pCR status
[68][69], disease-free survival, and overall survival
[70][71]. Furthermore, sequential sampling before, during, and after therapy allows for the real-time monitoring of the treatment response and disease evolution. Detection of ctDNA during neoadjuvant therapy correlated with primary tumor regression and shorter metastasis-free survival
[72].
In retrospective studies, ctDNA demonstrates a prognostic value after hepatectomy for patients with CRLM. In a single institution study of 63 patients with CRLM, 42 (67%) had ctDNA present after hepatectomy. These patients had significantly worse overall survival, especially when multiple gene mutations were detected
[73]. Similar results were demonstrated by Tie et al. in a cohort of 49 patients with resectable CRLM who underwent curative-intent hepatectomy
[74]. The 11 patients with positive ctDNA after resection had lower recurrence-free survival and overall survival. However, ctDNA clearance was achieved with adjuvant therapy in three patients, two of whom remained disease free. Patients with positive ctDNA at the completion of therapy (surgery +/− adjuvant chemotherapy) had a 5-year recurrence-free survival of 0% versus 75.6% for patients with undetectable ctDNA after therapy.
Among patients with unresectable metastatic CRC, ctDNA-derived mutational analyses were nearly 100% concordant with the solid tissue biopsies for the detection of multiple clinically relevant mutations, such as the BRAF
V600E, KRAS, and NRAS mutations
[75][76]. Of note, intra-tumor heterogeneity, the treatment effect, or a low disease burden were responsible for the few discordant results. Furthermore, recent phase II clinical trials demonstrate the efficacy of ctDNA-guided rechallenge therapy with anti-EGFR drugs in the RAS wild-type unresectable metastatic CRC. Sartore-Bianchi et al. evaluated 27 patients who were ctDNA negative for the RAS/BRAF/EGFR mutations. Rechallenge therapy with panitumumab was associated with disease control in 59% of patients with a median progression-free survival of 16 weeks
[77]. A separate study by Martinelli et al. included 77 patients with unresectable metastatic CRC that failed second-line therapy after acquiring resistance to first-line chemotherapy plus anti-EGFR drugs. Among the 48 patients with RAS/BRAF wild-type ctDNA, rechallenge therapy with cetuximab plus avelumab improved the median overall survival
[78].
Despite the growing evidence supporting the utility of ctDNA, its widespread incorporation into clinical decision-making algorithms has yet to occur due to some notable limitations
[79]. While patients with CRCs tend to shed higher levels of ctDNA compared with other cancer types, the concentration of ctDNA is still low, especially earlier in the disease process and compared with cfDNA. Highly sensitive tests are required to detect ctDNA accurately. Furthermore, the ratio of ctDNA to cfDNA (ctDNA fraction) significantly influences assay sensitivity and specificity. As a result, variation in the ctDNA fraction leads to unclear thresholds for the limit of detection
[62][65]. Finally, currently available evidence supporting the use of ctDNA is overwhelmingly retrospective. Results from active randomized controlled trials will hopefully provide clarity on the clinical utility of ctDNA in patients with CRLM.