Current Status of Cancer Genomics and Imaging Phenotypes: What Radiologists Need to Know

Abstract

Ongoing discoveries in cancer genomics and epigenomics have revolutionized clinical oncology and precision health care. This knowledge provides unprecedented insights into tumor biology and heterogeneity within a single tumor, among primary and metastatic lesions, and among patients with the same histologic type of cancer. Large-scale genomic sequencing studies also sparked the development of new tumor classifications, biomarkers, and targeted therapies. Because of the central role of imaging in cancer diagnosis and therapy, radiologists need to be familiar with the basic concepts of genomics, which are now becoming the new norm in oncologic clinical practice. By incorporating these concepts into clinical practice, radiologists can make their imaging interpretations more meaningful and specific, facilitate multidisciplinary clinical dialogue and interventions, and provide better patient-centric care. This review article highlights basic concepts of genomics and epigenomics, reviews the most common genetic alterations in cancer, and discusses the implications of these concepts on imaging by organ system in a case-based manner. This information will help stimulate new innovations in imaging research, accelerate the development and validation of new imaging biomarkers, and motivate efforts to bring new molecular and functional imaging methods to clinical radiology.

Keywords: Oncology, Cancer Genomics, Epignomics, Radiogenomics, Imaging Markers

Supplemental material is available for this article.

© RSNA, 2023

Summary

In the era of precision health care, basic genomic concepts and the associated implications in diagnostic imaging will help radiologists expand their impact on patient management decisions for patients with cancer.

Essentials

• ■ Clinical cancer care has entered an era of genomics and precision health care, with radiologists positioned to play key roles in improving outcomes for patients.

• ■ Radiogenomics represents a fast-evolving field aimed at correlating cancer genotypes with imaging phenotypes in the search for new noninvasive biomarkers.

• ■ Radiologists need to understand basic genomic concepts and their impact on imaging in cancer.

Introduction

Clinical oncology has entered an era where diagnosis, treatment options, and prognostication rely heavily on genomic profiles for each patient. Due to the central role of imaging in cancer, imaging specialists need to know the basic concepts of genomics that now underlie clinical oncology. Advances in high-throughput parallel technologies have brought about the “omics revolution,” where omics broadly means “study of the whole.” Cancer genomics focuses on genome-wide DNA aberrations, while transcriptomics and proteomics analyze messenger RNA and proteins, respectively. The growing field of epigenomics uses methods such as methylation profiling to detect reversible modifications in DNA or histones, such as methylation and acetylation, that affect gene expression without altering DNA sequences (Fig 1). Omics-level information continues to advance the understanding of genetic and molecular pathogenesis of cancers (1) and better informs patient stratification at all stages of health care, including prevention, screening, diagnosis, treatment, and follow-up.

Figure 1:

Recent advances in high-throughput parallel technologies have brought about the omics revolution. Epigenomics uses methylation profiling to study reversible modifications in a cell’s DNA or histones, such as methylation and acetylation, which affect gene expression without altering the DNA sequence. Genomics focuses on genome-wide mutations at the level of DNA, using whole-genome, whole-exome, or oncopanel sequencing. Transcriptomics and proteomics allow for consideration of the entire cellular messenger RNA (mRNA) or protein composition, respectively, with technologies including RNA-sequencing (RNA-seq), microarray, and mass spectrometry. Ac = acetylation, Me = methylation.

Cancer originates from genetic mutations in DNA, leading to hallmarks including uncontrollable cell proliferation and escape from cell death. Tumors can be broadly classified as hereditary or sporadic, depending on the predisposing mutations. Hereditary cancer arises from germline DNA mutations inherited from one parent. Although germline mutations exist in all cells, hereditary cancers occur only in characteristic organs, indicating that local environments regulate tumorigenesis. Cancer more commonly arises sporadically from acquired (or somatic) DNA mutations that develop during an individual’s lifetime. DNA replication errors during divisions of normal tissue stem cells and carcinogens like radiation, chemicals, or chronic inflammation increase somatic mutations (2,3). Some cancers arise from either germline or sporadic mutations. For example, 40% of retinoblastomas are heritable, often manifesting with bilateral and early onset disease (4), in opposition to the de novo mutation.

Companies continue to develop newer, faster, and more cost-effective next-generation DNA and RNA sequencing methods. Current sequencing technologies offer options for nontargeted characterization of the whole genome, methylation profiling (5) (epigenomics), and transcriptomics (messenger RNA) (Fig 1). To reduce costs and analysis, oncologists may sequence only a predefined panel of oncogenes, transcripts, or antibodies tailored for a specific malignancy (ie, oncopanel DNA or RNA sequencing, proteome microarrays) (Fig 1). Investigators also may sequence only known protein-coding segments of DNA, called exons (exome sequencing) (Fig 1). Omics tests are performed on primary tumors, metastases, or circulating tumor cells. Sequencing cell-free circulating DNA provides a “liquid biopsy“ for mutations in oncogenes, response to therapy, and disease recurrence (6).

As oncology shifts to omics-informed precision medicine, radiologists must understand how certain cancer genotypes manifest at imaging. Precision medicine also propels radiology reports to move toward more quantitative, reproducible metrics. Numerous standardized reporting systems, such as the Breast Imaging Reporting and Data System, or BI-RADS, and response criteria measures, including the Response Evaluation Criteria in Solid Tumors version 1.1, or RECIST 1.1, provide more quantitative assessments of disease risk and response to therapies. Ongoing advances in pixel-level analysis and artificial intelligence also offer new approaches to extracting quantitative tumor imaging features from conventional imaging studies (7).

In this review article, we demonstrate how a radiologist can advance cancer clinical care and research in the era of omics-informed medicine. We present the most common genetic and epigenetic mutations and pathways by organ system in a case-based manner. A list of genes mentioned in this review but not discussed in detail can be found in Table S1. We also present key imaging findings and implications in oncologic care.

New Cancer Classification and Review of Common Cancer Genomics and Epigenetics Alterations, Clinical Implications, and Imaging Manifestations by Organ

Current oncology practice is progressively moving toward genomics-informed management. Cancers from different sites of origin but harboring the same genetic or epigenetic alterations may be treated using similar therapy (8). For example, lapatinib, an inhibitor of human epidermal growth factor 2 (HER2) that was initially approved by the U.S. Food and Drug Administration for breast cancer in 2007, also shows efficacy in clinical trials for gastric cancer and colorectal cancers with HER2 mutations (9,10) (Fig 2).

Figure 2:

HER2 mutation in colon cancer metastasis in liver. (A–D) Axial intravenous contrast–enhanced CT images in the portal venous phase in a 70-year-old man with metastatic colon cancer show multiple liver metastases (arrow in A, B) that progressed under multiple treatment regimens. Tumor molecular profiling showed an HER2 mutation, and the patient was offered treatment with trastuzumab and lapatinib (selective HER2 inhibitors). On follow-up scans, there was partial response to treatment, with substantial size reduction and necrosis of the liver metastases (arrow in C, D).

Germline TP53 mutations in Li-Fraumeni syndrome predispose to early-onset cancers in diverse tissues, most commonly breast cancer, soft-tissue sarcoma, brain tumors, adrenocortical carcinoma, and bone sarcomas. Early identification and testing of these individuals is essential for adequate screening (11). Current screening protocols indicate that carriers should undergo abdominal US every 3–4 months, annual whole-body and brain MRI, and annual breast MRI for female carriers. Prophylactic mastectomy has been advocated in some instances. Patients require tailored cancer treatment, as radiation therapy and conventional chemotherapy increase the risk of subsequent malignancies (12). TP53 mutations occur very commonly in many other types of cancers. These mutations inform prognosis, such as in hepatocellular cancer, endometrial cancer, and bladder cancer (13,14). We review the most common cancer mutations, mechanisms of oncogenesis, and implications of these concepts on imaging by organ system.

Brain Cancer

Glioma is the most common type of primary brain tumor. Historically, gliomas were classified by histopathologic features into astrocytomas, oligodendrogliomas, oligodendroastrocytomas, ependymomas, and gangliogliomas. Gliomas can be broadly divided into low-grade gliomas and high-grade gliomas by histopathologic features. Treatment of low-grade gliomas entails surgical resection and adjuvant radiation and/or chemotherapy, with a life expectancy of 6–13 years (15). High-grade gliomas include anaplastic astrocytomas (grade III) and glioblastomas (grade IV). Treatment of glioblastomas remains challenging, with surgical resection followed by radiochemotherapy producing a median overall survival of only 14 months (16).

The Cancer Genome Atlas Research Network and the 2016 World Health Organization classification (17) revolutionized the diagnosis and classification of gliomas by using genomic, epigenomic, and molecular profiles. The Cancer Genome Atlas provides open access clinical, genomic, epigenomic, and molecular data from patients with glioblastomas, with corresponding imaging data in the Cancer Imaging Archive (18). These data supported several exploratory radiogenomics studies. The Visually Accessible Rembrandt Images (or, VASARI) project analyzed 24 semantic imaging observations (https://www.cancerimagingarchive.net/) at MRI against clinicogenomic features. This study produced radiogenomic signatures of key molecular markers of gliomas, including IDH1/2 mutation, 1p/19q codeletion, O⁶-methylguanine DNA methyltransferase (MGMT) methylation, amplification and mutation of the EGFR gene, and histone H3 K27M mutation (19,20) (Table 1), described as follows:

Table 1:

Summary of Most Clinically and Imaging-Relevant Genetic and Epigenetic Alterations in Glioblastoma

1. Isocitrate dehydrogenases (IDHs) regulate cellular metabolism and epigenetic modifications. Both wild-type mitochondrial (IDH2) and cytosolic (IDH1) IDH forms convert isocitrate into α-ketoglutarate. Mutant IDHs generate an oncometabolite, 2-hydroxyglutarate. Given structural similarity to α-ketoglutarate, 2-hydroxyglutarate competes with and inhibits pathways such as DNA methylation, leading to aberrant gene expression and oncogenesis (21). All patients with glioblastoma undergo genetic testing for IDH. Patients with glioblastomas with mutant IDH show better survival than do those with wild-type IDH1/2 tumors (22).

2. IDH-mutant glioblastoma localizes predominantly to the frontal lobe or subventricular region of the lateral ventricles, with rare invasion of the eloquent areas of brain (23). It generally has an indistinct tumor border, cortical involvement, less peritumoral edema, and limited or no enhancement (Fig 3) (24–26). Mean apparent diffusion coefficient (ADC) values are higher in IDH-mutant glioblastoma compared with IDH wild-type glioblastoma (27), with values of 0.9 x 10⁻³ mm²/sec and higher more likely representing IDH-mutant glioblastoma (26). IDH-mutant glioblastomas accumulate high amounts of 2-hydroxyglutarate and low choline to creatinine ratios at MR spectroscopy (28). IDH wild-type glioblastomas occur preferentially at the subcortical white matter of the subventricular zones of both hemispheres, mostly in the anterior horns. These glioblastomas exhibit marked peripheral enhancement, central necrosis, and substantial surrounding vasogenic edema (Fig 3).

3. 1p/19q codeletion impacts migration, apoptosis, and cell cycle regulation. Once identified clinically using fluorescence in situ hybridization (29), it indicates low-grade gliomas, supporting the diagnosis of an oligodendroglial tumor (30). 1p/19q codeletion is associated with improved patient survival and treatment response. If associated with IDH mutations, the survival rate is even higher (median survival of 8 years) when compared with patients with IDH mutations without codeletion (median survival of 6.3 years) or patients with IDH wild-type tumors (median survival of 1.7 years) (31). 1p/19q codeletion is often found together with IDH1/2 mutation, therefore sharing similar imaging features.

4. MGMT is a DNA repair enzyme. Methylation of the promoter region of the gene decreases levels of this enzyme, resulting in less efficient DNA repair following cytotoxic chemotherapy. MGMT methylation frequently occurs with IDH mutations and implicates a better prognosis. MGMT methylation strongly predicts a positive response to temozolomide in new and recurrent high-grade gliomas (32). Pseudoprogression following treatment occurs more commonly in glioblastomas with both MGMT and IDH mutations (33). Imaging phenotypes of MGMT-mutant glioblastomas include a temporal or frontal lobe distribution (34), higher permeability parameters (transfer constant) at dynamic contrast-enhanced MRI, and higher minimum ADC values (35).

5. Amplification and mutation of EGFR occurs more commonly in primary than in secondary glioblastoma, inducing rapid cellular proliferation and resultant poor prognosis. EGFR-mutant tumors most frequently localize to the left temporal lobe (34), demonstrate complete enhancement or incomplete ring enhancement (36), and show increased perfusion at MRI (37).

6. Histone H3 K27M mutation, a newly defined entity in the World Health Organization group of grade IV diffuse gliomas, correlates with a poor prognosis. K27M mutation occurs on the histone gene encoding H3.3 and causes epigenetic changes by altering chromatin accessibility. This malignancy can mimic gliomatosis cerebri with additional midline structure involvement. Imaging features include a central or midline location and predominant involvement of the brainstem, thalamus, and spine. K27M-mutant tumors show partial to no, diffuse, or irregular peripheral enhancement. These tumors commonly are solid with mildly restricted diffusion (38).

Figure 3:

IDH-mutant and IDH wild-type gliomas. Axial (A, E) T2-weighted, (B) diffusion-weighted, (C) apparent diffusion coefficient map, and (D, F) T1-weighted MR images with extracellular gadolinium. (A–D) Images in a 65-year-old male patient diagnosed with IDH-mutant glioma demonstrate a typical imaging phenotype. The patient presented with a large, predominantly frontal mass, with indistinct borders, minimal peritumoral edema, and involvement of the cortex (arrow in A). At diffusion-weighted imaging, the mass showed high signal intensity (arrow in B), with high apparent diffusion coefficient values (arrow in C). The contrast-enhanced image demonstrates a poorly enhancing mass (arrow in D). (E, F) Images in a 68-year-old male patient with IDH wild-type glioma show a large left temporal central necrotic mass with marked peripheral enhancement (arrow in F) and surrounding vasogenic edema (arrow in E). (G) Isocitrate dehydrogenase (IDH) function and oncogenic pathway. IDHs play central roles in cellular metabolism and epigenetic regulation. Wild-type mitochondrial (IDH2) and cytosolic (IDH1) IDH forms convert isocitrate into α-ketoglutarate (αKG). When mutated, IDHs generate an oncometabolite, 2-hydroxyglutarate (2HG), which leads to epigenetic changes with aberrant histone and DNA methylation, causing oncogenesis.

Lung Cancer

Genomic characterization of lung cancer classifies tumors according to distinct mutations and matched options for targeted, rather than conventional, cytotoxic chemotherapies. According to current guidelines, patients with advanced or metastatic non–small cell lung cancer (NSCLC) should undergo genetic testing for driver mutations, especially for adenocarcinomas by histologic analysis or if patients have a light or no smoking history. Testing includes EGFR, ALK, KRAS, ROS1, RET, MET, BRAF, and HER2 because activating mutations in these genes have potential targeted therapies (39). National Comprehensive Cancer Network guidelines on NSCLC recommend that all patients with metastatic disease undergo NTRK1/2/3 gene fusion testing, which may indicate a response to tropomyosin receptor kinase inhibitors (larotrectinib or entrectinib) (40). Unfortunately, tumor genomic sequencing is not standard of care worldwide. Furthermore, many patients with advanced disease and compromised status are not suitable for invasive diagnostic procedures. By identifying surrogate imaging markers, radiogenomics can complement genetic testing, predict certain driver mutations, and help direct targeted therapy (41).

NSCLC, the most common type of lung cancer, commonly shows mutations in TP53 (50%), KRAS (30%), and EGFR (15%). ALK, MET, and HER2 genetic alterations are less frequent (<10%) but are important for radiologists to know about, given the availability of targeted therapies and distinct imaging features (Table 2). Interestingly, mutations in EGFR, KRAS, and ALK are typically mutually exclusive in patients with NSCLC (42).

Table 2:

Summary of Most Clinically and Imaging-Relevant Genetic Alterations in Non–Small Cell Lung Cancer

EGFR and HER2 are two of the most frequently dysregulated oncogenes in solid cancers (43). Epidermal growth factor receptor (EGFR) and HER2 are members of the ErbB family of receptor tyrosine kinases that activate downstream signaling cascades driving proliferation, survival, and differentiation (44). Kinase-activating mutations in EGFR occur primarily in cancers like NSCLC and glioblastoma or secondary to acquired drug resistance. EGFR mutations occur in approximately 15% of NSCLC cases overall and in 50% of females with adenocarcinoma who do not smoke (41); it is more common in women than in men and has a higher incidence in Asian populations (45). Targeted molecular therapy involves EGFR tyrosine kinase inhibitors like erlotinib (46). Imaging features associated with EGFR mutations include the following: pure or mixed ground-glass opacity (odds ratio, 1.86) (47), with higher volume ground-glass opacity in patients with exon 21 missense EGFR mutation compared with exon 19 mutation or EGFR wild-type tumors (48); air bronchograms (odds ratio, 1.6); pleural retraction (odds ratio, 1.99); small size; and absence of fibrosis (49). For detection of an EGFR mutation, the previous studies found an odds ratio of 1.58 for early disease stage. Furthermore, emerging PET imaging probes are being developed to detect total EGFR and even activate mutations of this receptor, with initial clinical translation (50,51).

ALK mutations occur in approximately 4% of NSCLC cases, more frequently occurring in young patients who do not smoke (41). Targeted treatment involves ALK tyrosine kinase inhibitors like crizotinib and alectinib (52). Radiologic features associated with ALK mutations in NSCLC include central solid tumors with pleural effusions (49,53), lymphangitic spread, and advanced nodal, pleural, or pericardial metastases (54) (Fig 4). Treating patients with ALK-mutant NSCLC using the targeted therapy, crizotinib, commonly (45%) causes new or increasing simple or complex renal cysts, which often regress after cessation of treatment (55) (Fig 4). Awareness of these treatment-related changes will allow for a more confident and appropriate diagnosis.

Figure 4:

ALK mutation in non–small cell lung cancer. Axial images from (A, B) soft-tissue and lung window CT, (C, D) fusion PET/CT, and (E, F) postintravenous contrast CT of right kidney. Patient is a 53-year-old female who does not smoke, with non–small cell lung cancer with an ALK mutation who presented with (A–D) a central solid fluorodeoxyglucose-avid tumor (blue arrow) accompanied by pleural effusion (green arrow), metastatic mediastinal and hilar lymphadenopathy (yellow arrows), pulmonary lymphangitic carcinomatosis (red arrow in B) and vertebral bone metastasis (white arrow in D). Patients treated with crizotinib (ALK and ROS1 inhibitor) often develop new or increased renal cysts (orange arrow in E, F) with simple or complicated features as part of their treatment. These cysts often regress after stopping crizotinib.

Exon 14 mutations in MET, a receptor tyrosine kinase, occur in approximately 3%–4% of patients with NSCLC (56). U.S. Food and Drug Administration–approved molecular therapies against MET-mutated NSCLC include cabozantinib and crizotinib (57,58). Imaging features associated with MET mutations in NSCLC include peripheral masses typically located in the upper lobes, cavitation, and cystic changes. Mutations in MET increase metastases to the lungs, brain, and adrenal glands (59) (Fig 5).

Figure 5:

MET mutation in non–small cell lung cancer. (A) Axial soft-tissue window CT, (B) axial fusion PET/CT, and (C) coronal soft-tissue window CT images in a 35-year-old nonsmoking male patient with non–small cell lung cancer with a MET exon 14 mutation who presented with an aggressive peripheral fluorodeoxyglucose-avid tumor (green arrow) located in the upper zone of the left lower lobe, with a heterogenic appearance including cystic changes, septations, and calcifications. The tumor invaded the chest wall, compromising the 6th left rib and intercostal muscles. A distant brain metastasis was found during staging MRI (not shown).

HER2 mutations occur in 1%–3% of patients with NSCLC, predominantly in adenocarcinomas arising in people who do not smoke and in females (60). Patients may respond to trastuzumab-based regimens, which target the HER2 receptor (61). A pattern of disseminated lung nodules has been observed frequently with HER2 mutations in lung cancer (41).

Breast Cancer

Genetic mutations driving breast cancer can be somatic (such as TP53 and RB) or inherited germline mutations (62). Germline mutations, most commonly in BRCA1 or BRCA2, account for up to 5%–10% of breast cancers (63–66). BRCA 1 and 2 genes maintain genomic stability by promoting efficient and precise repair of DNA double-strand breaks. BRCA1 is one of the most commonly mutated tumor suppressor genes, accounting for 5% of unselected breast cancers. Patients with a BRCA mutation have up to an 80% lifetime risk of developing breast cancer and a 50% lifetime risk of ovarian cancer (67). BRCA mutations have also been found in several other cancers, including pancreatic, prostate, and peritoneal malignancies (41). Identifying BRCA mutations in breast cancer helps to stratify the risk for breast and other cancers (68) and manage patients by either intense screening or prophylactic mastectomy (69). BRCA mutations confer susceptibility to targeted therapies (poly [adenosine diphosphate-ribose] polymerase inhibitors, ie, olaparib), and patients carrying these mutations have a better response to immunotherapy (70,71).

Clinical guidelines categorize breast cancer into four major, biologically distinct subtypes on the basis of the immunohistochemical expression of hormone receptors (estrogen receptor [ER], progesterone receptor [PR], and HER2 status) and distinct molecular and genetic profiles (Table 3) (72).

Table 3:

Summary of Clinical Features, Implications, and Imaging Findings of Current Clinically Used Breast Cancer Molecular and Genomic Subtypes

As many as 75% of invasive breast carcinomas are associated with significantly higher ER expression (ER+), rendering it an important diagnostic and treatment biomarker. PR expression is modulated by ERs, and as such, provides information about the current functionality of the ER pathway. Higher levels of PR expression (PR+) are associated with improved overall survival and time to recurrence, while lower levels (PR−) are associated with a more aggressive course and worse prognosis (73). HER2 overexpression (HER2+) occurs in approximately 20% of breast cancers and has historically been associated with decreased survival (74).

Subtypes of breast cancer provide essential predictive information for therapy. Patients in the luminal A (ER+, PR+, HER2−) and luminal B (ER+, PR− or low, HER2± ) subgroups receive targeted antiestrogen therapies, such as selective ER modulators (eg, tamoxifen), selective ER degraders (eg, fulvestrant), or aromatase inhibitors (eg, anastrozole and letrozole) (75). Anti-HER2 therapies, such as trastuzumab and lapatinib, improve survival in patients with the HER2+ subtype (76,77). HER2 overexpression is not unique to breast cancer, and overexpression and/or amplifications or other mutations of HER2 occur in gastric, biliary, colorectal, lung, and bladder cancer (10). Checkpoint inhibitor immunotherapy and poly (adenosine diphosphate-ribose) polymerase inhibitors remain the only targeted therapies for patients with the basal or triple-negative subtype, and conventional chemotherapy remains a mainstay of care. Patients with basal or triple-negative breast cancer have increased risk for metastases and poorer prognosis relative to those with other subtypes (78,79).

Distinct imaging features have been associated with different molecular subtypes of breast cancer (Table 3). Luminal A tumors have irregular shapes, spiculated margins, clustered calcifications, and size less than 2 cm (80,81). At MRI, luminal A tumors usually lack peritumoral enhancement. The luminal B subtype shows irregular shape, noncircumscribed margins, and skin or nipple extension at mammography and lower ADC values at MRI. Imaging findings associated with the HER2 subtype include irregular or round shape, spiculated or noncircumscribed margins, higher ADC values, peritumoral edema, and persistent inflow type enhancement on delayed MR images. Branching or fine linear calcifications, microcalcifications, and increased breast density also occur commonly in tumors with HER2 overexpression (81,82). Basal-like or triple-negative tumors frequently have BRCA mutations. These tumors frequently present as round or oval well-circumscribed lesions with smooth margins and no calcifications (83). US often reveals a hypoechoic mass with microlobulated or angular margins with parallel orientation, with one-third of cases showing posterior acoustic enhancement (83). At MRI, this cancer subtype commonly shows high internal signal intensity on T2-weighted images (84), rim enhancement, higher ADC value than other subtypes, and peritumoral edema (81) (Fig 6). When irregular margins and intratumoral necrosis are present, these features correlate with a poor response to neoadjuvant chemotherapy (85). Peritumoral edema has also been related to reduced disease-free survival (86). Kinetic curves for contrast enhancement and tumor size in basal-like tumors have variable reporting, excluding these indicators from reliably differentiating the basal-like subtype from other subtypes (81). While imaging features aid in breast cancer molecular or genetic subtype differentiation, biopsy still serves as the reference standard.

Figure 6:

Non–triple-negative and triple-negative breast cancer with BRCA1 mutation. (A, G) Axial diffusion-weighted images (b = 850 sec/mm² and b = 800 sec/mm², respectively), (B, H) apparent diffusion coefficient maps, (C, I) maximum intensity projection images, (D, J) high-resolution dynamic contrast-enhanced subtraction images, (E, K) fat-saturated images, and (F, L) fat-saturated T2-weighted images. (A–F) Images in a 31-year-old female patient with BRCA1 overexpression and cT2N1 triple-negative left breast cancer who presented with an oval-shaped, rim-enhancing, centrally necrotic, and well-circumscribed mass (yellow arrow) with restricted diffusivity (apparent diffusion coefficient, 0.86 × 10⁻³ mm²/sec) in the upper outer quadrant of the left breast mid depth. This patient was treated with neoadjuvant chemotherapy. (G–L) Images in a 45-year-old female patient with a 3.2-cm biopsy-proven non–triple-negative estrogen receptor– and progesterone receptor–positive, human epidermal growth factor receptor 2–negative left breast cancer with a positive axillary lymph node. MR images show multicentric disease, with nonnecrotic enhancing foci (yellow arrow).

Gastrointestinal System

Hepatocellular carcinoma.— Hepatocellular carcinoma (HCC) represents about 90% of all primary liver cancers. Genetic and molecular pathogenesis of HCC is heterogeneous as outlined by data in The Cancer Genome Atlas. The two most common mutations in HCC are in the TP53 (31%) and the CTNNB1 (27%) genes (87). The genomic and histologic features classify HCC into two main groups: (a) proliferative, poorly differentiated tumors characterized by TP53 mutation and (b) nonproliferative, well-differentiated tumors characterized by CTNNB1 mutations (Table 4) (88).

Table 4:

Summary of Genetic and Histologic Subtypes of Hepatocellular Carcinoma and Key Imaging Features

The nonproliferative, well-differentiated HCC subtype has good prognosis, with low levels of α-fetoprotein and a nonaggressive phenotype. Imaging usually shows a single small lesion with smooth margins and a capsule (89), mild hyperintensity on T2-weighted images, and high ADC at diffusion-weighted imaging (90). Other imaging biomarkers include intratumoral fat and low fluorine 18 (¹⁸F)–labeled fluorodeoxyglucose (FDG) accumulation (89). The value of intratumoral fat remains controversial as it may also occur in poorly differentiated HCCs (91). Nonproliferative HCCs comprise the β-catenin (CTNNB1 gene) and the steatohepatitic subgroups. CTNNB1–mutant HCCs are associated with alcohol consumption (92) and are characterized by activation of the Wnt/β-catenin pathway. CTNNB1-mutant HCC shows hyperintensity at the hepatobiliary phase of gadoxetic acid–enhanced MRI because of overexpression of the OATP1B3 transporter (Fig 7) (93,94). Steatohepatitic HCCs demonstrate activation of the JAK/STAT3 pathway and lack of a CTNNB1 mutation. These HCCs correlate with metabolic syndromes and steatosis in the background liver. At imaging, steatohepatitic HCCs show abundant fat infiltration with loss of signal at T1-weighted MRI in the opposed phase, no necrosis, and hypoenhancement in the hepatobiliary phase. At dynamic CT, this subtype exhibits hyperenhancement in the arterial phase, isoenhancement in the portal phase, and washout in the late phase (95).

Figure 7:

CTNNB1-mutant and wild-type, biopsy-proven hepatocellular carcinoma (HCC). Images from axial contrast-enhanced liver MRI, using an intravenous hepatobiliary-specific contrast agent (gadoxetate disodium): (A, D) arterial, (B, E) portal venous, and (C, F) hepatobiliary phase. (A–C) Images in a 67-year-old man with a background of liver cirrhosis who presented with a focal liver lesion (arrow), with (A) nonrim arterial phase hyperenhancement, (B) washout in the portal venous phase, (C) and hyperintensity in the hepatobiliary phase. (D–F) Images in a 59-year-old man with a background of liver cirrhosis who presented with a focal liver lesion (arrow) with (D) nonrim arterial phase hyperenhancement, (E) washout in the portal venous phase, (F) and hypointensity in the hepatobiliary phase. Hyperintensity at hepatobiliary phase can distinguish CTNNB1-mutant HCC from CTNNB1 wild-type HCC.

The proliferative, poorly differentiated HCC subtype includes the macrotrabecular massive and scirrhous subgroups, which are associated with high serum α-fetoprotein levels and poor survival (96). Macrotrabecular massive HCC frequently occurs in the setting of chronic hepatitis B virus infection, TP53 gene mutation, VEGFA overexpression, and FGF19 gene amplifications (88,97), with patients often presenting with early relapse. At imaging, the macrotrabecular massive subgroup characteristically shows greater than 20% necrosis (98) and other aggressive features, such as large lesions, multifocality, nonsmooth tumor margins, arterial phase hyperenhancement, confluent multinodularity, infiltrative behavior, macro- and microvascular invasion, bile duct dilatation, and increased ¹⁸F-FDG uptake (89,99). Several MRI features have been reported to predict microvascular invasion, namely the presence of internal arteries, absence of a hypoattenuating halo, infiltrative behavior, decreased ADC, peritumoral enhancement, nonsmooth tumor margins, peritumoral hypointensity during the hepatobiliary phase of gadoxetic acid–enhanced MRI, and high ¹⁸F-FDG accumulation (100–104).

Scirrhous HCCs resemble intrahepatic cholangiocarcinomas, with significant desmoplasia, progenitor cell phenotype, activation of the transforming growth factor β pathway (88,97), and common expression of cytokeratin 19 (105). Imaging differentiation from intrahepatic cholangiocarcinoma is usually the main challenge. Ancillary findings can help make the diagnosis, including central hypointensity on T2-weighted images, intratumoral blood products, and the presence of multifocal intratumoral septa, an enhancing capsule, and a nontargetoid appearance at diffusion-weighted imaging for scirrhous HCC (106,107).

Further studies are still required to establish new and optimized imaging-based stratification of HCCs as new morphologic-molecular HCC subtypes emerge (13). Such classification will enhance patient care by tailoring treatment decisions and monitoring to individual patients.

Colorectal cancer.— Colorectal cancer (CRC) evolves through a stepwise progression from normal tissue epithelium to precursor lesions and then to carcinoma. Three main molecular mechanisms have been documented in CRC tumorigenesis: (a) chromosomal instability (85%), (b) microsatellite instability (MSI) (15%), and (c) CpG island methylator phenotype (17%) (108).

CRC tumors driven by chromosomal instability, defined by alterations in number and structure of chromosomes, are often associated with mutations in the APC, TP53, KRAS, and BRAF genes (109), or via Wnt and mitogen-activating protein kinase, or MAPK, pathway activation. These molecular alterations underlie the classic adenoma-carcinoma sequence. CRC tumors with mutant KRAS and BRAF tend to be more aggressive, with a poorer prognosis. In metastatic CRC, anti-EGFR monoclonal antibodies such as cetuximab and panitumumab demonstrate some clinical benefit. These antibodies block the extracellular domain of EGFR and subsequent downstream activation of the RAS/RAF/MEK/ERK pathway (110) (Fig 8E). However, if downstream signaling molecules are mutated, most commonly KRAS and BRAF mutations in 30%–50% and 10% of CRC cases, respectively, the anti-EGFR antibodies get bypassed, with resulting drug resistance (108,111,112). All patients with metastatic CRC who are candidates for anti-EGFR antibody therapy should be tested for mutant KRAS and BRAF as standard of care (108,111). Because these mutations increase membrane glucose transporter levels, several articles have reported the possible use of ¹⁸F-FDG PET/CT uptake as a surrogate marker for KRAS/BRAF mutations (113). However, a recent meta-analysis found low sensitivity and specificity of ¹⁸F-FDG PET/CT findings for this purpose (114). Other studies have shown the following additional characteristics of BRAF-mutated colonic tumors: right-sided predominance involving the cecum and up to the proximal two-thirds of the transverse colon, larger primary tumors, heterogeneous enhancement, lack of nonperitoneal metastasis, higher rates of serrated polyp formation, mucinous pathologic features, poor differentiation, and female predominance, particularly in individuals more than 70 years old (115) (Fig 8).

Figure 8:

BRAF-mutant and BRAF wild-type colon cancer. (A, C) Axial and (B, D, E) coronal portal venous phase intravenous contrast–enhanced CT images. (A–D) Images in an 82-year-old woman with a BRAF mutation who presented with a right-sided colonic mass with heterogeneous enhancement (yellow arrow). (E) Image in a 75-year-old man who presented with a BRAF wild-type left colonic mass with overall homogeneous enhancement (yellow arrow). (F) BRAF function and oncogenic pathway. BRAF is a serine/threonine protein kinase involved in the signaling cascade of the mitogen-activated protein (MAP) kinase/extracellular signal-regulated (ERK) pathway, which is crucial for cell proliferation, differentiation, apoptosis, and survival. Activation of this pathway can be triggered by substrate engagement with the receptors on the cell membrane. This leads to subsequent activation of the Raf kinase through direct interaction with Ras and phosphorylation of the immediate downstream mitogen-activated protein kinase (MEK), which in turn activates ERKs. There are various subtypes of BRAF mutations, with the most common at codon 600 (BRAF V600E), accounting for about 95% of BRAF-mutant cases. This pathway has been the subject for the development of many new targeted therapies.

MSI is defined by instability of short, tandemly repeated DNA sequences known as microsatellites. MSI tumors result from mutations in mismatch repair genes, including MLH1, MSH2, MSH6, and PMS2, or from epigenetic silencing of the MLH1 promoter (108). MSI causes the most common hereditary CRC syndrome, Lynch syndrome, which accounts for 3% of all CRCs (108). Other syndromes with increased risk of CRCs include familial adenomatous polyposis, MUTYH-associated polyposis, juvenile polyposis syndrome, Peutz-Jeghers syndrome, and Cowden syndrome (116). MSI can also occur in 15% of sporadic CRC tumors, often by methylation of the CpG-rich promoter sequence of MLH1. These sporadic MSI tumors tend to arise in the proximal colon and exhibit poor differentiation, mucinous histologic findings, and prominent lymphocytic infiltration. At imaging, more than 70% of MSI-driven CRC develop proximally to the splenic flexure and typically show high lymphocyte infiltration at histopathologic analysis (108,111,116). However, no specific radiogenomic signature has been reported to date. The excellent treatment response and prognosis of MSI-driven tumors treated with checkpoint inhibitor antibodies supports routine testing for MSI status in CRC (108,111).

The CpG island methylator phenotype features hypermethylation of promoter CpG island loci, resulting in silencing of tumor suppressor genes. It is considered a distinct molecular subtype of sporadic CRC and the major driver for the serrated neoplasia pathway in CRC tumorigenesis (108). It often correlates with female sex, older age, proximal and right location, MSI-high status, BRAF mutation, and mucinous histologic findings (117). At imaging, these tumors often have similar characteristics to BRAF tumors.

Genitourinary System

Pheochromocytoma.— Pheochromocytomas and paragangliomas are rare neuroendocrine cancers arising from either the adrenal medulla or extra-adrenal paraganglia, respectively. Paragangliomas occur along the sympathetic nervous system of the chest, abdomen, or pelvis or in the parasympathetic ganglia of the head and neck. Contrary to head and neck paragangliomas, both pheochromocytomas and sympathetic paragangliomas commonly release catecholamines. Approximately 10%–15% of paragangliomas are metastatic at the time of diagnosis, with metastases generally involving lymph nodes, bones, lungs, or liver (118).

Pheochromocytomas and paragangliomas harbor somatic mutations in 30% of cases and germline mutations in about 40% of cases (119). Germline mutations affecting subunits of the Krebs cycle enzyme, succinate dehydrogenase (SDH), are among the most prevalent, occurring in the SDHA, SDHB, SDHC, SDHD, or SDHAF2 genes. Dysfunction of SDH enzymes leads to impaired conversion of succinate to fumarate and consequent substantial accumulation of succinate. As with 2-hydroxyglutarate, succinate acts as an oncometabolite and inhibits several α-ketoglutarate dioxygenases, inducing pseudohypoxia and epigenetic changes (Fig 9E). These tumors often express a pseudohypoxic phenotype, reflected by avid contrast enhancement on imaging studies because of extensive angiogenesis caused by stabilization of hypoxia-inducible factor-1α, a key regulator of angiogenesis and metabolic reprogramming.

Figure 9:

SDHD mutation in paraganglioma. (A) Axial and (B) coronal intravenous contrast–enhanced CT images in the neck and upper abdomen in the (C) arterial and (D) portal venous phases in a 24-year-old man with an SDHD mutation who presented with an arterially enhancing left cervical paraganglioma (yellow arrow). One year after resection, the patient developed a left adrenal pheochromocytoma (red arrow) showing (C) arterial enhancement and (D) washout in the portal venous phase. (E) Succinate dehydrogenase (SDH) function in the pseudohypoxia pathway. Hypoxia inducible factor (HIF)–1 is a key player in reprogramming metabolism in the lack of oxygen. In normoxia, HIF-1 rapidly undergoes degradation by prolyl-hydroxylases and proteasomal degradation facilitated by von Hippel−Lindau (VHL) protein. However, under hypoxia or pseudohypoxia, HIF-1 gets stabilized and transcribed, leading to the activation of multiple downstream pathways such as angiogenesis. This can occur in cancer cells, either by accumulation of oncometabolites (ie, succinate or 2-hydroxyglutarate) or other oncogenic activations, such as VHL mutations. ARNT = aryl hydrocarbon receptor nuclear translocator, HRE = hypoxia response element, OH = hydroxylation, TCA = tricarboxylic acid.

Comprehensive molecular analysis of pheochromocytomas and paragangliomas has unveiled four genomic clusters, each with specific molecular signatures (120): (a) cluster 1, pseudohypoxia; (b) cluster 2, abnormal kinase signaling and activation of the mTOR pathway, including mutations in RET, NF1, TMEM127, MAX, MET, and HRAS; (c) cluster 3, activation of the Wnt-signaling pathway with MAML3 anomalies; and (d) cluster 4, cortical admixture.

The pseudohypoxic cluster can be divided into two subtypes: (a) tricarboxylic acid cycle–related mutations, accounting for 10%–15% of all pheochromocytomas and paragangliomas, and (b) VHL/EPAS1-related mutations, which occur in 15%–20% of cases. The tricarboxylic acid cycle–related subtype involves mutations in SDH gene subunits (SDHx) and other mitochondrial enzymes (120,121). SDHB and SDHD mutations are the most common. SDHB and SDHD mutations predominately occur in abdominal and head and neck paragangliomas, respectively (122). Among all SDH mutations, SDHD has the highest penetrance, with 75% of affected patients manifesting disease by 40 years of age, often with a multifocal distribution. Head and neck paragangliomas occur in more than 90% of patients with SDHD mutations, and multifocal disease occurs most commonly in patients with these mutations compared with those having mutations in other SDH subunits (Fig 9) (123).

Metabolic imaging alongside functional anatomic imaging (MRI and CT) have enhanced phenotypic understanding of the different genomic clusters of pheochromocytomas and paragangliomas. However, more research is needed in this field, with the few available studies focusing on the pseudohypoxic cluster (124–126). Based on their molecular profiles, most paragangliomas exhibit moderate to low ¹⁸F-FDG uptake, except cluster 1 paragangliomas (SDHx-related and VHL-related tumors) (124,126,127). This is particularly true for SDHB- and SDHD-related tumors (126). Additionally, a high expression of somatostatin receptor 2A in the tricarboxylic acid cycle–related cluster represents a good imaging target for diagnosis, staging, and surveillance of paragangliomas with gallium 68 DOTATATE PET/CT (128). More studies are required to better understand how this probe may help define the different genetic clusters.

Renal cell carcinoma.— Clear cell renal cell carcinoma (ccRCC) accounts for 70% of kidney cancers. Unlike many other cancers, a few well-recognized genetic mutations drive ccRCC. Most mutations occur on the short arm of chromosome 3, which contains the VHL, PBRM1, BAP1, and SETD2 genes; additional relevant mutations include KDM5C and MUC-4 (129):

1. Germline or somatic DNA mutation of VHL occur in familial or sporadic ccRCC, respectively. VHL encodes part of a protein complex that degrades hypoxia-inducible factor-1α, a key orchestrator of angiogenesis and metabolic reprogramming (Fig 9E). Inactivation of VHL often produces hypervascular and hypoxic tumors. Imaging features include tumors with well-defined margins, nodular enhancement, intra-tumoral vascularity, and solid appearance (rather than cystic) (130) (Fig 10). Germline mutations in VHL also predispose to adrenal pheochromocytomas, pancreatic cysts, and central nervous system hemangioblastomas (Fig 10). Despite being mutated in greater than 90% of ccRCC, a VHL mutation bears no prognostic value (131). Effects of mutant VHL have motivated development of antiangiogenic therapies like tyrosine kinase inhibitors targeting vascular endothelial growth factor receptor 2 and neutralizing VEGF antibodies.

2. PBRM1, an epigenetic regulator, encodes a nucleosome remodeling complex that limits DNA accessibility to RNA polymerase and transcription factors. PBRM1 mutations occur in 40%–50% of ccRCC cases, often with mutant VHL (132). Imaging shows tumors with solid structure. Mutant PBRM1 correlates with poor survival and higher stage disease. Mutations may predict response to immune checkpoint therapy (133).

3. BAP1, a tumor suppressor gene mutated in both hereditary and sporadic ccRCC, regulates cell breakdown and replication (132). Imaging features include ill-defined margins, calcifications, and renal vein invasion. Loss of BAP1 in multicystic ccRCC correlates with higher histopathologic grade and more aggressive disease.

4. KDM5C mutations occur in 6%–7% of ccRCC and are consistently absent in multicystic ccRCC (132). Imaging features include hypoenhancement relative to the renal cortex at nephrographic phase CT and renal vein invasion. Patients with this mutation often show higher total and visceral abdominal fat (134). KDMC5 mutations paradoxically correlate with unfavorable prognosis for localized cancer and prolonged survival in advanced metastatic disease (135).

5. SETD2 is a tumor suppressor gene mutated in approximately 10%–15% of ccRCC. Loss of this protein occurs in multicystic ccRCC, with tumors showing predominately exophytic growth (132). SETD2 mutations are associated with a poor prognosis in localized disease.

6. MUC4 is a transmembrane mucin. MUC4 mutations confer ccRCC tumors a preferential exophytic tumor growth pattern and a decreased overall survival.

Figure 10:

Clear cell renal cell carcinoma (ccRCC) with a germline VHL mutation. (A, B, D) Axial and (C) coronal soft-tissue CT images, as well as sagittal (E) T1-weighted image with extracellular gadolinium and (F) T2-weighted image of the cervicothoracic spine in a 35-year-old man with a germline VHL mutation (von Hippel−Lindau syndrome) who presented with bilateral arterially enhancing renal lesions, consistent with ccRCC (yellow arrows in A and C), and bilateral renal cysts. Other syndromic lesions were also identified in this patient, namely an avidly enhancing left adrenal pheochromocytoma (yellow arrow in B), multiple pancreatic cysts (yellow arrows in D), and several spinal hemangioblastomas (yellow arrows in E, F).

Bladder cancer.— Urothelial carcinoma, the most common urinary tract malignancy, involves the bladder in 90% of cases and the upper urinary tract in 10% of cases (136). Bladder cancer occurs as a superficial tumor (70%), which often has good prognosis, or as a muscle-invasive urothelial carcinoma (MIUC) (30%) (137) with poor survival due to metastases (138). Superficial tumors are routinely treated by transurethral resection, followed, in cases of high-grade histologic features and/or early subepithelial invasion, by bacillus Calmette–Guerin intravesical immunotherapy. Patients with invasive bladder cancer generally undergo radical cystectomy, often preceded by neoadjuvant chemotherapy (139). Given the incremental risk for metastases with certain genetic subtypes, patients will inevitably require more imaging tests and follow-up. Hence, it is important for a radiologist to understand the phenotypic translation of key genetic mutations.

The Cancer Genome Atlas Research Network has reported 58 significantly mutated genes in MIUC (140), with many of these genes involved in chromatin structure, cell cycle regulation, and kinase signaling. Molecular profiling defines five subtypes of MIUC: (a) luminal-papillary (35%), strongly associated with FGFR3 mutations; (b) luminal-infiltrated (19%), with best response rate to immunotherapy using checkpoint inhibitors; (c) luminal (6%), with high expression of urothelial differentiation markers and uroplakins; (d) basal-squamous (35%), with high expression of basal markers, CD44 stem cell-like markers, and TP53 mutations; and (e) neuronal (5%), featuring TP53 and RB1 mutations and elevated expression of neuronal and developmental genes (140). The best and worst 5-year overall survival rates are for the luminal-papillary and the neuronal subtypes, respectively.

The few radiogenomic studies have focused on specific low-risk (ARID1A, FGFR3, PIK3CA, STAG2, and TSC1) and high-risk (TP53, RB1, and KDM6A) mutations rather than molecular subtypes and clusters. Low-risk mutations demonstrate better outcomes, particularly in the setting of FGFR3 mutations (corresponding to the luminal-papillary subtype) (141), lower tumor grade and stage for STAG2 mutations (142), and reduced recurrence in non–muscle-invasive bladder cancer with a PIK3CA mutation (143). These mutations are associated with longer metastasis-free and greater overall survival in MIUC (144).

For high-risk mutations, TP53 mutations correlate with higher grade tumors, more invasive behavior, greater risk of recurrence, worse outcome, higher frequency of nodal and osseous metastases, and shorter metastasis-free and overall survival (Fig 11) (144–146). RB1 and KDM6A mutations correlate with an increased risk of metastatic disease and recurrence rates, as well as a greater likelihood of early progression and lower survival rates (Fig 12) (144,147).

Figure 11:

TP53 mutation in bladder cancer. (A, C) Axial soft-tissue and (B, D) lung window CT images and (E) sagittal T1-weighted image of the lumbar spine in a 75-year-old man diagnosed with resectable TP53-mutant bladder cancer (not shown) and no metastatic spread at initial diagnosis (A, B). After 3 months, the patient developed multiple lung and nodal metastases (arrows in C, D), followed by bone metastases 1 month later (arrow in E). This patient had a metastasis-free survival of 3 months and an overall survival of 6.6 months. TP53-mutant bladder cancers usually have an aggressive phenotype with early metastatic spread and poor survival compared with TP53 wild-type bladder cancers.

Figure 12:

RB1 mutation in bladder cancer. (A–C) Coronal oral contrast–enhanced images of the abdomen and pelvis without (A) and with intravenous contrast media (B, C) in a 61-year-old man diagnosed with a large, resectable RB1-mutant bladder mass (arrow in A) and no metastatic spread. After 2 months, the patient developed multiple nodal metastases (arrows in B), which further progressed at the subsequent CT scan performed 2 months later (arrows in C). This patient had a metastasis-free survival of 2 months and an overall survival of 4.7 months. Like bladder cancers with TP53 mutations, bladder cancers with RB1 mutations have aggressive phenotypes with early metastatic spread and poor survival.

Understanding the clinical significance of certain bladder cancer mutations can aid interpretation of imaging scans, particularly for patients with MIUC, given their higher risk of metastasis. When reporting a scan in a patient with TP53 or RB1 mutations, the radiologist should be more critical about equivocally enlarged or heterogeneous small lymph nodes, particularly in the pelvis, and search carefully for early signs of metastasis.

Conclusions

Omics technologies are now part of clinical care in oncology, underscoring the need for imaging experts to understand how these technologies integrate into risk assessment, diagnosis, and treatment. We highlighted key genetic and epigenetic regulators of cancer pathogenesis and treatment response with examples in different organ systems to familiarize radiologists with the basic concepts of cancer genomics. Applying these concepts will position radiologists to provide better patient-centered cancer care.