Early mutational events and clonal dynamics in normal crypts: implications for colorectal tumorigenesis

Abstract

Colorectal cancer has widely been described by the classic adenoma-carcinoma sequence, where sequential mutations in APC, KRAS, and TP53 drive the transition from normal epithelium to carcinoma. However, recent high-resolution genomic analyses have expanded this framework by revealing the presence of multiple driver mutations within morphologically normal colonic crypts. In this review, we summarize the emerging landscape of several somatic driver mutations in normal crypts, highlighting genes that were reported to be under positive selection, such as FBXW7, STAG2, AXIN2, PIK3CA, ERBB2, and ERBB3, and discuss their known functions in both normal epithelium and colorectal cancer. We also postulate the functional consequences of these mutations in the overall priming of colonic crypts to potential tumorigenesis. Despite their presence, the frequencies of these mutations in both normal epithelial and colorectal cancers are low, and most of them do not progress to malignancy. We therefore discuss the intrinsic and extrinsic factors that dictate the fate of these crypts. Together, these findings underscore that tumor-initiating events in colorectal cancer may occur much earlier than previously recognized. A deeper understanding of these early driver events may help to better inform strategies for early detection, risk stratification, and prevention before these morphologically normal crypts transform into neoplasia.

Introduction

Colorectal cancer (CRC) has traditionally been framed through the lens of the adenoma-carcinoma sequence, in which normal epithelial cells acquire a series of somatic mutations that confer selective advantages, transforming them into abnormal lesions and eventually carcinoma [1]. This model outlined the genetic roadmap of CRC initiation and progression through stepwise acquisition of canonical Vogelstein driver mutations in APC, KRAS, and TP53 [2].

The classic Vogelstein model implies a homogeneous mutational trajectory in CRC. However, multiregional sequencing studies showed extensive intratumoral heterogeneity, indicating that CRC progression is more complex than simple linear clonal expansion [3, 4]. The existence of significant heterogeneity supports the Big Bang model, in which there is an early burst of genetic diversity, followed by neutral evolution with little subsequent selection and selective sweeps [5–7]. Branching evolution, in which multiple subclones are derived from a common ancestor and persist in parallel before diverging, was also proposed as a viable model [8]. In addition, intratumoral heterogeneity can arise from niche-dependent selection where tumor microenvironment and spatial factors play critical roles in determining the fate of each clone, independent of their intrinsic genetic factors [9]. Mathematical and computational models support the co-existence of different evolutionary models at different stages [10], suggesting that CRC evolution may not be model-exclusive.

Regardless of the model, the underlying triggers for this marked intratumoral heterogeneity have not been uncovered. The advent of next-generation sequencing has made it possible to peer into the early stages of tumor initiation, in which tissue appears phenotypically normal, with an eye towards uncovering the first changes that set a cell on its path towards cancer. In the colon, somatic driver mutations have been detected in a small fraction of morphologically normal colonic crypts [11]. The functional impact of these driver mutations and how they fit into the evolutionary trajectory of CRC remains elusive. Broadly, these mutations could function in one of the following ways in relation to oncogenesis. They may act as (1) a bystander with minimal effect, (2) in a cell-interactive manner, or (3) cell-autonomously. Here, we aim to summarize the current landscape of somatic mutations which occur in the normal colon and hypothesize the mechanisms on how such mutations either inhibit or stimulate cancer. By bridging crypt-level mutational events with the broader process of CRC evolution, we aim to map the events that precede and potentially initiate the adenoma-carcinoma sequence, thereby highlighting opportunities for early risk stratification and clinical intervention before neoplastic transformation occurs.

Somatic mutational landscape of normal colonic crypts

The colonic epithelium is organized into approximately 15 million crypts, each comprising about 2000 cells derived from a single ancestral stem cell located at the base of each crypt. Thus, the presence of a somatic mutation in one stem cell will be reflected in all cells within that particular crypt. This crypt clonality allows for the study of early mutational accumulation, evolution, and selection in normal colon [11].

Recent advances in high-resolution sequencing and genomics have enabled us to dissect the evolution of somatic mutations at the resolution of individual crypts. It was found that virtually all normal colonic crypts from healthy, cancer-free 50–60 year old individuals harbor somatic mutations, including both neutral and potentially functional changes, with a median of around 3000 single nucleotide variants and 300 insertions/deletions, and a linear accumulation rate of approximately 43.6 mutations each year [11]. Of these accumulated mutations, only ~ 1% of normal crypts harbor mutations that are under positive selection and confer clonal advantages.

It is important to note that mutant crypts do not exist in isolation and their ability to achieve clonal fixation is influenced by their spatial organization within the colonic epithelium, where crypt-dense regions show an increased likelihood of mutant crypt persistence [12]. Once a mutation persists within the crypt, the mutant crypt can expand through the process of clonal fission, dividing and colonizing adjacent crypts to form larger clonal patches. This clonal expansion through fission contributes to field cancerization, in which multiple mutant crypts form patches of genetically related epithelium. These expanded fields increase the likelihood that subsequent driver mutations arise within an already advantaged clone, making them more relevant to early tumor initiation despite the tissue remaining morphologically normal [7].

Despite the accumulation of mutations over time, most of these mutations do not uniformly translate to a malignant phenotype, with only a small but significant subset of normal colonic crypts harboring mutations in genes under positive selection. These positively selected mutations include truncating variants in AXIN2 and STAG2, and hotspot missense mutations in FBXW7, PIK3CA, ERBB2, and ERBB3, which are enriched beyond expectation based on mutational burden, and are likely involved in crypt expansion [11]. In contrast, the canonical Vogelstein CRC drivers, such as APC, KRAS, and TP53, are rarely mutated in normal crypts [13], suggesting that their appearance may mark a transition to well-established neoplasia rather than early survival or expansion advantages.

The functional consequences of these positively selected mutations are not well-understood. They may provide subtle stem cell proliferative advantages that drive crypt fission [14], or may act as bystanders and exert effects upon acquiring additional advantageous mutations [15]. Additionally, their effects may be crypt-autonomous with minimal influence from external factors, or crypt-interactive, shaped by their interaction with neighboring crypts and microenvironment [16].

The phenotypic effect of a given driver mutation is also dependent on the order of acquisition. Experimental and mathematical models indicate that mutation order is dynamic and can significantly influence the trajectory and spatial progression of tumorigenesis, as well as shaping tumor heterogeneity, metastatic potential, and subtype differentiation. For example, the order sequence in which FBXW7 and other driver mutations like APC occur can dictate whether cells remain in a near-normal state or acquire cancer stem cell features. Another study demonstrated that Vogelstein-order editing of APC → KRAS → TP53 in normal organoids resulted in upregulation of hydrogen sulfide-producing enzymes and metabolic reprogramming to promote tumorigenesis features, however, when this order was altered, they yielded varying phenotypic outcomes [17]. Taken together, crypt-level analyses show that normal epithelium is not devoid of mutations, rather it consists of a variety of mutant clones that accumulate with age, and are influenced by genetic alterations, as well as the crypt microenvironment. Mapping early mutational events and their dynamic order in normal colon can provide insights into clonal evolution of CRC and understanding why certain clones remain benign while others evolve towards malignancy.

Functions & implications of driver mutations in normal colonic epithelium

The detection of positively selected driver mutations in phenotypically normal colon crypts indicates that genetic alterations can occur before any notable neoplastic changes. While functional consequences of these early alterations remain uncertain, there is a possibility that they influence the trajectory of colonic crypts in ways relevant to CRC development. In this review, we focus specifically on genes identified by Lee-Six et al. as being under significant positive selection in normal colonic crypts based on dN/dS modelling of truncating and hotspot missense mutations. To better understand the significance of these early driver events, we will review the known functions of these genes and their respective mutations in both normal epithelium and CRCs, and hypothesize their potential roles in contributing to CRC initiation and tumorigenesis (Table 1).

Table 1.

Summarized known and postulated functions of candidate positively selected driver mutations identified by Lee-Six et al. in normal colonic crypts

Gene	Type of mutation(s) or inactivation	Known function of mutations	Potential consequence in normal colonic crypts	Context-dependent effects
FBXW7	Missense	Impaired degradation of oncoproteins	Cell-autonomous (inhibitory): Alter genomic landscape and cellular plasticity in normal crypts, abrogates the effect of later APC mutation, thus preventing colorectal cancer tumorigenesis	Order of mutations confer varying phenotypic effects and cell states (may be cell-autonomous inhibitory or stimulatory)
STAG2	Truncating, Loss of expression	Chromosomal instability, telomere recombination, prolonged stem cell lifespan	Bystander: Insufficient alone, Prolonged lifespan may prime the stem cell for the acquisition of additional mutations	-
			Cell-autonomous (stimulatory): Increased double stranded breaks and chromosomal instability. Exhibited early clonal dominance
			Cell-interactive: Loss results in increased oncogeninc pathways, including TNF-α and KRAS, and Ki67 marker in wild-type colon organoids
AXIN2	Truncating, CN-LOH, promoter methylation	Impaired β-catenin degradation, mild Wnt activation, regulate Snail1 to induce EMT	Cell-autonomous (stimulatory): primes Wnt signaling and crypt expansion	Expression is epigenetically regulated. Methylation context (MSI vs. MSS, age, serrated vs. adenoma)
PIK3CA	Missense	Constitutive PI3K signaling	Bystander: Insufficient alone, prime the cells for additional mutations, such as KRAS, BRAF, and TP53	Bystander effect in early stage of colorectal cancer, but stimulatory at late stage of colorectal cancer
			Cell-autonomous (stimulatory): Upon acquiring additional mutations, PIK3CA mutations may contribute to colorectal cancer progression
			Cell-interactive (stimulatory): Paracrine secretion via exosomes
ERBB2/3	Missense	Ligand-independent signaling, increased crypt fission	Cell autonomous (stimulatory): Proliferative advantages that enable crypt expansion, may be an early-stage event	ERBB3 mutation may increase or decrease polyp number and size, as well as proliferation in different genetic backgrounds
Other genes	Various	Genomic instability, cell cycle, chromatin remodeling, altered Wnt signaling	Bystander: Likely acting as proto-drivers, insufficient individually	-

FBXW7

FBXW7 encodes for F-box protein that is part of the substrate recognition component of the SKIP1-Cullin1-F-box E3 ubiquitin ligase complex [18]. This complex is responsible for the degradation of several oncoproteins including cyclin E, c-Myc, Notch, c-Jun, β-catenin, and EGFR, thereby regulating cell cycle progression, differentiation, and maintaining genomic stability [19]. Thus, it functions as a tumor suppressor in normal colon.

Mutations in FBXW7 are frequent, ranging from 7.5% to 18% in large patient cohorts. In CRC, this frequency is approximately 16% [20]. Majority of these mutations are missense and occur in hotspot mutations within the WD repeat domains, notably the R465, R479, and R505, that are critical for substrate recognition [21]. These mutations lead to impaired degradation of oncogenic proteins and are associated with aberrant nuclear/cytoplasmic localization of β-catenin. Knockout of FBXW7 resulted in rapid budding events in the crypt regions of colon organoids [22]. Additionally, these mutations are associated with higher tumor mutation burdens, higher microsatellite-unstable (MSI), and lower chromosomal instability scores in CRC [23].

The loss of FBXW7 function also resulted in resistance to various therapies. Notably, the mutations results in the stabilization of Myc and cyclin E that allow the bypass of external Wnt signaling, leading to intrinsic resistance to Wnt-blocking therapies such as ETC-159 [24]. CRC tumors carrying FBXW7 mutations were also reported to exhibit resistance to multi-kinase inhibitors targeting the RAS/RAF/MEK/ERK pathway such as regorafenib and sorafenib, caused by the inability to degrade the anti-apoptotic protein Mcl-1, thus preventing inhibitor-induced apoptosis [25].

Despite their oncogenic potential, the clinical impact of FBXW7 in CRC remains inconclusive. A large-scale study found no association between FBXW7 mutational status and patient prognosis [26], whereas another study reported that metastatic colorectal adenocarcinoma patients with FBXW7 missense mutations had significantly worse overall survival compared to those with wild-type FBXW7, and were significantly associated with PIK3CA mutations with a trend towards co-occurrence [27]. Liu et al. reported that while patients with FBXW7 mutations showed better overall survival, the R465C showed a contrasting trend, suggesting specific mutational effect. Beyond prognosis, FBXW7 mutations are also associated with higher immune infiltrations, specifically M1 macrophages and CD8 + T cells, as well as significant enrichment of multiple immune-related gene sets [23]. In contrast, Cho et al. reported that FBXW7 mutations in CRC enabled the evasion of lytic T-cell immunosurveillance [28].

There are also contrasting reports on the metastatic potential of FBXW7 mutations in CRCs. Preclinical study by Li et al. showed loss of FBXW7 resulted in impaired degradation of epithelial- mesenchymal transition (EMT) transcription factor ZEB2 in a phosphorylation-dependent manner, thus promoting EMT [29]. Complementary to this study, it was reported that FBXW7 mutations are associated with higher risk of lymph-node metastasis [30]. Conversely, FBXW7 mutations were observed to be more frequent in patients with non-organ metastasis, with negative correlation observed between organ metastasis and FBXW7/Notch mutations in CRC [31].

The conflicting results on the role of FBXW7 mutations in CRC may be attributed to several factors, such as the differential functional consequences of distinct FBXW7 mutations, the regulatory relationship between TP53 status and FBXW7-mutants [23], and the order of mutations within the adenoma-carcinoma sequence. Specifically, normal colon organoids maintained a near-normal phenotype and repressed transcriptional response to APC loss when FBXW7 was mutated before APC. In contrast, APC mutation preceding FBXW7 mutation results in activation of transcriptional programs associated with cancer stemness and dedifferentiation [32]. This highlights a more context-specific effect of FBXW7 mutations in the overall CRC initiation. The shift in cell state induced by FBXW7 mutations in normal colon may initially reduce susceptibility to subsequent effects of APC mutation in a cell-autonomous inhibitory manner (Fig. 1a). However, FBXW7 loss may also create selective pressure for additional oncogenic hits that drive transformation in CRC.

Fig. 1.

Functional consequences and potential trajectories of positively selected driver mutations in normal colon crypts. a Known functional consequences of FBXW7 and STAG2 mutations and their trajectories in normal colon. b Postulated functional consequences of AXIN2, PIK3CA, ERBB2, and ERBB3 mutations and their potential trajectories in normal colon. Created with BioRender.com

Evidence showed that TP53 and KRAS were the most frequently co-occurring mutations in FBXW7-mutated CRCs [18]. Following FBXW7 loss, accumulation of TP53 loss-of-function mutations typically occurs [33]. While the exact mechanisms of how FBXW7 mutation promotes additional mutations in colonic epithelial cells remain unclear, it was found that silencing of FBXW7 resulted in cellular senescence in p53 wild-type cancer cell lines [34]. Complementary to this study, FBXW7-mutated CRC cells also demonstrated increased phosphorylation of p53 at Ser15 residue compared to its wild-type counterparts, and this increased p53 activation promoted resistance to oxaliplatin [35]. These studies highlight that FBXW7 mutation creates a selective pressure for TP53 inactivation, thereby promoting CRC progression. Given the paradoxical, context-dependent functions of FBXW7 in normal colon, more research is warranted to determine how the temporal order of FBXW7 mutations influences the mutational trajectory of CRC tumorigenesis.

STAG2

The STAG2 gene encodes for a key component of the cohesin complex and is crucial for the maintenance of genomic stability through its role in sister chromatid cohesion, DNA repair, gene expression regulation, and chromatin structure [36]. Mutations or loss-of-function in STAG2 are amongst the most common cohesin-related gene alterations in cancer, with truncating mutations identified as potential driver events in several tumor types [37]. STAG2 mutation occurs at 0.9% frequency in colonic adenocarcinomas [38], a number lower than other reported tumor types. Lee-Six et al., identified two truncating mutations that confer proliferative advantage in normal colon crypts [11]. Somatic mutations are infrequent in the context of CRC, and the main mechanism of STAG2 inactivation is loss of expression which was detected in 23% of all CRCs and is considered to be an early event in CRC development [39].

There have been conflicting reports on STAG2’s role in CRC. On one hand, knockout of STAG2 has been shown to increase double-stranded breaks and chromosomal aberrations by impairing homologous recombination repair. This conferred increased sensitivity to ATM and PARP inhibitors, suggesting a tumor-suppressive function. Moreover, STAG2 mRNA expression is observed to be higher in tumor tissue than normal colon [40]. Our recent study involving the co-culture of STAG2 knockout colon organoids with wild-type organoids observed an increase in Ki67 proliferation marker and upregulation of oncogenic pathways such as KRAS, TNF-α, and EMT in wild-type organoids [41]. Conversely, functional studies conducted by Zhou et al. suggested the role of STAG2 as a proto-oncogene in CRC, with its knockdown causing decreased proliferation, migration and invasion, as well as increased cisplatin sensitivity, while overexpression being associated with poorer prognosis and increased cisplatin resistance [42].

Currently, more research on STAG2 mutations and their role in CRC tumorigenesis is warranted. Evidence points to the STAG2 mutation or inactivation being an early-event driver in CRC, as STAG2 mutant stem cells exhibited clonal dominance pattern with approximately 99% chance of taking over colon crypts compared to wild-type cells, a figure that is higher than neutral mutations [43]. STAG2 inactivation alone in normal colonic crypts may be insufficient to induce CRC and is likely to require additional hits for malignant transformation. It was reported that silencing of STAG2 in normal human cells lacking telomerase resulted in increased telomere recombination, delayed telomere shortening and eventual senescence events [44]. This prolonged lifespan extension of normal cells induced by STAG2 inactivation, along with its role in promoting chromosomal instability, may allow for additional accumulation of other driver mutations that confer selective advantages, thus promoting CRC tumorigenesis (Fig. 1a). Another possible role of STAG2 loss-of-function mutations or inactivation in driving tumorigenesis is through immune evasion, as evidenced by an in vitro osteocarcinoma model which showed increased PD-L1 expression, thereby allowing for additional mutations to co-occur with time [45].

AXIN2

AXIN2 is a scaffold protein important for the assembly of the β-catenin destruction complex and plays a key role as a negative regulator of the Wnt/β-catenin signaling pathway, essential for intestinal homeostasis and regulation of stem cell renewal in the colonic epithelium [46, 47]. It is also a direct transcriptional target of the Wnt pathway, forming a negative feedback loop to suppress Wnt activity [48]. In normal colon, AXIN2 is predominantly expressed in the stem cells located at the crypt base [49].

Truncating mutations in AXIN2 leads to impaired degradation of β-catenin, resulting in increased Wnt signaling activity [48]. Lee-Six et al., identified multiple truncating mutations in AXIN2 within these morphologically normal colon crypts, and it is one of the mutated genes that exhibit statistically significant positive selection. Besides truncating mutations, Copy-Neutral Loss of Heterozygosity (CN-LOH) was observed in AXIN2-mutated crypts, resulting in inactivation of the AXIN2 function [11]. Clonal expansion was observed in these crypts, suggesting that the loss of both allelic copies may confer a proliferative advantage. These mutational events are likely to mimic early-stage APC loss, which results in constitutively active Wnt signaling. Although its mutations are relatively rare in colorectal adenocarcinomas (2.3%), AXIN2 is among the few genes under positive selection in normal colonic crypts and plays a key role in the Wnt signaling pathway central to CRC [38]. Interestingly, AXIN2 mutation is more frequently observed in MSI CRC tumors, with 20% of tumors harboring frameshift mutation in exon 7 due to mononucleotide repeat instability [50]. This results in premature protein truncation that deletes the DIX domain responsible for DVL binding and homo-oligomerization [51].

Aside from mutations and CN-LOH, AXIN2 expression is subjected to epigenetic regulation. Hypermethylation in AXIN2 promoter was observed in serrated lesions and MSI cancers [52]. This corresponds to a decrease in AXIN2 expression. The level of methylation was observed to increase from sessible serrated adenomas to MSI carcinomas, suggesting that decreased expression of AXIN2 may contribute to the progression of cancer progression [53]. Another study also reported that AXIN2 methylation is correlated with microsatellite instability [54]. In contrast, AXIN2 was reported to be frequently hypomethylated in microsatellite stable (MSS) CRC, indicating context-specific regulation [55]. The level of methylation may also differ in a spatial and age-related manner. It was reported that methylation level of AXIN2 was significantly lower in tumor tissue than normal mucosa. Additionally, AXIN2 is one of the genes that were associated with age-related increases in CpG island methylation [52]. AXIN2 is demethylated by ALKBH5, resulting in the hyperactivation of Wnt signaling. Subsequently, the Wnt/β-catenin target DKK1 was induced by ALKBH5 to recruit myeloid-derived suppressor cells to induce immunosuppression in [56]. Taken together, these studies imply that epigenetic regulation of AXIN2 is context-dependent, with methylation patterns varying by tumor subtype, tissue, and age, thereby influencing its role in CRC initiation and progression.

AXIN2 is also implicated in tumor progression beyond Wnt signaling. It was reported to upregulate Snail1 activity, driving EMT and promoting invasion and metastasis in CRC. Silencing AXIN2 reduces Snail1 activity, in turn suppresses metastasis and reverses EMT [57].

In summary, mutations in AXIN2 in normal tissue may partially mimic early APC loss and may provide clonal advantage in the early crypt expansion. While its frequency in CRC is relatively low and its role in contributing to CRC tumorigenesis remains elusive, it is possible that mutations in AXIN2 set the stage for additional driver mutations, by acting as a low-amplitude early Wnt-activating hit that primes tissue for transformation prior to APC mutations. Such mutation may slightly enhance Wnt signaling without triggering immune surveillance in normal colon crypts, as a supporting role before the initiation of a full-scale Wnt signaling activation that induces neoplasia (Fig. 1b). Given its role in driving EMT, AXIN2 mutations may also be important in CRC progression when acquired in late-stage neoplasia.

PIK3CA

The PIK3CA gene encodes for p110α catalytic subunit of phosphatidylinositol 3-kinase (PI3K), a key regulator of the PI3K/AKT/mTOR signaling pathway involved in cellular growth, proliferation, survival and metabolism [58]. Dysregulation of PIK3CA is often implicated in many cancers [59]. Approximately 15–20% of all CRC cases harbor mutations in the PIK3CA, and out of these mutations, 80% cluster in three hotspots, E542 and E545 in the helical domains located at exon 9, and H1047 in the kinase domain at exon 20 [60]. These mutations are typically missense and result in a constitutively active PI3K, along with its downstream effectors of AKT and mTOR, promoting proliferation, invasion, metastasis, and apoptotic evasion [61, 62]. Activating PIK3CA mutations have also been associated with clinical resistance to EGFR-targeted monoclonal antibodies (cetuximab, panitumumab) in metastatic CRC, likely by bypassing the need for upstream EGFR activation [63].

While many studies have investigated the role of PIK3CA mutations during or after the transition to neoplasia, their role in normal tissue is under-reported. Lee-Six et al. identified two hotspot mutations in PIK3CA, E542K and R38H, associated with positive selection in normal colon epithelium [11]. The frequency of such mutations is low in normal colon, and gradually increases during the intramucosal to invasive carcinoma progression [64], suggesting that they are likely inducing subtle changes in the early events of CRC tumorigenesis, but play a more prominent role in the progression and invasion at the later stages of this malignancy. Recent findings show that PIK3CA-mutant tumor cells have the ability to influence adjacent normal intestinal epithelial cells through paracrine mechanisms. Specifically, exosomal arachidonic acid is released by the tumor cells and are taken up by the normal cells, inducing chromatin remodeling (H3K4 trimethylation) and promoting malignant transformation [65].

Similar to the other driver mutations in normal colon, PIK3CA mutations alone do not induce CRC tumorigenesis, and additional genetic and epigenetic changes are required to achieve full malignant transformation. In the context of established neoplasia, tumors with PIK3CA mutations are often accompanied by other mutations, namely KRAS, BRAF, and TP53 [66, 67]. Additional PIK3CA mutation, in the presence of other co-occurring mutations such as KRAS and BRAF, is associated with worse prognosis, more aggressive phenotype and increased drug resistance in CRC [59, 68]. Another study also demonstrated that PIK3CA H1047R mutant alone failed to initiate intestinal tumorigenesis, however, upon APC loss, this mutation resulted in the development of highly aggressive and invasive adenocarcinomas in both small and large intestines [69]. Taken together, this suggests that PIK3CA mutation, although exhibiting positive selection in normal cells, is not penetrant alone. Rather, it may prime the crypt for additional mutations in the early stages, and contribute significantly to CRC progression when acquired at the later stages of the adenoma-carcinoma framework (Fig. 1b).

ERBB2 & 3

ERBB2 and ERBB3, also known as HER2 and HER3, are members of the epidermal growth factor receptor (EGFR) family that plays an important role in cell signaling, growth, and differentiation [70]. Mutations and amplification of the EGFR family have been studied extensively as key events implicated in multiple cancers, including CRC. They can form homodimers with each other or heterodimerize with members of their own family to activate their downstream signaling cascades, such as PI3K and MAPK pathways [71].

According to The Cancer Genome Atlas, the mutational frequencies of both ERBB2 and ERBB3 mutations are 4.0% and 6.3% in CRC respectively [72]. Most common hotspot mutations of ERBB2 in CRCs are R678Q, V842I, and S310F/Y [73]. For ERBB3, those affecting V104 residue are reported to transform colonic epithelial cells in a ligand-independent manner, however, full oncogenic activity requires ERBB2 [74]. Both ERBB2 and ERBB3 mutations are significantly associated with MSI CRCs and frequently co-occur with PIK3CA mutations via cooperative activation of the PI3K-AKT/MAPK signaling [72], but not with KRAS, BRAF, and TP53 mutations [72], suggesting a distinct molecular pathway for tumorigenesis of CRC. Additionally, they have been implicated in resistance to anti-EGFR therapy in CRC via activation of compensatory EGFR feedback loop [75]. On the other hand, ERBB2/3 mutations have potential predictive factor of benefiting from regorafenib + anti-PD-1 combination therapy in refractory MSS metastatic CRC patients [76].

In normal surface colon epithelium, both ERBB2 and ERBB3 are weakly expressed, and mutations are uncommon [77, 78]. The mRNA levels of ERBB2 and ERBB3 were found to be 2.3-fold and 1.5-fold higher in CRC samples than normal colon samples respectively [78]. In CRC, ERBB2 overexpression is associated with the transcriptional upregulation of Wnt/ β-catenin-target genes, indicating transcriptional cross-talk between ERBB2 activity and Wnt pathway dysregulation [79].

Despite this, ERBB2 and ERBB3 mutations are relatively enriched within the minority of crypts that harbor driver mutations, accounting for much higher frequency than CRCs [11]. The hotspot mutations of R678Q, V842I, and T862A in ERBB2 and R475W AND R667L in ERBB3 were reported to be associated with positive growth selection in normal colon [11]. Moreover, the presence of mutations, specifically ERBB2’s R678Q and V842I mutations, in serrated polyps mark early neoplastic changes, with significantly higher risk of polyp recurrence and advanced neoplastic lesion than wild-type and a shorter median polyp-free interval of 15 months compared to 26 months for wild-type [80]. In profiling studies, ERBB3 alterations were observed to frequently co-occur with ERBB2 changes and are associated with high microsatellite instability and tumor mutational burden [81]. The role of ERBB3 may be context-dependent with contrasting effects depending on genetic background. More specifically, the deletion of ERBB3 in heterozygous APC mutant, C57BL/6-ApcMin/+ mice, resulted in increased polyp count and proliferation, but smaller size mediated by increased EGFR signaling. However, on a B6;129–ApcMin/+ background, the same deletion induced a reduction in both polyp number and size, and reduced proliferation. In pure C57BL/6J background mice without APC mutations, the same ERBB3 loss resulted increased polyp number without affecting size [82].

While the role of ERBB2 and ERBB3 driver mutations in normal colonic epithelium remains underexplored, they may confer an early clonal advantage by activating the pro-survival and proliferative signaling through both PI3K and MAPK pathways, in a ligand-independent manner, thus exhibiting increased crypt fission rates. It was reported that ERBB2 mutations are more frequent in early-stage and primary CRC than in metastatic CRCs [83]. Moreover, the overexpression of ERBB3 was observed to decrease at later stages of CRC, compared to early-stage where it was often coexpressed with ERBB2 [84]. These studies suggest that the effects of ERBB2/3 aberrations are more notable in the early stages of CRC, possibly enabling local clonal expansion, but are not always retained during malignant progression. Their significant association with PIK3CA mutations, but not KRAS and BRAF, suggests that the driver mutations in ERBB2 and ERBB3 can bypass RAS-dependent MAPK signaling to allow mutant crypts to expand in KRAS/BRAF wild-type backgrounds. Furthermore, the co-occurrence between ERBB2/3 mutations and PIK3CA mutations may confer additional selective advantages in normal colonic epithelium to transform them into full-blown neoplasia (Fig. 1b). However, it is important to note that different crypt mutational backgrounds can yield varying context-dependent effects of ERBB2/3 mutations.

Taken together with the other early driver candidates, these observations highlight how alterations across distinct pathways can influence early crypt behaviour, including potential points of interaction between genes. To summarize these pathway-level relationships, the reported direct and indirect associations among the six genes are presented in Table 2. Direct associations refer to established genetic or physical interactions, whereas indirect associations refer to genes that do not interact directly but influence shared regulatory pathways.

Table 2.

Summary of reported direct and indirect associations among FBXW7, STAG2, AXIN2, PIK3CA, ERBB2, and ERBB3 in the context of CRC

	FBXW7	STAG2	AXIN2	PIK3CA	ERBB2	ERBB3
FBXW7	-	Indirect	Indirect	Indirect	Indirect	Indirect
STAG2	Indirect	-	No	No	No	No
AXIN2	Indirect	No	-	No	No	No
PIK3CA	Indirect	No	No	-	Yes	Yes
ERBB2	Indirect	No	No	Yes	-	Yes
ERBB3	Indirect	No	No	Yes	Yes	-

Emerging early driver candidates in normal crypt

In addition to the positively selected mutations mentioned earlier, several cancer-associated genes were found to harbor heterozygous truncating mutations in morphologically normal colonic crypts, including ARID2, ATM, ATR, BRCA2, CDK12, CDKN1B, RNF43, TBL1XR1, and TP53, although none reached statistical significance for positive selection across the panel of 90 CRC driver genes [11]. While individually infrequent, these mutations reveal a broader landscape of proto-driver alterations that may subtly reshape crypt biology, and they tend to converge on a limited number of functional pathways relevant to CRC tumorigenesis.

DNA damage response and replication stress (ATM, ATR, BRCA2, CDK12)

ATM, ATR, BRCA2, and CDK12 are regulators of DNA damage response and homologous recombination repair and play important roles in maintaining genomic stability under replication stress [85]. In MSI-positive tumors, heterozygous ATR mutations are frequently observed, and ATR loss contributes to increased genetic instability and early tumor development in mice models [86, 87]. Thus, the detection of mutations in these four genes in morphologically normal crypt may imply early attenuation of DNA repair and checkpoint fidelity. Loss of function of these pathway regulators may allow for persistence of mutant crypts by weakening checkpoint response, allowing cells with replication errors to survive and clonally expand, rather than undergoing normal epithelial turnover.

Cell cycle regulation (TP53, CDKN1B, ARID2)

TP53, CDKN1B, and ARID2 are key regulators of G1/S cell cycle regulation, ensuring that DNA is properly replicated in G1 prior to entry into S phase [88, 89]. Among them, TP53 is the most well-characterized gene in CRC and mutations are typically considered as a late-stage event in the adenoma-carcinoma sequence [90], however, their detection in morphologically normal crypts suggests that early checkpoint attenuation may occur much earlier. Matas et al. reported that TP53 mutations in normal colon are less pathogenic and formed smaller clones than those in CRC. Additionally, among CRC patients, TP53 mutations were found to be different in normal colon samples than in synchronous CRC tumors, suggesting that these mutations arise independently in normal tissue and are not from by seeding events from its adjacent tumor [91]. Mutations in these checkpoint regulators imply impaired G1/S checkpoint regulation, which may permit uncontrolled cell proliferation even in the absence of additional oncogenic hits. Over time, this deregulation could enable mutant crypts to persist and acquire new mutations that may increase their likelihood of malignant transformation.

Wnt pathway modulation (RNF43, TBL1XR1)

RNF43 and TBL1XR1 are both directly involved in the Wnt/β-catenin pathway, and their mutations in morphologically normal crypts suggest early alterations to Wnt sensitivity prior to APC loss [92]. RNF43 encodes an E3 ubiquitin ligase that targets Frizzled receptors for degradation, thereby restricting Wnt ligand responsiveness. Loss-of-function RNF43 mutations are well-characterized in MSI-high CRCs, and result in increased Frizzled expression and heightened Wnt sensitivity [93]. Although infrequent in normal crypts, similar to AXIN2 mutations, RNF43 mutations may enhance Wnt signaling, and confer subtle Wnt-driven growth advantage, to allow mutant crypts to persist long enough to acquire more penetrant oncogenic events to occur.

Taken together, the presence of these proto-driver events suggests that defects in DNA damage response, cell cycle regulation, and Wnt-pathway modulations may subtly contribute to early deviation from epithelial homeostasis, potentially setting the stage of the acquisition of later, more penetrant driver mutations in morphologically normal crypts.

Transformation of mutated colonic crypt to a tumor-initiating clone

The earliest stage of CRC tumorigenesis typically begins with the emergence of mutant stem cells at the crypt base within normal colonic crypts [15]. Positive selection derived from mutations may promote clonal expansion through crypt fission, a process in which one crypt divides into two. In normal colon, this event is infrequent and only occurs once every 27 years per crypt on average [14]. However, the presence of these advantageous driver mutations may accelerate this process. For example, colonic crypt with STAG2 loss-of-function mutation was reported to yield 13-fold higher lifetime expansion coefficient and a 3-fold increase in crypt fission rates compared to wild-type crypts [14]. In the same study, KRAS G12D mutation yielded 155-fold higher lifetime expansion coefficient compared to wild-type crypts. These findings are in line with the findings from another study showing elevated crypt fission rates of KRAS mutants compared to wild-type [94]. These rapidly expanding mutant crypts can propagate across adjacent crypts and gather additional mutations or epigenetic alterations that have the potential to drive CRC tumorigenesis. However, clonal expansion does not always result in malignant transformation. Despite the ever-increasing rates of mutations accumulated in the colon epithelium, most will not be transformed fully into CRC and these mutated crypts can coexist within morphologically normal tissue [91].

One of the main obstacles that a mutated crypt has to overcome to be a tumor-initiating clone is insufficient oncogenic potency. Mutant crypts must outcompete the neighboring wild-type crypts to expand at a rate that is competitive enough to overcome the threshold for malignancy [95, 96]. This phenomenon was observed in a study which quantified that KRAS and APC mutant intestinal stem cells (ISCs) have increased probability of clonal fixation at 75% and 55% respectively. Despite the increased clonal fixation and growth advantages, these numbers mean that about 25% and 45% of these mutant cells can still be replaced by normal cells [94]. Mechanistic study showed that APC-mutant ISCs function as supercompetitors via secretion of Wnt antagonists that induced differentiation and elimination of adjacent wild-type ISCs to facilitate clonal fixation [97]. It is plausible that the clonal fixation rates of the positively enriched driver mutations specified by Lee-Six et al., are lower than canonical drivers such as KRAS and APC, and may require sequential mutations to outcompete replacement by normal cells and establish themselves as a tumor-initiating clone.

Even in the presence of advantageous driver mutations, epigenetic regulations are critical in determining whether a mutant crypt can progress to adenoma or remain benign. It was observed APC-null crypts that are not part of an adenoma do not possess an increased open enhancer state than those that are part of adenoma [12], suggesting that despite identical genetic landscape, their epigenetic landscape may differ significantly. Furthermore, as mentioned previously, age-related CpG island hypermethylation in AXIN2 is context-dependent, in which hypermethylation was observed in serrated lesions and MSI CRCs while hypomethylation was observed in MSS CRCs [52, 55]. Histone modifications and alterations in chromatin remodelers, such as SWI/SNF complex members [98], as well as FBXW7 and STAG2, may contribute to lineage plasticity and stemness, priming crypts for potential tumorigenesis. Overall, these studies highlight that even with identical genetic makeup, epigenetic regulation is influenced by the location and disease subtype, and plays a key role in determining the fate of a crypt harboring driver mutations.

The progression of a mutated crypt to a tumor-initiating clone is not only shaped by its intrinsic genetic and epigenetic alterations, but also spatial and microenvironmental factors. Studies showed that mutant crypts located in regions of low density and lacking neighboring crypts have reduced likelihood to establish clonal dominance as crypt-crypt cooperation is typically required to promote clonal expansion [12]. The immune microenvironment also plays a part in placing selective pressure on mutant crypts. The presence of CD8⁺ T cells, macrophages, and innate lymphoid cells can limit early neoplastic growth, and immune editing mechanisms are required for persistence of mutant crypts. Conversely, inflammation induced by immune cells may also promote tumorigenesis by secreting reactive oxygen species and pro-inflammation cytokines that activate the NF-κB signaling pathway [99]. Aside from the immune microenvironment, the gut microbiota is also reported to influence tumor initiation. For example, Fusobacterium nucleatum can induce inflammation, and activation of the Wnt pathway, while Enterobacteriaceae secretes specific metabolites that induce inflammation and genomic instability to promote tumorigenesis. Conversely, commensal-derived metabolites such as butyrate were shown to support epithelial differentiation and suppress proliferation [100].

Lastly, the order of acquired mutations is critical in determining phenotypic outcome. As highlighted earlier, there are distinct phenotypic changes in cell state when FBXW7 is mutated before or after APC [32]. A mathematical model simulating the order of 3 Vogelstein mutations also predicted probabilities of APC loss as the first event, followed by KRAS mutation in a large majority of CRC cases. However, when KRAS mutation becomes the first event, it is virtually always followed by APC → TP53 loss [13]. Similarly, TP53 and KRAS mutations often co-occur after FBXW7 mutation [33, 101], suggesting that certain sequences are preferentially selected for CRC tumor formation. In summary, these positively selected driver mutations in normal tissues may drive early clonal expansion, but are typically insufficient for cancer initiation. The context such as tissue architecture, competitive landscape, order of mutation, and microenvironmental conditions can influence the outcome of a mutant crypt, whether it remains benign or transforms into a malignant tumor (Fig. 2).

Fig. 2.

The trajectory of a mutated colonic crypt to either benign or carcinoma. Created with BioRender.com

Conclusion and future perspectives

The discovery of positively selected driver mutations within histologically normal colonic crypts expands the knowledge of tumor-initiating events that occur before the widely recognized earliest initiating APC mutation in normal colonic crypts. Our review highlights that within a normal individual, several mutations arise, some of which may influence crypt dynamics that drive potential malignant transformation. In the right context and environment, these driver mutations may gain several oncogenic hits or interact with neighboring crypts to accelerate crypt fission and subsequent progression to CRC.

It is important to note that current insights into early mutational events in normal colon are largely based on key crypt-sequencing studies, such as those by Lee-Six et al., and many proposed early roles for these mutations remain provisional in comparison to their roles in established CRC studies. Expansion of crypt-level genomic datasets will be essential to strengthen these emerging frameworks. Furthermore, additional research is warranted to delineate the functional consequences of these early driver mutations and the order of mutations within individual crypts, as well as their interaction with adjacent cells. The evolutionary paths from phenotypically normal mutant crypts to adenomas and carcinomas may be uncovered by longitudinal studies that integrate spatial genomics, lineage tracing, and single-cell multi-omics. Additionally, understanding the mutational combinations or environmental factors that determine whether a mutated crypt stays benign or develops into neoplasia may inform early diagnosis. Ultimately, incorporating the knowledge of early driver events into CRC is clinically meaningful in terms of intercepting CRC at its initial, most treatable stages.

Abbreviations

CRC

Colorectal cancer

MSI

Microsatellite-unstable

CN-LOH

Copy-Neutral Loss of Heterozygosity

MSS

Microsatellite stable

EMT

Epithelial-mesenchymal transition

EGFR

Epidermal growth factor receptor

ISCs

Intestinal stem cells

Author contributions

C.M. conceived and designed the review, conducted the literature search, compiled and analyzed relevant data, prepared the figures and tables, and drafted and edited the manuscript. D.K.H.C. contributed to the conceptual development of the review, provided critical revisions, suggested key references, and offered intellectual input for the manuscript. All authors read and approved the final manuscript.

Funding

Not applicable.

Data availability

Not applicable.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.