3835 Metabolomics CHAPTER 489

associated with the development or progression of disease. Efforts to

characterize these “metabolic signatures” have been focused primarily

on common, multifactorial diseases such as diabetes, cardiovascular

disease, and various cancers that are well represented in large prospective cohort studies. These studies have, for example, identified altered

levels of amino acids that are associated with a future diagnosis of

diabetes or pancreatic cancer.

Additional efforts have been made to assess the metabolome in

patient samples at the time of an acute presentation. Because altered

metabolite levels can be associated with a specific clinical diagnosis and/

or outcome, the idea is to identify a metabolite signature that facilitates

diagnosis or provides prognostic information. This approach has been

studied, for example, in the context of sepsis and septic shock, in which

blood lactate levels are assessed in combination with the use of clinical

tools such as the Acute Physiology and Chronic Health Evaluation

(APACHE II) or the Sequential Organ Failure Assessment Score (SOFA).

One key limitation in all of these studies is that researchers are primarily assessing correlations between blood plasma metabolite levels

and complex, multisystem diseases. It is often difficult to obtain a

biological understanding of the mechanisms driving these changes or,

even more simply, the primary tissue source(s) of these alterations from

human data alone, without further experimentation in model systems.

■ REFINING DIAGNOSIS AND PREDICTION OF

DRUG SUSCEPTIBILITY

In contrast to the above-described use of metabolomics-based

approaches in multifactorial diseases, the application of these

approaches in some specific contexts can yield an immediate diagnosis and suggest actionable therapeutic interventions. One specific

example in oncology involves an understanding of the pathogenesis of

oncogenic mutations in the metabolic enzyme isocitrate dehydrogenase

(IDH) isoforms 1 and 2. The normal function of these enzymes is to

interconvert isocitrate and α-ketoglutarate; however, cancer-specific

point mutations in these enzymes alter the enzymes’ function in a

manner conferring neomorphic activity that converts isocitrate into

2-hydroxyglutarate (2-HG). 2-HG is a metabolite that is typically

present only at very low levels in cells, but when mutant IDH protein

is present, 2-HG is produced and accumulates to high levels. Elevation

of 2-HG can promote changes that directly contribute to malignancy;

IDH mutations and 2-HG accumulation are found in several human

cancers, including specific clinical subsets of acute myeloid leukemia

and glioma. Given the unique and specific accumulation of 2-HG in

these mutant tumors, detection of this metabolite by LC-MS and NMRbased approaches has been studied both for diagnostic purposes and

as a means of assessing drug response. For example, researchers have

applied MRS-based approaches to assess the accumulation of 2-HG

in gliomas, as this finding can noninvasively identify patients with an

IDH-mutant subset of this cancer (Fig. 489-4). This diagnosis provides

prognostic information and, in the future, may help direct therapeutic options. In principle, metabolomics may identify other disease

biomarkers to aid with diagnosis or therapy assessment in a similar way.

■ PHARMACOMETABOLOMICS

The previous example positions metabolomics as a possible mechanism for achieving a more personalized approach to medicine. The

emerging field of pharmacometabolomics aims to take personalization

further by making this approach more widely applicable across drugs

and disease states. The general workflow is to take a sample population

and perform baseline metabolomics studies on blood from its members. The individuals then receive a given drug with subsequent bloodbased measurement of drug metabolites to gain pharmacokinetics (PK)

and pharmacodynamics (PD) information. These PK and PD data are

then correlated with baseline metabolomic profiling, with the goal of

generating a predictive model for individual PK and PD responses

based on a naïve patient’s metabolomic profile. Correlation with posttreatment metabolomics is also used to provide insight into how

expected responses to therapy could be monitored. Ideally, this

approach would allow clinicians to take a baseline set of measurements

and then—a priori—choose a specific dose of a specific drug to produce

the desired effect in that specific patient. Monitoring of the expected

metabolite changes in response to the drug could also be used to ensure

therapeutic efficacy. If successful, this method could limit both prolonged titration of medications and medication switching, dramatically

shortening and simplifying the current approach to medical therapy.

EMERGING TECHNOLOGIES

While efforts to improve the existing capabilities discussed above are

ongoing, innovations in instrumentation and computation are allowing

collection of metabolite information that previously was not possible.

■ MASS SPECTROMETRY IMAGING

Most clinical metabolomics relies on analysis of bulk material, but

in an individual patient there are areas of normal and diseased tissue, and understanding the differences in metabolism in these areas

requires both spatially sensitive resolution (imaging) and interrogation

(metabolomics). While MRS can perform some of these functions,

it is limited to macroscopic imaging (MRI) and relatively insensitive

metabolomics approaches (NMR). In contrast, MS-based approaches,

while more sensitive, by their nature rely on specimen destruction and

homogenization. The premise of mass spectrometry imaging (MSI)

is to overcome these limitations of MRS and mass spectrometry. MSI

combines histologic evaluation of tissue with MS-based approaches to

assess spatial differences in metabolism. MSI as a technique has been

FIGURE 489-4 In vivo 1

H spectra and analysis demonstrating 2HG detection in IDH mutant brain tumors. A–C In vivo spectra from normal brain (A) and tumors (B–C), are

shown. Components of 2HG, GABA, gluamate and glutamine are displayed. Measurement location indicated by yellow box (voxel). 2HG is seen only in mutant IDH brain

tumors, but not normal brain or wildtype tumors. Shown in brackets is the estimated metabolite concentration (mM) ± s.d. Cho, choline; Cr, creatine; Glu, glutamate; Gln,

glutamine; Gly, glycine; Lac, lactate; Lip, lipids. Scale bars, 1 cm. (Reproduced with permission from Choi et al: 2012).

A B C

3836 PART 20 Frontiers

most highly refined in the neurosciences and can provide subcellular

resolution. In general, thin slices of tissue are mounted on a slide, and

metabolomics is performed at defined points across the slide, yielding

spatial information on where in the tissue section metabolites are measured. One specific approach utilizes matrix-assisted laser desorption/

ionization (MALDI) coupled to MS. In MALDI, tissues are coated

with a special matrix and the MALDI laser scans point-by-point across

a tissue slice, ionizing the metabolites at each location for analysis by

a mass spectrometer. These data can then be referenced back to an

image of the original tissue slice (Fig. 489-5). This approach is being

tested for defining brain tumor margins in real time during resection

and thereby providing insight into boundaries between normal and

abnormal tissues.

■ IMPROVING UNTARGETED METABOLOMICS

Identifying unknown signals in an untargeted metabolomics analysis

remains one of the central challenges in the field. As discussed above,

NMR can definitively identify unknown signals but lags significantly

behind MS-based approaches in its sensitivity and therefore in the

number of signals it can detect. To leverage the sensitivity of MS-based

detection and overcome the challenge of metabolite identification,

researchers are applying computational techniques, using networkstyle analyses to streamline the process. The general approach is to

combine information from known biological perturbations (e.g.,

changes in experimental conditions or disease states), empirical mass

and structural information from MS analysis, and correlations with

known metabolites/pathways to place unknown metabolites within

existing metabolic networks.

SUMMARY

Metabolomics is part of a growing list of “-omics” techniques that have

emerged over recent decades. Despite its immaturity relative to genomics, the power of metabolomics comes from being directly connected to

phenotype and being very sensitive in measuring perturbations within

a system or a patient. While the clinical applications of metabolomics

are currently limited to specific indications, researchers are pushing

to expand these technologies toward broader use in medicine. If these

methods are to be used appropriately, clinicians need to be aware of

established biological and practical confounders. Similarly, a basic

knowledge of the technologies in use and their inherent limitations is

critical. With time and further technical development, metabolomics

could become a routine part of the clinical armamentarium for diagnosis, monitoring, and treatment of disease.

■ FURTHER READING

Bertholdo D et al: Brain proton magnetic resonance spectroscopy:

Introduction and overview. Neuroimaging Clin N Am 23:359, 2013.

Choi C et al: 2-Hydroxyglutarate detection by magnetic resonance

spectroscopy in IDH-mutated patients with gliomas. Nat Med 18:624,

2012.

Emwas AH et al: NMR spectroscopy for metabolomics research.

Metabolites 9:123, 2019.

Gencheva R et al: Clinical benefits of direct-to-definitive testing

for monitoring compliance in pain management. Pain Physician

21:E583, 2018.

Ionization

FIGURE 489-5 Mass spectrometry imaging provides spatial information around metabolites in tissues. Tissue is mounted onto a slide, and a laser or another method is used

to ionize metabolites in a discreet section of the tissue for detection by mass spectrometry. The process is repeated as the laser scans across the tissue, generating an

“image” based on the levels of a metabolite detected at each point in the tissue section.

Kantae V: Integration of pharmacometabolomics with pharmacokinetics and pharmacodynamics: Towards personalized drug therapy.

Metabolomics 13:9, 2017.

Langley RJ et al: An integrated clinico-metabolomic model improves

prediction of death in sepsis. Sci Transl Med 5:195ra95, 2013.

Mayers JR et al: Elevation of circulating branched-chain amino acids

is an early event in human pancreatic adenocarcinoma development.

Nat Med 20:1193, 2014.

Townsend MK et al: Reproducibility of metabolomic profiles among

men and women in 2 large cohort studies. Clin Chem 59:1657, 2013.

Wang TJ et al: Metabolite profiles and the risk of developing diabetes.

Nat Med 17:448, 2011.

While this chapter is focused on the modern use of liquid biopsies and

their applications in the context of malignancies, transplantation, and

noninvasive prenatal testing, it is worth considering that the history

of liquid biopsies dates back many centuries. Indeed, as early as the

fourth century b.c., Hippocrates is noted to have studied body fluids and humors to diagnose various maladies in his patients. By the

mid-seventeenth century, the analysis of bodily fluids (especially urine)

increasingly became a cornerstone of European medicine. However,

arguably, the most important use of bodily fluids as liquid biopsies

currently involves key clinical applications in oncology, solid organ

transplantation, and prenatal diagnosis (Fig. 490-1). Below, we discuss

these applications of liquid biopsies in the context of modern medical

management of patients with risk of cancer and tumor progression,

posttransplantation outcomes, and obstetric management.

CELL-FREE DNA IN ONCOLOGY

Major advances in the twenty-first century in cancer biology have

helped transform oncology, particularly with the advent of personalized cancer therapies, which are typically selected using molecular biomarkers to identify tumor-specific vulnerabilities. However,

because tumor heterogeneity remains a major obstacle to cancer monitoring and treatment, better approaches are needed to define clinically

actionable targets and to facilitate the individualized therapies that

embody precision medicine. While much can be learned from studying tumor biopsies directly, such as molecular profiling that informs

490 Circulating Nucleic Acids

as Liquid Biopsies and

Noninvasive Disease

Biomarkers

Ash A. Alizadeh, Kiran K. Khush,

Yair J. Blumenfeld

3837Circulating Nucleic Acids as Liquid Biopsies and Noninvasive Disease Biomarkers CHAPTER 490

Tumors are well known to release analytes into body

fluids, including the peripheral blood circulation, and the

diagnostic use of such analytes has often been referred to

as liquid biopsies (Fig. 490-2). For hematologic malignancies such as leukemias, the direct examination of

circulating hematopoietic elements is a cornerstone for

diagnosis, response, assessment, and disease monitoring.

Indeed, molecular tools allowed the specific detection of

lymphoma B cells in the blood circulation using in situ

hybridization in the mid-1980s, even before the advent of

polymerase chain reaction for more sensitive detection.

However, even when considering hematologic neoplasms

such as lymphomas, many do not circulate in large numbers in the blood. Indeed, the dominant anatomic distribution of most human cancers results in the absence of

detectable circulating tumor cells as evidence for minimal

residual disease (MRD) in the cellular compartment of

the blood.

Aside from extravasated tumor cells circulating in the

bloodstream as circulating tumor cells (CTCs), liquid

biopsy methods using peripheral blood also include

membrane-bound vesicles released by tumor cells called

exosomes, as well as non–membrane-bound derivatives such as cell-free DNA (cfDNA) and cell-free RNA

(cfRNA), which are released by apoptotic or necrotic

tumor cells. Although the most common source of liquid

biopsy is the peripheral blood, various bodily fluids have

been used for specific liquid biopsy applications to interrogate proximally associated anatomic compartments,

whether using urine, feces, pleural fluid, peritoneal fluid,

bronchoalveolar lavage fluid, saliva, or cerebrospinal

fluid. For example, urinary cfDNA has shown promise for

noninvasive detection of genitourinary tumors including bladder carcinomas, and colorectal cancer screening can be done using multianalyte

assays leveraging fecal DNA.

Among the most intensively studied tumor-derived biomarkers is

circulating tumor DNA in the blood plasma, as a subset of cfDNA.

cfDNA in the blood was first described in 1948 by Mandel and Metais,

• Acute Allograft

 Rejection

(dd-cfDNA)

• Microbiome/Virome

• Monitoring Host

Immunity

• Donor-specific

 Antibodies

Key Medical

Uses of cfDNA

Liquid Biopsies

Transplant Medicine

• Early Detection

• Noninvasive

Genotyping (ctDNA)

• Molecular Response

Evaluation

• Minimal Residual

 Disease (MRD)

Oncology

• Noninvasive

Prenatal Testing

(NIPT/fetal DNA)

• Screening/

 Diagnosis

• Fetal Aneuploidies (+21, +18, etc)

• Microdeletions & Duplications

• Single-Gene Disorders

Obstetrics

FIGURE 490-1 Key medical applications of liquid biopsies employing cell-free DNA (cfDNA),

with specific focus on oncology, transplantation, and obstetrics. ctDNA, circulating tumor DNA;

dd-cfDNA, donor-derived cfDNA; NIPT, noninvasive prenatal testing.

diagnosis and monitoring of pathologic response, these are invasive

clinical procedures. In addition, tumor biopsies may not yield enough

material for analysis and can be risky to the patient. Therefore, an

approach to analyze cancer less invasively, such as from a blood sample,

serves as a potentially attractive alternative in the clinical evaluation of

patients with cancer.

Tumor genotyping

MRD detection

Mechanisms of resistance

Circulating

DNA Profiling

Healthy Tissue

Malignant Tissue

FIGURE 490-2 Cell-free DNA (cfDNA) is released both from healthy and malignant tissues. Tumors shed circulating tumor DNA (ctDNA), a minor fraction of all circulating

cfDNA. Molecular profiling of cfDNA enables cancer detection and monitoring applications. MRD, minimal residual disease. (Source: Adapted from BJ Sworder et al:

Hematological Oncology, 39, 2021. https://doi.org/10.1002/hon.6_2879.)

3838 PART 20 Frontiers

Diagnosis

Profiling of tumor DNA

and/or ctDNA using:

Tissue

biopsies

Plasma

samples

Therapy

Subclone 1

Subclone 2

Subclone 3

Surveillance

ctDNA profiling

Tumor evolution

Progression

or disease

transformation

FIGURE 490-3 Framework for noninvasive identification of cancer risk groups.

Schematic illustrating the application of ctDNA profiling for the identification

of adverse risk at different disease milestones. A cancer patient is imagined

as experiencing these disease milestones over time, depicted as an arrow

progressing from left to right. During this temporal sequence, ctDNA can inform

risk at diagnosis, during therapy, in surveillance of disease, and at progression or

disease transformation. At diagnosis, profiling of tumor DNA obtained from either

tissue biopsies (indicated by a scalpel) or plasma (depicted as blood collection

tubes) allows for the identification of patients with high tumor burden and disease

subtypes defined by specific genomic aberrations. Assessment of ctDNA during and

after treatment facilitates the detection of both emerging resistance mutations and

minimal residual disease (MRD) before progression, with potential for noninvasive

prediction of relapse and treatment resistance. Tumor evolution in an anecdotal

patient is illustrated, showing tumor response and clonal evolution over the course

of the disease (detectable subclones at diagnosis are shown in blue/gray; an

emergent subclone after therapy is shown in red). The profiling of tumor DNA and

ctDNA at each milestone is shown by a double-stranded DNA molecule. (Adapted

from F Scherer et al: Distinct biological subtypes and patterns of genome evolution

in lymphoma revealed by circulating tumor DNA. Sci Transl Med 8.364:364ra155-

364ra155, 2016.)

but it was not for another four decades that tumor-derived DNA was

first observed in the plasma of cancer patients. Interestingly, cfDNA is

naturally fragmented with lengths that correlate with DNA wrapped

around individual nucleosomes. Since both healthy cells and cancer

cells release their DNA into the circulation, a major challenge is attaining enough sensitivity to identify and quantify tumor-derived circulating DNA (circulating tumor DNA [ctDNA]) against a potentially large

background of normal constitutional or germline DNA, the majority

of which derives from hematopoietic sources. Methods based on next-generation sequencing offer a way to achieve this. Over the past two

decades, ctDNA has been established as an important biomarker for

studying tumor biology and for detection of cancers. Here, we summarize key applications of ctDNA in the context of early cancer detection,

noninvasive tumor genotyping and classification, molecular response

during therapy, and MRD after definitive therapy (Fig. 490-3).

■ EARLY CANCER DETECTION

For several common cancer types including carcinomas of the lung,

colorectal tract, and breast, early cancer detection via screening can

improve outcomes for adults with established risk factors. However,

such screening can pose significant risks and associated expenses that

limit broad adoption. For example, while annual radiologic screening

by low-dose computed tomography is recommended to screen for lung

cancers in high-risk populations, implementation has been complicated by a high false discovery rate (~90%) and low compliance. Separately, distinct tumor types currently require unique corresponding

screening tests. Finally, most existing cancer screening modalities have

not fully integrated some of the key insights gained through molecular

profiling of cancer genomes. Therefore, there is an unmet need for

new methods for the early detection of cancers. Analysis of ctDNA is

a promising approach that could facilitate blood-based screening. In

this context, several efforts to leverage ctDNA for early detection are

emerging, including approaches focused on individual cancer types,

multianalyte testing combining ctDNA with other biomarkers including proteins, and broader multicancer early detection (MCED) assays

leveraging mutations, tissue-specific methylation, cfDNA fragmentation profiles, and other features.

Despite much excitement about their promise, most such cancer

detection efforts are relatively early in their development. As of 2021,

no screening studies have prospectively applied liquid biopsies as

interventions in randomized studies, with survival outcomes as clinical

endpoints. Indeed, no studies performed to date have demonstrated a

survival advantage of early detection using liquid biopsies in randomized trials, and this remains a key hurdle limiting widespread clinical

adoption of these tests. With these limitations in mind, future prospective studies with substantially longer-term follow-up are needed to

better establish the sensitivity and specificity of MCED tests for early

cancer detection, when considering the common cancer types most

amenable to localized interventions with curative intent. Separately,

because early cancer detection may not improve cancer-specific survival, studies will also be required to determine if these tests can yield

more true positives than false positives, especially because false-positive

results can generate a substantial number of expensive secondary invasive procedures.

■ NONINVASIVE TUMOR GENOTYPING

Molecular subgroups defined by activating mutations in key oncogenic

drivers are the basis for many targeted therapies in oncology across

diverse cancers. For example, targetable mutations in the EGFR gene

are the most common oncogenic driver in non-small-lung cancer

(NSCLC), and the presence of EGFR mutations in NSCLC tumor

biopsies is strongly correlated with a positive response to smallmolecule EGFR tyrosine kinase inhibitors (TKIs). While tissue-based

determination of tumor genotype is still considered by many to represent the gold standard for diagnostic purposes, liquid biopsies have

demonstrated very high positive predictive value for noninvasively

determining tumor genotype in the setting of advanced malignancies.

Indeed, several liquid biopsy assays have been approved by the U.S.

Food and Drug Administration (FDA) for companion diagnostic use

to identify mutations associated with response to targeted therapies in

diverse tumor types. Currently, several comprehensive genomic profiling liquid biopsy assays targeting >50 frequently mutated genes are

currently available as FDA-approved noninvasive tumor genotyping

tests for various solid tumors. For example, liquid biopsy testing using

the Guardant360 test (Guardant Health, Inc.) in patients with NSCLC

can be performed to identify EGFR mutations associated with response

to osimertinib or amivantamab or to identify KRAS mutations associated with sotorasib response. Similarly, liquid biopsy testing using

the FoundationOne Liquid CDx test (Foundation Medicine, Inc.) can

be performed to identify BRCA1/2 gene mutations associated with

rucaparib response in ovarian cancer, ALK rearrangements associated

with alectinib response in NSCLC, PIK3CA mutations associated with

alpelisib response in breast cancer, and BRCA1, BRCA2, and ATM

mutations associated with olaparib response in metastatic castration-resistant prostate cancer.

Of note, such assays have largely been developed for noninvasive

tumor genotyping in the setting of advanced malignancies and are not

well suited in the context of early-stage tumors. Indeed, a recent analysis

by the Sequencing Quality Control Phase 2 (SEQC2) Project Working

Group led by the FDA found that ctDNA detection was less reliable

below a circulating variant allele fraction of 0.5% for mutations of interest. Accordingly, the currently available assays designed for the primary

purpose of noninvasive genotyping in advanced disease are generally

not optimal for detection of MRD because the residual ctDNA allele

fractions after definitive treatment of localized solid cancers are typically much lower than 0.5%. Furthermore, even prior to treatment of

advanced cancers and especially in the context of low metastatic tumor

burden, the modest sensitivity of such liquid biopsy tests currently

requires that if specific lesions of interest are not initially detected in the

blood, a tumor biopsy is still needed to determine if the specific mutations and alterations are present, as this could inform therapy selection.

3839Circulating Nucleic Acids as Liquid Biopsies and Noninvasive Disease Biomarkers CHAPTER 490

■ BIOLOGIC CONSIDERATIONS

In addition to the somatic mutations found in cfDNA that originate

from tumor cells, somatic mutations arising in nontumor tissues can

pose as a source of biologic “background.” Such mutations can confound the use of ctDNA for cancer detection and monitoring. Hematopoietic stem cells can acquire mutations through a process called

age-related clonal hematopoiesis (CH), resulting in variants that can

be found in both cfDNA and circulating peripheral blood leukocytes.

When considering peripheral blood cells in patients who do not otherwise meet criteria for a leukemia diagnosis, identification of mutations

in ~20 genes canonically associated with hematologic neoplasms (and

exceeding 2% in allelic fraction) is termed clonal hematopoiesis of indeterminate potential (CHIP). CHIP has been identified as a risk factor

for cardiovascular disease and hematologic neoplasms and is also

relevant for liquid biopsies (Fig. 490-4). For example, because most

cfDNA derives from hematopoietic sources, CH represents a major

contributor to biologic mutational background for various applications

of liquid biopsies across cancers. Importantly, the prevalence of CH

variants increases with patient age, larger gene panels, and more sensitive testing, with prevalence approaching 100% in adults >60 years old.

Indeed, in liquid biopsy applications for noninvasively identifying

somatic alterations using cfDNA, CH represents the dominant biologic

source of false-positive findings, even if this can vary as a function of

the specific genes or lesions being considered. While CHIP disproportionately affects genes such as DNMT3A, TET2, ASXL1, and JAK2,

which are associated with myeloid cell fitness, population sequencing studies have shown that many more genes in the genome can

be affected by somatic mutations arising during CH in aging adults.

For example, independent liquid biopsy studies have identified that

15–41% of cfDNA mutations in the TP53 gene can be attributable to

CH because these same lesions were found in matched blood leukocytes but not in biopsied tumor tissues.

Accordingly, direct genotyping of peripheral blood leukocytes can

be very helpful to avoid such cfDNA mutations arising from CH,

which can potentially be false-positive results masquerading as ctDNA.

However, studies suggest that ~10% of mutations found in cfDNA of

otherwise healthy adults may not be found in matched leukocytes.

This suggests other noncirculating sources of such cfDNA mutations,

including CH arising in noncirculating hematopoietic precursors

in tissues such as the bone marrow or from clonal proliferations of

nonmalignant, nonhematopoietic cell types. Indeed, recurrent somatic

mutations in genes such as KRAS, MED12, BRAF, and others can

be present in diverse nonmalignant cell types. These mutations (and

others) can arise in a range of epithelial, endothelial, and stromal constituents of vascular malformations, endometrial leiomyomas, melanocytic nevi, and other nonmalignant proliferations. As detailed below,

incorporating sequencing of tumor tissue to initially identify somatic

mutations that are later monitored in plasma samples can help to guard

against these nonmalignant sources of biologic background.

■ PRETREATMENT TUMOR BURDEN

In patients with an established cancer diagnosis, liquid biopsies can be

useful as a noninvasive means of tumor genotyping using plasma, in a

manner that informs therapy selection in such scenarios as described

above. Aside from this use, quantitative assessments of tumor burden

in the plasma prior to therapy can also provide valuable information

via liquid biopsies. For example, pretreatment ctDNA levels have

been shown to have significant associations with established measures

of tumor burden and disease risk including stage, metabolic tumor

volume, and serum protein markers such as lactate dehydrogenase in

lymphomas, CA19-9 levels in pancreatic cancers, and carcinoembryonic antigen levels in colorectal cancers, among others. Importantly,

in many of these scenarios and in other cancers such as NSCLC where

noninvasive tumor biomarkers are not available, pretreatment ctDNA

levels have been shown to have strong prognostic value for treatment

failure and disease progression, providing an independently prognostic

measure of risk. Indeed, such pretreatment ctDNA levels can be used

to noninvasively measure tumor burden poorly captured by other

indices that can lead to biases in clinical trials, including the diagnosis

to treatment interval in lymphomas. Accordingly, pretreatment ctDNA

levels may be used to prevent selection bias in prospective clinical

trials. However, while such independent prognostic value of liquid

biopsies has been validated for several tumors (e.g., using CTC levels

in breast cancers), randomized trials demonstrating the clinical utility

of these measurements for predicting therapeutic benefit from specific

treatments have not yet been performed.

■ MONITORING RESPONSE TO TREATMENT

Beyond the application of liquid biopsies for early cancer detection and

noninvasive tumor genotyping, their use for monitoring therapeutic

responses in a quantitative and qualitative fashion deserves discussion.

In many cancer types, functional imaging has utility for monitoring

systemic response to treatment, when considering relative changes in

volumetric and/or metabolic tumor burden to assess complete versus

partial responses or stable disease. Similarly, in hematologic malignancies such as chronic myelogenous leukemia, the magnitude of response

to systemic therapy with TKIs can be monitored at defined milestones

while on continuous treatment, using defined thresholds for reductions

in BCR-ABL1 transcript levels in the blood. Liquid biopsies can be

similarly useful for measuring quantitative changes in ctDNA while

on therapy. Importantly, as with the use of functional imaging, it is

critical to consider several key factors when using quantitative changes

in ctDNA levels to monitor response. These include the specific tumor

histology and treatment setting (frontline vs relapsed disease), treatment type, the timing of interim ctDNA response assessment, the

type of ctDNA assay being used and associated sensitivity and specificity characteristics, and the thresholds for change in ctDNA levels

observed as informing early molecular response. For example, liquid

biopsy applications for measuring interim responses have been shown

to strongly predict survival outcomes in diffuse large B-cell lymphoma

and Hodgkin’s lymphoma in the frontline setting. Here, 100-fold

reductions (2-log) in ctDNA levels after one cycle of induction chemotherapy have been shown to reliably define an early molecular response

threshold, and 2.5-log reductions in ctDNA levels after two cycles can

be used to define a major molecular response, which are both strongly

associated with event-free and overall survival.

■ MINIMAL RESIDUAL DISEASE

As in the case of several hematologic malignancies, an emerging

body of evidence now demonstrates that liquid biopsies including

ctDNA can detect MRD following treatment of diverse solid tumor

types as a predictor for relapse risk. Indeed, detection of MRD using

ctDNA-based techniques has shown strikingly high positive predictive value for predicting relapse risk in many cancer types, including

carcinomas of the lung, colon, rectum, bladder, and breast, among

others. For example, when considering adjuvant immunotherapy following resection of localized bladder cancers, a retrospective analysis

of a randomized clinical trial strongly suggests that clinical utility for

the checkpoint inhibitor atezolizumab is likely limited only to patients

Nonmalignant

circulating leukocytes

harboring clonal

mutations, called

clonal hematopoiesis

of indeterminate

potential (CHIP), also

release mutant cfDNA

into the blood.

Healthy and cancer

tissues release

cell-free DNA

fragments (cfDNA)

into the blood as

cells turn over.

Tumor

Tissue

Healthy

Tissue

FIGURE 490-4 Contribution of clonal hematopoiesis of indeterminate potential

(CHIP) to cell-free DNA, as relevant for tumor genotyping and monitoring using

circulating tumor DNA. (Source: Adapted from J Boegeholz et al: Hematological

Oncology 39, 2021. https://doi.org/10.1002/hon.23_2879.)

3840 PART 20 Frontiers

with ctDNA MRD detectable in the blood plasma after surgery using

bespoke assays. Separately, the Centers for Medicare and Medicaid

Services recently finalized the first local coverage determination to

provide coverage for ctDNA MRD testing for monitoring colorectal

cancers after surgery using the Signatera MRD test from Natera, and a

broader draft local coverage determination is currently under consideration to enable coverage of ctDNA MRD across tumor types.

Based on these and other similar results, pivotal clinical trials are

now underway that select patients for adjuvant therapy integrating

liquid biopsies to detect MRD and to measure response to such adjuvant and consolidative treatments. Of note, despite the high positive

predictive value of ctDNA MRD for predicting relapse, the clinical

sensitivity of several current assays is generally modest, with a substantial fraction of relapses occurring in patients falsely classified as

MRD negative after initial maneuvers. As in the case of early detection

and noninvasive genotyping described above, for applications of liquid

biopsies to MRD, it is critically important to consider the specific

therapies administered, the biology of ctDNA release, the timing of

MRD measurements, the liquid biopsy assay performance characteristics, and the sources of both technical and biologic background. For

instance, the optimal use of liquid biopsies for their negative predictive

value (e.g., by withholding unnecessary adjuvant therapy in those

without evidence of MRD) is likely to require substantial additional

improvements in the analytical and clinical sensitivity of MRD detection for solid tumors. Nevertheless, although the broad clinical utility

of ctDNA MRD for treatment personalization has yet to be fully established, liquid biopsies hold substantial promise for guiding adjuvant

and consolidative therapies.

■ TECHNICAL CONSIDERATIONS

Despite generally representing a small fraction of nucleic acids that

circulate in the blood plasma, tumor-derived ctDNA molecules can be

identified through a range of techniques related to amplification and

detection. These methods broadly include assays that target tumorspecific mutations, structural variants, somatic copy number alterations, and epigenetic features and generally involve using polymerase

chain reaction (PCR) and high-throughput sequencing. Recently,

substantial improvements in the analytical sensitivity of liquid biopsies

have been achieved through a combination of refinements in these

molecular techniques and the associated computational methods for

analyzing the corresponding data.

Currently, liquid biopsy applications for early cancer detection,

noninvasive tumor genotyping, response monitoring, and MRD rely

mainly either on sequencing-based methods or on amplicon-based

techniques that do not require sequencing, such as digital droplet PCR

(ddPCR) or allele-specific PCR (AS-PCR). Key factors that distinguish

these methods include the cost and turnaround time for the assays

(which are generally in favor of amplicon-based methods), as well as

the scope of genomic aberrations simultaneously being evaluated and

the breadth and depth of the associated molecular profile (which are

generally in favor of sequencing-based methods). For MRD applications, the integration of multiple tumor-specific somatic alterations

into a single assay can allow multiplexed sequencing-based strategies to

achieve analytical sensitivities in the parts-per-million range, especially

in the context of bespoke assays leveraging tumor genotypes.

Among various sequencing-based techniques, liquid biopsy applications have variably profiled the whole human genome or targeted

subregions of the genome, depending on the specific applications

in oncology. More specifically, selective targeting of portions of the

human genome can be achieved either through enrichment by hybridization affinity capture or using locus-specific amplicons, whether

focused on the whole coding exome or smaller, more focused genomic

regions of interest from tens to hundreds of genes.

The biologic and technical sources of background can limit the

sensitivity and specificity of liquid biopsies. Apart from the various

strategies described to reduce these sources of error, many liquid

biopsy studies monitoring therapeutic response and ctDNA MRD

have leveraged tumor genotype–informed analyses to improve performance. Unlike noninvasive genotyping methods that rely entirely on

blood plasma, this tumor genotype–informed approach includes the

profiling of tumor tissue to identify mutations that are then tracked

in posttreatment blood plasma. In reducing the number of mutations

under consideration, this approach reduces the risk of false positives

due to technical and biologic background sources of error. Separately,

the tumor genotype–informed approach can be less demanding for

blood sample volumes. However, due to the very low circulating levels

of tumor-derived DNA at posttreatment milestones, tracking multiple tumor genotype–informed mutations and minimizing unwanted

effects of biologic and technical errors are critical for optimally capturing disease risk in diverse cancers using liquid biopsies to detect

ctDNA MRD.

CELL-FREE DNA IN TRANSPLANTATION

cfDNA testing offers very powerful tools for clinical monitoring of

organ transplant recipients. After transplantation, cfDNA analyses

have been used to assess for development of acute allograft rejection, to

study microbial diversity and infection, and to quantify host immunity.

This section describes current and potential future clinical applications

of cfDNA testing in the transplant arena.

■ NONINVASIVE DETECTION OF ACUTE

ALLOGRAFT REJECTION

In the setting of transplantation, cfDNA is derived from both the recipient tissues and the donated organ or cells. A transplant procedure is

essentially a “genome transplant,” and methods have been developed

to detect and quantify levels of donor-derived cfDNA (dd-cfDNA)

after transplant, with elevated levels reflecting graft injury due to acute

rejection and other forms of graft damage.

In 1998, Lo and colleagues first reported detection of DNA from the

organ donor in the plasma of transplant recipients. By performing PCR

amplification using Y chromosome–specific primers, they were able to

identify dd-cfDNA in the blood of female kidney and liver transplant

recipients. This early work provided proof-of-concept of this unique

approach but was limited to female recipients of male donor organs

(<25% of transplant procedures).

Subsequently, a universal, sex-independent strategy was developed

using whole genome shotgun sequencing to measure single nucleotide

polymorphism (SNP) differences between individuals to quantify the

donor signal. This approach is applicable to any organ donor and recipient combination, regardless of sex, by first genotyping the donor and

recipient to identify sequence differences that can then be used to identify donor cfDNA in the recipient’s blood after transplantation. Prospective studies of this approach have demonstrated that dd-cfDNA is

present at very high levels during the first few days after the transplant

procedure, reflecting cell death within the allograft due to ischemia and

reperfusion injury. Within 1–2 weeks after transplantation, however,

dd-cfDNA levels fall to a low baseline level and remain constant in the

absence of acute rejection.

In the setting of acute rejection, dd-cfDNA levels increase significantly in the transplant recipient’s circulation and correlate with severity of rejection. Initial studies in heart transplantation showed that, at

a threshold value of 0.25%, dd-cfDNA had an area under the receiver

operating characteristic curve (AUC) of 0.60 for distinguishing mild,

0.83 for distinguishing moderate-to-severe, and 0.95 for distinguishing

severe rejection events, each compared to the absence of rejection.

Notably, this assay can be used for surveillance of both acute cellular

and antibody-mediated rejection, as both processes result in graft

damage (Fig. 490-5). A subsequent multicenter study confirmed the

utility of dd-cfDNA monitoring for acute rejection, showing that, at

a threshold of 0.25%, dd-cfDNA had a 99% negative predictive value

for acute rejection and would have safely eliminated 81% of routine

surveillance biopsy procedures.

Notably, increasing dd-cfDNA levels were detected several weeks

to months prior to the rejection event, suggesting that dd-cfDNA is a

highly sensitive marker of graft injury and can enable earlier rejection

diagnosis. Early detection of graft injury may prompt augmentation of

immunosuppressive therapy to halt graft damage in its early stages and

to prevent a subsequent rejection event.