Molecular and functional imaging in cancer-targeted therapy: current applications and future directions

Abstract

Targeted anticancer drugs block cancer cell growth by interfering with specific signaling pathways vital to carcinogenesis and tumor growth rather than harming all rapidly dividing cells as in cytotoxic chemotherapy. The Response Evaluation Criteria in Solid Tumor (RECIST) system has been used to assess tumor response to therapy via changes in the size of target lesions as measured by calipers, conventional anatomically based imaging modalities such as computed tomography (CT), and magnetic resonance imaging (MRI), and other imaging methods. However, RECIST is sometimes inaccurate in assessing the efficacy of targeted therapy drugs because of the poor correlation between tumor size and treatment-induced tumor necrosis or shrinkage. This approach might also result in delayed identification of response when the therapy does confer a reduction in tumor size. Innovative molecular imaging techniques have rapidly gained importance in the dawning era of targeted therapy as they can visualize, characterize, and quantify biological processes at the cellular, subcellular, or even molecular level rather than at the anatomical level. This review summarizes different targeted cell signaling pathways, various molecular imaging techniques, and developed probes. Moreover, the application of molecular imaging for evaluating treatment response and related clinical outcome is also systematically outlined. In the future, more attention should be paid to promoting the clinical translation of molecular imaging in evaluating the sensitivity to targeted therapy with biocompatible probes. In particular, multimodal imaging technologies incorporating advanced artificial intelligence should be developed to comprehensively and accurately assess cancer-targeted therapy, in addition to RECIST-based methods.

Introduction

Cancer is the leading cause of mortality globally. It was estimated that approximately 2,370,000 and 4,820,000 new cancer cases and 640,000 and 3,210,000 cancer deaths would occur in the United States and China in 2022, respectively.¹ Surgery and radiotherapy (RT) are two primary treatment cornerstones of locoregional and nonmetastatic cancers, while chemotherapeutics can be used in all cancer stages. Current chemotherapeutics are often limited by undesirable side effects due to an inability to distinguish between tumorous and normal tissues. To mitigate these side effects, it is possible to develop targeted therapy using monoclonal antibodies or small-molecule inhibitors directed against specific signal transduction pathways for angiogenesis, proliferation, survival, and invasiveness, which are often dysregulated in tumor cells.² The development of targeted therapies is thus a valuable advance for cancer treatment.

To evaluate the efficacy of anticancer treatment, both the World Health Organization (WHO) response criteria³ and Response Evaluation Criteria in Solid Tumors (RECIST)⁴ utilize changes in tumor size as determined using imaging techniques such as CT, MRI, and positron emission tomography (PET). RECIST 1.1 is currently the gold standard for assessing treatment response in solid tumors in a clinical context, but it is insufficient for some targeted medications such as antiangiogenic agents and immunotherapy.⁵ For immunotherapy, a 5% rate of pseudo-progression, involving the tumor regressing after initial disease progression or the appearance of new lesions, was reported in non-small-cell lung cancer (NSCLC) patients treated with nivolumab.⁶ According to reports, this phenomenon was caused by an insufficient immune response or edema and necrosis related to immune-cell infiltration into tumor tissue.⁷ RECIST 1.1 relies solely on tumor size and does not consider changes in cellular events such as apoptosis, inhibition of proliferation, cell cycle arrest, tumor metabolism within the tumor microenvironment, and the density and number of intra-tumoral vessels. Thus, it remains unclear how best to evaluate the efficacy of targeted therapies and optimize the therapeutic strategy.

New functional and molecular imaging biomarkers are being developed to better evaluate targeted therapy’s effects. Molecular imaging combines biomedical imaging and molecular biology to visualize and quantify the spatiotemporal distribution of biological processes within living organisms in a noninvasive way for biochemical, biologic, diagnostic, and therapeutic applications.^8,9 Representative examples of molecular imaging techniques include radionuclide imaging (PET), single-photon emission computed tomography (SPECT), molecular magnetic resonance imaging (mMRI), magnetic resonance spectroscopy (MRS), optical imaging (optical bioluminescence, optical fluorescence), photoacoustic imaging, and multimodal imaging. Some modalities, such as radionuclide and optical imaging, require the injection of molecular probes to acquire the imaging signal. In contrast, mMRI and photoacoustic imaging can track drug effectiveness through endogenous molecules or exogenous molecular probes.

This review focuses on how to use novel imaging modalities to visualize the response to cancer-targeted therapies. Signaling pathway-based targeted therapies are concisely summarized. In addition, functional and molecular imaging modalities are discussed in detail about basic principles, imaging probes, and their application in targeted therapies against different molecular pathways. Lastly, future directions for molecular imaging in targeted therapies are prospectively reviewed.

Targeted therapies and anticancer drugs

Dysregulation of oncogenic signaling pathways plays a key role in the occurrence and progression of cancer. Substantial efforts have been made in treating cancer through “targeted” therapies that specifically disrupt pro-oncogenic signaling pathways. Specifically, there are two types of targeted therapy: small-molecule (enzyme)-based therapies,¹⁰ such as with tyrosine kinase inhibitors (TKIs) like sunitinib, and antibody-based targeted therapies,¹¹ such as with vascular endothelial growth factor (VEGF)-targeted antibodies including bevacizumab. Cancer immunotherapies, such as immune checkpoint inhibitors (ICIs), targeting the interaction between cancer and immune cells, broaden the scope of targetable tumors.¹² In this review, we introduce targeted therapies according to each signaling pathway.

VEGF/VEGFR signaling pathway

Angiogenesis is a crucial step in the successful growth, invasion, and metastasis of tumors, without which tumors could not grow beyond 1–2 mm in diameter.¹³ Multiple growth factors and their receptors are dysregulated in the complex process of tumor angiogenesis. The VEGF/VEGF receptor (VEGFR) signaling pathway is the pivotal inducer of angiogenesis, so antiangiogenic approaches have primarily focused on inhibiting this pathway.

A large number of drugs have been developed for targeting the VEGF/VEGFR signaling pathway¹⁴: (1) ligand binding agents that block the binding of VEGF ligands to receptors, such as bevacizumab,¹⁵ which binds to VEGF alone, and aflibercept,¹⁶ which binds to VEGF and placental growth factor (PlGF); (2) antibodies that block signaling through VEGFR, such as ramucirumab¹⁷ that targets VEGFR2; and (3) small‐molecule TKIs that block the kinase activity of VEGFR,¹⁸ such as lenvatinib, sorafenib, sunitinib, pazopanib, and regorafenib. TKIs can often inhibit the activity of other receptor tyrosine kinases, such as platelet-derived growth factor receptors (PDGFRs), fibroblast growth factor receptors (FGFRs), and epidermal growth factor receptors (EGFRs).

Besides, integrin α_Vβ3 is highly expressed in tumors and neovascular endothelial cells¹⁹ and is recognized as an ideal marker for distinguishing between cancerous and normal states. RGD (Arg-Gly-Asp) is a polypeptide including cyclic and linear categories screened by phage peptide display technology, which can specifically bind to the extracellular region of the α chain of α_Vβ3. Recently, many studies have reported that RGD peptides can carry drugs to tumor sites precisely and described that radionuclide-labeled RGD peptides have many uses for tumor imaging and therapy.²⁰

EGFR signaling pathway

The epidermal growth factor receptor (EGFR) is a member of the ERBB receptor tyrosine kinase family consisting of EGFR/human epidermal growth factor receptor 1 (HER1), HER2, HER3, and HER4.²¹ EGF ligand binding to EGFR results in EGFR dimerization and activation of intracellular kinase activity. Autophosphorylation of tyrosine residues of EGFR activates two main downstream signaling pathways,²² PI3K/AKT and RAS/RAF/MEK/ERK, which regulates cell proliferation, differentiation, and survival.²³

Inhibitors of the EGFR signaling pathway include small-molecule TKIs and anti-EGFR monoclonal antibodies (MoAbs). TKIs act on the ATP binding pocket of EGFR, inhibit EGFR autophosphorylation and antagonize tyrosine kinase signal transduction.²⁴ First-generation (erlotinib, gefitinib, and icotinib) and second-generation (afatinib and dacomitinib) EGFR TKIs have been approved for the treatment of advanced NSCLC patients harboring EGFR-activating mutations.²⁵ Unfortunately, resistance is inevitably acquired in most patients, at a median of 10–14 months after treatment.²⁶ The most common reason for acquired resistance is the T790M mutation in exon 20 of EGFR. Third-generation mutation-selective EGFR TKIs,²⁷ such as rociletinib,²⁸ osimertinib,²⁹ and almonertinib,³⁰ have been developed to overcome this resistance mutation.

The related anti-EGFR antibodies target the EGFR extracellular domain and competitively bind to receptors, which impedes dimer formation, thereby inhibiting intracellular signal transduction.³¹ The antibodies against EGFR include nimotuzumab,³² panitumumab,³³ matuzumab,³⁴ and cetuximab.³⁵ Antibodies are specific to EGFR, while TKIs can cross-link with other EGFR tyrosine kinases (HER2 and HER4).

HER2 signaling pathway

Unlike other members of the EGFR family, HER2 has no identified ligand. The HER2 signaling pathway is mediated by its heterodimeric form, created by HER2’s binding to other members of the EGFR family. This heterodimer can transactivate HER2 tyrosine kinase activity, further activating its downstream signaling pathways like PI3K/Akt/mTOR, MAPK, phospholipase C, and protein kinase C.³⁶

Numerous HER2 inhibitors had been developed in the last few decades, primarily consisting of monoclonal antibodies (MoAbs), small-molecule TKIs, and antibody–drug conjugates (ADCs).³⁷ Trastuzumab, pertuzumab, ertumaxomab, and margetuximab are common monoclonal antibodies targeting HER2. Various HER1/HER2 TKIs, pan-HER TKIs, and dual HER2/VEGF TKIs are in different stages of clinical trials or clinical practice.³⁸ Currently, the most commonly used TKI drugs include lapatinib, neratinib, pyrotinib, and tucatinib. HER2 ADCs direct drug delivery to HER2-expressing cancer cells while limiting exposure to normal tissue.³⁹ The currently available ADCs for HER2-positive cancer include trastuzumab emtansine (T-DM1) and trastuzumab deruxtecan (T-DXd).

In recent years, several novel emerging groups of anti-HER2 agents, including antibody-based fragments (Fabs), diabodies, minibodies, nanobodies (Nbs), and affibodies, have been explored for HER2-positive breast cancer imaging and targeted radionuclide therapy.⁴⁰

CDK4/6 signaling pathway

The binding of cyclin D1 to CDKs (CDK4 and CDK6) drives G₁ to S phase transition and disease progression in tumors.⁴¹ CDK4/6 inhibitors prevent RB1 phosphorylation and E2F transcription and thereby induce G₁ cell cycle arrest and block cancer progression. There are three CDK4/6 inhibitors, palbociclib, ribociclib, and abemaciclib, that are approved for treating advanced HR-positive (HR+), HER2-negative (HER2−) breast cancer, either combined with an aromatase inhibitor (AI) as a first-line treatment option or combined with fulvestrant as a second-line treatment option. Among these, abemaciclib is the first US Food and Drug Administration (FDA)-approved CDK4/6 inhibitor for adjuvant therapy in HR⁺ HER2⁻ early-stage breast cancer.⁴²

PI3K/Akt/mTOR pathway

The phosphatidylinositol-3 kinase (PI3K)/Akt/mammalian target of rapamycin (mTOR) signaling pathway is frequently activated in response to various extracellular stimuli, such as growth factors, hormones, and cytokines.^43,44,45 PI3K is mainly triggered by the two largest groups of membrane receptors: receptor tyrosine kinases (RTKs) and G-protein-coupled receptors (GPCRs).⁴⁶ Activated PI3K phosphorylates phosphatidylinositol 4,5-bisphosphate (PIP2), which is converted to phosphatidylinositol 3,4,5-triphosphate (PIP3). PIP3 binds to Akt and promotes Akt phosphorylation.⁴⁷ Then, phosphorylated Akt triggers the downstream effector mTOR and results in gene transcription to facilitate cell growth and metabolism, motility, and angiogenesis and suppress apoptosis. Besides, activated mTORC2 (mTOR complex 2) can also promote the hyperactivation of Akt by phosphorylating Akt.⁴⁸ Two tumor suppressors [phosphatase and tensin homolog (PTEN)⁴⁹ and inositol polyphosphate 4-phosphatase type II (INPP4B)]⁵⁰ prevent the activation of downstream of PI3K by dephosphorylating PIP3 and PIP2.^51,52

The PI3K/AKT/mTOR pathway is often genetically altered in different human cancers.^53,54 Although many developed small-molecule inhibitors target this pathway, only a few have been approved by the FDA for therapeutic use.

Some isoform-specific inhibitors of PI3K have been approved for treating lymphoma/leukemia and breast cancer,⁵⁵ including a pan-PI3K inhibitor (copanlisib/BAY 80-6946/Aliqopa),⁵⁶ a dual PI3Kγ/δ inhibitor (duvelisib/IPI-145/Copiktra), an α-selective PI3K inhibitor (alpelisib/NVP-BYL719/Piqray),⁵⁷ and a δ-selective PI3K inhibitor (umbralisib/TGR-1202, idelalisib/CAL-101/GS-1101/Zydelig).⁵⁸

As the critical effector of the PI3K/Akt/mTOR pathway, genetic alterations of Akt or its abnormal expression initiate tumor development and lead to resistance to chemotherapy and radiotherapy.⁵⁹ Many small-molecule inhibitors of Akt have been evaluated in clinical trials,⁶⁰ but none have been approved for clinical use as of July 2022.

As a downstream effector of PI3K/Akt, mTOR is usually hyperactive in various tumor types. mTOR-selective or dual mTOR/PI3K small-molecule inhibitors⁶¹ have been developed, and four anticancer mTOR-specific inhibitors have been approved by the FDA: (1) sirolimus (rapamycin) to treat lymphangioleiomyomatosis⁶²; (2) everolimus to treat advanced renal cell carcinoma (RCC), renal angiomyolipoma (AML), postmenopausal advanced HR-positive, HER2-negative breast cancer, progressive neuroendocrine tumors of pancreatic origin, and subependymal giant cell astrocytoma (SEGA)⁶³; (3) temsirolimus to treat advanced RCC⁶⁴; and (4) Fyarro (sirolimus albumin-bound nanoparticles, nab-sirolimus, ABI-009), the latest mTOR inhibitor approved in November 2021, to treat patients with locally advanced unresectable or metastatic malignant perivascular epithelioid cell tumor.⁶⁵ PI3K/mTOR dual inhibitors, combining multiple therapeutic effects in a single molecule, reduce the activity of PI3K and mTOR by competitive interaction with the ATP-binding cleft of these enzymes. Recently, an array of inhibitors,^66,67,68,69 such as dactolisib (BEZ235), samotolisib (LY3023414), and bimiralisib (PQR309), have been studied in clinical trials, but none has yet been approved for use clinically.

Immunomodulatory signaling pathways

As newcomers to the human body, tumor cells generate and express specific antigens on their surface, which can be recognized and eliminated by immune cells such as cytotoxic T lymphocytes cells (CTLs). However, malignant tumor cells develop multiple escape mechanisms to evade immune recognition and surveillance. Reversing these immune evasion strategies is a promising approach for antitumor therapy. Immune checkpoint blockade (ICB) therapies targeting the programmed cell death protein 1 (PD1)/programmed cell death ligand 1 (PD-L1)^70,71 pathway or cytotoxic T lymphocyte antigen 4 (CTLA4)^72,73,74 have revolutionized the treatment landscape for multiple cancer types.

PD-1/PD-L1 signaling pathway

PD-1 is expressed on the surface of activated T cells and acts as an immunosuppressant, while its ligand PD-L1 is mainly overexpressed on tumor cells. The binding of PD-1 to PD-L1 suppresses T-cell-mediated immune responses by inhibiting T-cell proliferation, limiting cytokine production, and ultimately resulting in immune evasion of tumors.^75,76,77

Blockade of the PD-1/PD-L1 interaction with specific antibodies results in the rescue of functionally exhausted T cells and the reactivation of immune responses. As of 2022, based on highly successful clinical trials, the FDA, European Medicines Agency (EMA), and National Medical Products Administration (NMPA) have approved 10 anti-PD-1 (nivolumab, pembrolizumab, cemiplimab, sintilimab, camrelizumab, toripalimab, tislelizumab, zimberelimab, prolgolimab, and dostarlimab) and three anti-PD-L1 antibodies (atezolizumab, durvalumab, and avelumab) for various hematological and solid malignancies.⁷⁸

CTLA-4 signaling pathway

During the process of T-cell activation, membrane CTLA-4 and secreted soluble CTLA-4 are upregulated on CD8⁺ T cells and CD4⁺ T cells. Then, CD28 binds to the costimulatory molecules B7-1 (CD80) and B7-2 (CD86) as a secondary signal. CTLA-4 competitively binds to B7 to block the B7-CD28 signaling pathway.⁷⁹ Moreover, the intracellular domain of CTLA-4 becomes phosphorylated, which generates a negative signal blocking the activation and function of T cells.⁸⁰ In addition, CTLA-4 is abundantly found in forkhead box p3 (Foxp3)⁺ regulatory T cells (Tregs), which mainly works to suppress T-cell activity.⁸¹ Therefore, CTLA4 is considered a negative regulator of T-cell activation.

CTLA-4 inhibitors interfere with the interaction of CTLA-4 and B7 to erase the suppressive impact of CTLA-4 on T-cell activity and promote antitumor immune response, leading to tumor regression. Considerable evidence has also shown that CTLA-4 inhibitors function through antibody-dependent cell-mediated cytotoxicity (ADCC) via the Fc receptor (FcR) to remove Treg and downregulate the immunosuppressive effect effectively.⁸² Ipilimumab is a commonly used CTLA-4-blocking antibody approved by the FDA.⁸³

The signaling pathways discussed above are deemed as potentials therapeutic targets. We summarized the main signal transduction pathways in Fig. 1 by illustrating key signal transduction processes. Besides, based on the defined signal pathways, these FDA-approved and commonly used targeted anti-cancer drugs were summarized in Table 1, as well as their related targets, indications, and categories (i.e., antibody or small-molecule inhibitors or ADCs).

Fig. 1

The pathway diagram of VEGFR, EGFR, HER2, CDK4/6, and PD1/PDL1 signaling. The signal transduction signaling pathways and some targeted inhibitors

Table 1 FDA-approved and commonly used targeted anti-cancer drugs

Molecular imaging

Molecular imaging is a noninvasive medical imaging method that enables the visualization, characterization, and measurement of biological processes at the molecular and cellular levels in tumors.^84,85 In contrast to conventional imaging modalities that primarily image differences in the structure of tissues or organs, molecular imaging reveals the physiological activities or expression status of specific molecules within a tissue or organ by employing medical imaging modalities with or without tracers.

From the phenomenon of magnetic resonance spectroscopy observed in 1966,⁸⁶ to the first SPECT instrument developed in 1976,⁸⁷ to the first whole-body MRI scanner in 1977,⁸⁸ to luciferase (Luc) used as a reporter of gene expression in vivo in 1986,⁸⁹ to near-infrared fluorescence (NIRF) imaging developed in 1994,⁹⁰ to the first PET-CT completing the unity of both function and anatomical imaging in clinical practice in 1998,⁹¹ to the photoacoustic imaging first used in human in 2002,⁹² to the NIR-II imaging proposed in 2009.⁹³ Some other historic steps promote the development of molecular imaging (Fig. 2). Given this excellent work, in September 1999, Weissleder and other imaging authorities held an international imaging conference in Jackson, the capital of Mississippi. The participating experts agreed that molecular imaging has emerged as a new field. Since then, molecular imaging has accelerated oncology detection, surgical guidance, targeted drug delivery, imaging-guided therapy, and efficacy evaluation.⁹⁴ Given that several valuable reviews on molecular imaging have recently been published,^{95,96,97,98,99,100} here we only highlight the modalities most commonly used and their application for evaluating the efficacy of cancer-targeted therapy.

Fig. 2

The historic steps in molecular imaging technology

Nuclear imaging

PET imaging

PET is a molecular imaging technique that uses radiotracers to visualize and quantify the biological characteristics of tumors. PET is based on the principle that radionuclides emit positrons when decaying. The emitted positron is annihilated with an electron to create two 511 keV gamma rays at an angle of 180°.¹⁰¹ A ring of detectors is used to detect these emitted gamma rays. Radionuclides that are available for PET imaging in clinical and research applications include fluorine-18 [¹⁸F], carbon-11 [¹¹C], zirconium-89 [⁸⁹Zr], gallium-68 [⁶⁸Ga], and copper-64 [⁶⁴Cu]. Among these, ¹⁸F is most commonly used for clinical applications because of its beneficial half-life (T_1/2 = 1.8 h), positron yield, and associated detection sensitivity.¹⁰²

Owing to the development of nuclear medicine, there is now a wide variety of radiopharmaceuticals available in clinical practice to evaluate the biological features of tumors, such as ¹⁸F-FDG for tumor metabolism,^{101 18}F-FMISO for hypoxia,^{103 18}F-FLT for tumor cell proliferation,^{104 18}F-labeled amino acids for protein synthesis,¹⁰⁵ and ¹⁵O–water for blood flow.¹⁰⁶ Furthermore, PET, in combination with tumor-specific monoclonal antibodies (immune-PET), has broadened the application of PET imaging. Various monoclonal antibodies and radionuclides have been explored to develop immune-PET tracers.¹⁰⁷ In addition, peptides and other receptor-targeting compounds, such as nanobodies or bispecific antibodies, are being used to design novel immune-PET tracers.^107,108 By dynamically monitoring tumor antigens’ expression, immune-PET imaging is a promising technique for evaluating the efficacy of cancer-targeted therapy.

Inherent advantages of PET imaging include its high sensitivity and quantifiable imaging parameters, such as standardized uptake value (SUV). PET-CT combining anatomical and functional imaging information becomes possible to assess the molecular features of tumors with highly accurate anatomical structure correction. Furthermore, PET–MRI combining PET with MRI extends the scope of multimodality imaging and reduces patients’ exposure to radiation.¹⁰⁹

SPECT imaging

SPECT is another nuclear imaging modality that uses radionuclides that emit single photons, such as technetium-99m (^99mTc), iodine-123 (¹²³I), and indium-111 (¹¹¹In). The emitted single photons are subsequently detected by a gamma camera to image the organs of interest or the whole body. SPECT has limited spatial resolution and lower sensitivity in tumor detection compared to PET. The quantification of SPECT is also more challenging. However, despite these limitations, it is more commonly used than PET in clinical practice. This is mainly due to its advantage of having a large number of radiopharmaceuticals that are readily available for clinical use. Specifically, ~85% of radiopharmaceuticals can be detected in clinical practice by SPECT imaging.¹¹⁰ Tumor-specific biological compounds, such as antibodies or peptides, can be used for radiolabeling to produce tumor-targeted SPECT radiopharmaceuticals, e.g., ¹²³I-VEGF for VEGFR targeting, ¹¹¹In-bevacizumab for VEGF targeting, ¹¹¹In-trastuzumab for HER2 targeting, and ¹¹¹In-EGF for EGFR targeting.^111,112,113 These approaches have been explored for monitoring the responses to anticancer treatment in humans.^110,111

Magnetic resonance imaging

MRI is a noninvasive imaging technique often classified as an anatomical imaging modality. When placed in a strong magnetic field, specific atomic nuclei can absorb radiofrequency energy and align like small magnets because of their spin. The absorption of energy by the nuclei causes a transition from a high- to a low-energy state. This further induces a voltage that can be detected, amplified, and displayed as “free-induction decay (FID),” which can be resolved by a mathematical process to generate high-resolution anatomical images.¹¹⁴ Moreover, using specific techniques, MRI demonstrates the ability to image molecular processes within a tumor. This provides functional information on tumor vascular permeability, perfusion,¹¹⁵ vascular volume and flow, tortuosity of extracellular space,¹¹⁵ and hypoxia.¹¹⁶ As such, these MRI techniques can be classified as forms of functional imaging.

Dynamic contrast-enhanced MRI

As suggested by its name, dynamic contrast-enhanced MRI (DCE-MRI) demonstrates the temporal enhancement pattern of a tissue following the injection of a paramagnetic contrast agent (CA), such as gadolinium-diethylenetriamine pentaacetic acid (Gd-DTPA). Factors influencing the tumor uptake of CA include blood perfusion, tissue vascularization, vessel permeability, cell density, extravascular extracellular volume fraction, and extracellular matrix density.¹¹⁷ A CA has two important physicochemical properties, namely, the relaxation effect and the susceptibility effect. MRI sequences studying the relaxation effect are termed DCE-MRI or T1-W DCE, while those assessing the susceptibility effect are termed dynamic susceptibility contrast (DSC)-MRI or T2*-W DCE.

T1-W DCE is sensitive to the presence of CA in the extravascular extracellular space. Tofts’ standard pharmacokinetic model is the mathematical model most commonly used in clinical studies for analyzing human DCE-MRI data.¹¹⁸ Tofts’ model introduces three key parameters, namely, K^trans, V_e, and V_p.^119,120 K^trans is the bulk transfer coefficient, which reflects the leakage of contrast from the vascular to the extravascular compartment.¹¹⁹ V_e is the fractional volume of the extravascular extracellular space (EES),¹¹⁹ while V_p represents the concentration of CA in plasma space.¹²⁰ The efflux rate constant from EES to plasma (K_ep), which is the ratio of K^trans to V_e, is also frequently described in DCE-MRI.¹²⁰ DCE-MRI has been used to evaluate the treatment response and demonstrate prognostic value in patients receiving cancer-targeted therapy. This is elaborated on in the following section.

T2*-W DCE is sensitive to the vascular phase of CA and is used to evaluate tissue perfusion and blood volume. The following parameters are measured in T2*-W DCE: (1) regional blood volume (rBV), which is defined as the volume (ml) of blood perfusing vessels in a voxel divided by the tissue mass in that voxel (g); (2) mean transit time (MTT), which is the average transit time of a CA particle through the capillary bed; and (3) regional blood flow (rBF), which reflects the tissue perfusion and is measured in milliliters per minute.

Diffusion-weighted imaging

Diffusion-weighted imaging (DWI) is an MRI technique that generates signal contrast based on differences in the diffusion of water molecules. Within the body, water is distributed in the intracellular and extracellular compartments. The water molecules in the extracellular compartments diffuse relatively freely, while those within the intracellular compartments experience more restricted diffusion. The diffusion of water molecules can be quantitatively evaluated by the apparent diffusion coefficient (ADC). A lower ADC value reflects restricted diffusion. In tumor tissue, especially those with a high histological grade, the ADC value is lower than that in surrounding normal tissue. This is due to the high cellularity of tumors, as cellularity is positively correlated with the degree to which the diffusion of water is restricted.^{121,122,123,124} The use of DWI for evaluating the response to anticancer treatments has been explored. An increase in the ADC value can be observed at an earlier stage than a decrease in tumor size.¹²⁵

Magnetic resonance spectroscopy

Magnetic resonance spectroscopy (MRS) uses the same principles of signal acquisition as other MRI techniques. MRS studies in a medical context usually involve the detection of radiofrequency electromagnetic signals that are produced by chemical compounds. As a result, this technique provides chemical information on tissue metabolites.¹²⁶ Therefore, MRS can be used to monitor the metabolic variations caused by treatments and treatment efficacy.¹²⁷

Chemical exchange saturation transfer

Chemical exchange saturation transfer (CEST) is an advanced MRI technique in which the exchangeable protons on target metabolites are selectively saturated and exchanged with water protons. The relative concentration of target metabolites can be measured with enhanced sensitivity by detecting the attenuation of the water proton signal indirectly.¹¹² Amide proton transfer (APT) imaging is the CEST technique most commonly used in a clinical context to indirectly detect proteins and peptides in tissue. This technique provides important information for the diagnosis and monitoring of tumors.

Targeted magnetic resonance imaging

Like other tumor-targeted imaging modalities, targeted magnetic resonance imaging (TMRI) can visualize tumor-specific molecular markers with targeted magnetic nanoparticles. This provides the opportunity to classify patients, deliver individualized therapy directly to tumors, and monitor the treatment response through MRI for tumors expressing specific biomarkers.^128,129

Optical imaging

Optical imaging is a noninvasive technique that uses light and optical properties of protons to image tissues, cells, and even molecules within the body.¹³⁰ One of the major advantages of optical imaging is that it uses nonionizing radiation, making it much safer than techniques that use ionizing radiation such as X-rays. This makes optical imaging suitable for being repeatedly used to monitor gene expression, disease progression, or treatment response. Optical imaging includes multiple submodalities, such as bioluminescence imaging (BLI), chemiluminescence, Cherenkov imaging, and fluorescence imaging (FLI).¹³¹

Bioluminescence imaging makes use of the reaction between luciferases and their substrates to produce light.¹³¹ Luciferases, such as firefly luciferase, can be constitutively or inducibly expressed, and as such used for tracking the expression of targeted genes or monitoring tumor growth or regression to evaluate drug efficacy. These applications of BLI have been routinely used in preclinical studies.^{132,133,134,135} However, the drawbacks of BLI, such as the need for cell transfection and administration of a reactive substrate, prevent its clinical translation.

Fluorescence imaging provides images of tumors by detecting the emitted light that is generated from genetically encoded fluorescent proteins or fluorescent dyes after excitation by light of a different wavelength.¹³¹ The fluorescent dyes can be used to label tumor-specific antibodies, peptides, or nanobodies, enabling tumor-targeted molecular imaging. In comparison with BLI, a significant disadvantage of FLI is that its signal-to-background ratio is lower due to the auto-fluorescent noise coming from endogenous fluorophores within the tissue, which absorbs the excitation light. Other fundamental factors influencing the image quality of FLI are diffraction and diffusion.¹³¹ Diffraction reduces the spatial resolution of fluorescent images, while diffusion is caused by the tissue scattering of light, which limits the tissue penetration depth. Near-infrared (NIR) fluorescence imaging, including the first NIR (NIR-I) window (650–950 nm) and second NIR (NIR-II) window (1000–1700 nm), was emerged recently as an attractive imaging modality with high sensitivity, relative safety, and low cost. Compared with traditional NIR-I imaging, NIR-II imaging has less autofluorescence, absorption and scattering of light, higher penetration depths, and spatiotemporal resolution for biological tissues.¹³⁶ FLI is primarily used in preclinical studies for tumor detection, fluorescence image-guided surgery, and monitoring of response to therapy.^{137,138,139,140} Along with the development of good manufacturing practice (GMP) tracers and clinical imaging systems, NIR FLI has been actively explored in early clinical trials for guiding cancer surgeries.^{141,142,143,144} However, the clinical application of FLI for evaluating the efficacy of targeted treatment has not been reported.

Photoacoustic imaging

Photoacoustic imaging (PAI) is a novel noninvasive molecular imaging modality, which generates an ultrasound signal based on the photoacoustic effect. When laser pulses are delivered to a material, some of the energy is absorbed and converted to heat, resulting in a thermoelastic expansion that generates an ultrasonic signal from which images can be produced.¹⁴⁵ As a hybrid of optical imaging and ultrasound imaging, PAI combines the high contrast and sensitivity of the optical property and high ultrasonic spatial resolution in a single imaging modality. In addition, the unprecedented imaging depth (up to centimeters) makes this a promising technique for various clinical applications.^146,147 PAI can be used to analyze various endogenous contrast agents such as oxygenated and deoxygenated hemoglobin, lipids, melanin, and water.¹⁴⁷ By using multiwavelength measurement, PAI can simultaneously quantify the concentrations of these endogenous chromophores and further provide biological information on tissues that reflects their different physiological or pathophysiological status.¹⁴⁸ Furthermore, by using tumor-specific exogenous contrast agents, PAI can identify tumor cells and monitor the expression of tumor-specific biomarkers.^{149,150,151,152,153} Examples of these agents include gold nanoparticle-conjugated peptides or antibodies for EGFR and HER2 molecular imaging.^152,154 The ability to perform both functional (by endogenous contrast) and molecular (by exogenous contrast) imaging makes PAI an attractive technique for evaluating tumor-targeted therapy.

Ultrasound imaging

Ultrasound is a technique that uses high-frequency sound waves to produce anatomical images. It possesses several advantages, such as high availability, lack of radioactivity, and cost-effectiveness. These merits make it suitable to be repeatedly used in clinical practice. The Doppler technique allows ultrasound to be used to assess the blood flow in tumors. More importantly, with the use of contrast agents such as microbubbles, dynamic contrast-enhanced ultrasound (DCE-US) can measure longitudinal changes in hemodynamic parameters (e.g., perfusion, flow velocity) and morphological parameters (e.g., blood volume, vascular heterogeneity) of a given tumor relative to the findings in a pretreatment baseline assessment. The value of information on the changes in these parameters for monitoring the therapeutic response induced by anti-angiogenic therapies has been explored.¹⁵⁵

Molecular imaging probes

Among the above-mentioned imaging modalities, some modalities, such as PET, SPECT, and optical imaging, require the injection of imaging probes into the studied subjects to acquire an imaging signal. On the other hand, other modalities, such as photoacoustic imaging and MRI, can monitor the biological change associated with diseases either through the injection of exogenous molecular probes or by using endogenous molecules.¹⁵⁶ Tumor nonspecific imaging probes, such as ¹⁸F-FDG for PET imaging and ^99mTc-sulfur colloid for SPECT imaging, have been widely used in clinical practice for detecting lesions or lymph nodes in cancer patients and for evaluating the efficacy of anticancer treatments.^157,158,159 However, concerning tumor-specific imaging probes, many are still under development and are being tested in preclinical or early-stage clinical studies.^160,161,162 A comprehensive review of all molecular imaging probes is beyond the scope of this paper. Therefore, here we only briefly describe the tumor-specific imaging probes that have been employed to evaluate the efficacy of cancer-targeted therapy. The mechanism of action of tumor-specific imaging probes is based on the concept that a carrier molecule that is labeled with a positron emitter for PET imaging, a single-photon emitter for SPECT imaging, or a fluorophore for fluorescent imaging specifically binds to a certain tumor target. Carrier molecules can be monoclonal antibodies, monoclonal antibody fragments, affibody molecules, small peptides, or small molecules that specifically target certain cell surface markers that are overexpressed in tumors.^163,164,165 Moreover, carrier molecules can also be small molecules that detect the acidic microenvironment of tumors.¹⁴²

Monoclonal antibody

Monoclonal antibody (mAb)-based probes have been most commonly studied in evaluations of cancer-targeted therapy using molecular imaging, mainly due to their high specificity and high binding affinity. In addition, the relative stability and tolerance of chemical modifications to mAbs make them desirable for creating moieties targeting radionuclides or fluorophores. For PET and SPECT imaging, owing to the prolonged circulation time of mAbs in the body, radionuclides with long physical half-lives, such as ¹²⁴I (t_½ = 100.3 h),^{166 89}Zr (t_½ = 78.4 h),¹⁶⁶ and ¹¹¹In (t^½ = 2.8 days),¹⁶⁷ should be chosen for radiolabeling. PET imaging of ⁸⁹Zr-mAb has shown it to be a biomarker for predicting the efficacy of cancer-targeted treatments in xenograft models and early clinical trials.^163,168 Nevertheless, the potential of mAb PET imaging or fluorescent imaging is limited by the slow clearance of intact antibodies from the blood, which causes undesirable high background signals and excessive nonspecific tissue accumulation such as in the liver.¹⁶⁹

Monoclonal antibody fragments

To reduce the undesirable high background signal and nonspecific tissue accumulation of mAb probes, specifically in PET and SPECT imaging, there is a demand for the use of antibody fragments that are cleared more rapidly from circulation. Furthermore, the shorter circulation time of antibody fragments requires shorter-lasting radionuclides such as ⁶⁸Ga (t_½ = 68 min),^{170 99m}Tc (t_½ = 6 h),¹⁶⁷ and ⁶⁴Cu (t_½ = 12.7 h) for radiolabeling, which in turn decreases radiation exposure in patients. All of these advantages make antibody fragments an attractive alternative to employ for molecular imaging.

Various antibody-derived fragments with different sizes, serum half-lives, bio-distributions, and levels of tumor penetration have been developed.¹⁶⁹ Full-length antibodies can be digested by enzymes to produce F(ab’)₂ (110 kDa) and F(ab) (50 kDa) fragments, or genetically engineered to generate a variety of antibody derivatives such as minibodies (75 kDa), scFvs (26 kDa), diabodies (50 kDa), and nanobodies (12–15 kDa).^167,169 These antibody fragments retain the specificity of binding to the molecular targets of their parental IgG. However, much of the data on their application in molecular imaging has been obtained from preclinical studies.¹⁶⁷ Clinical translational studies on them have been performed in only limited numbers of cancer types and patients.¹⁶⁷ Examples include the use of a ⁶⁸Ga-labeled anti-HER2 VHH for detecting lesions in patients with breast cancer.¹⁷¹

Affibody molecules

Affibody molecules are engineered scaffold proteins consisting of 58 amino acids with a molecular weight of 6–7 kDa, which meet the requirements for an optimal balance between clearance from circulation and extravasation.¹⁷² This ensures the high absolute tumor uptake of affibody molecules and further increases the signal-to-background ratio. Several affibody molecules with high affinity to VEGFR2, EGFR, HER2, HER3, and IGF-1R have been demonstrated as probes for radionuclide or fluorescent molecular imaging in preclinical settings.^167,173 Clinical evaluation of radiolabeled affibody molecules has been explored for HER2 imaging^{174,175,176,177} and fluorescent dye-labeled affibody molecules (ABY-029) for EGFR imaging have been tested in patients with recurrent glioma.¹⁷⁸

Peptides

According to the definition by the United States Food and Drug Administration (FDA), peptides are proteins consisting of fewer than 40 amino acids. Although their binding affinity is lower than that of antibodies, they are small, easy to synthesize, and capable of flexible chemical modification.¹⁷⁹ With these distinctive advantages over antibodies, peptides have been increasingly used as probes for tumor receptor imaging, such as peptides for PD-L1 imaging,^{180,181,182,183} integrin αvβ3 imaging,^184,185,186 and somatostatin receptor imaging.¹⁸⁷ To date, many peptide-based probes have been developed and clinically used for cancer diagnosis and treatment.^187,188,189

Application of molecular imaging in targeted therapy

As discussed above, both conventional imaging modalities and standard-of-care assessment of tumor responses to treatment are based on morphological indicators such as tumor size. Molecular imaging is more sensitive and may detect changes earlier than morphological changes in tumors because it can visualize the expression of a specific molecule when used to monitor therapeutic efficacy.¹⁹⁰ Figure 3 shows the representative diagram of how molecular imaging probes are designed and how responses to targeted therapy are evaluated.

Fig. 3

The diagram of molecular imaging and functional imaging in cancer-targeted therapy. The probes are constructed by imageable agent and their targets, respectively. On the cellular level, probes can bind to cell surface receptors or targets in the cytoplasm or nucleus to visualize and measure the target. After the probes are injected into animals or the human body, quantitative measures of probe uptake are used as predictive or evaluative assays for response to targeted therapy by different molecular imaging technologies

Nuclear imaging

Because of temporal and spatial heterogeneity and discordance in gene expression status between primary and distant metastatic lesions,¹⁹¹ the reproducible, noninvasive, whole-body evaluation of the efficacy of targeted therapy is critical for determining optimal treatment options. Among various types of molecular imaging, radionuclide imaging is advantageous for clinical usage because of the excellent sensitivity and tissue penetration of radionuclides. This approach would facilitate treatment optimization, prevent useless prescriptions, avoid unnecessary side effects, and, more importantly, prevent treatment failure in nonresponding patients.

Although nuclear imaging has been mainly utilized for monitoring the pharmacokinetics and pharmacodynamics of targeted drugs noninvasively,¹⁹² a discussion of this is outside the scope of this review. We instead discuss the application of nuclear imaging, including PET/CT and SPECT, in evaluating the response and predicting the prognosis after targeted therapy, among which the clinical application is summarized in Table 2.

Table 2 Clinical application of nuclear imaging in cancer-targeted therapy

Evaluation of therapeutic responses

In clinical practice, ¹⁸F-labeled fluorodeoxyglucose (¹⁸F-FDG) is the most commonly used radiotracer for PET-based imaging. PET/CT with 18F-FDG has been a proven staging modality for various neoplasms for many years. Besides staging, it is used increasingly frequently to categorize the metabolic response to antineoplastic therapy, called PET Response Criteria in Solid Tumors (PERCIST).¹⁹³ Several studies have confirmed that ¹⁸F-FDG PET or PET/CT can predict the response to EGFR-TKI treatment in 1–2 weeks, while conventional CT requires 2–3 months in patients with lung cancer.^{194,195,196,197,198} Similar results were also seen upon the use of HER2-targeted therapy,¹⁹⁹ antiangiogenic agents,^200,201,202 CDK4/6 kinase inhibitor,²⁰³ and mTOR inhibitor.²⁰⁴ Therefore, by using ¹⁸F-FDG, it is possible to evaluate the efficacy of targeted therapy at an early stage without waiting until the middle or end of treatment. For example, in patients with metastatic renal cell cancer treated with an mTOR inhibitor everolimus, the 2-week relative changes of ¹⁸F-FDG uptake (SUV_max) from baseline were predictive of the 8-week change in tumor size as evaluated by conventional computed tomography.²⁰⁴

¹⁸F-FLT is another PET tracer for assessing cell proliferation in vivo. It is also a potential candidate for evaluating response to targeted therapy, such as VEGFR TKI sunitinib,²⁰⁵ EGFR inhibitors,²⁰⁶ HER2 inhibitor trastuzumab,²⁰⁷ PI3K inhibitor GDC-0941, and the MEK inhibitor PD 0325901,²⁰⁸ mTOR inhibition everolimus,²⁰⁹ but most of the studies were in preclinical models.

Despite the optimal performance of ¹⁸F-FDG or ¹⁸F-FLT in a clinical context, they suffer from certain limitations in evaluating therapeutic response as it only detects the glucose metabolism of tumors and cannot reflect all biometric features of tumors. Against this background, many studies are underway to explore and develop specific radiotracers capable of binding to particular targets and to identify efficacy as accurately as possible.

In addition to ¹⁸F, some other PET radioisotopes are commonly used, including ¹²⁴I, ⁸⁹Zr, ⁶⁸Ga, and ⁶⁴Cu.²¹⁰ Among these, ¹²⁴I and ⁸⁹Zr with long half-lives have been used for radiolabeling intact antibodies. In contrast, short half-life nuclides have been used for antibody fragments, nanobodies, peptides, affibodies, and small molecules.²¹¹

For antiangiogenic therapy, iodinated VEGF is the most studied radiolabeled VEGF tracer.^{212,213,214,215} VEGF, VEGFR, and related integrins can also be labeled with ¹⁸F,^{216,217 99}mTc,^{218,219,220 111}In,^{220,221 64}Cu,^218,222,223 and ⁸⁹Zr,²²⁴ but most of these are simply for imaging of the tumor vasculature and can help to visualize tumors and metastatic lesions usually overexpressing VEGFR or VEGF. Recently, reports have been published on several clinical trials on the efficacy of ⁸⁹Zr-labeled bevacizumab imaging for monitoring various cancer-targeted therapy such as the mTOR inhibitor everolimus,²²⁵ bevacizumab/interferon-α,²²⁶ multi-targeted tyrosine kinase inhibitors like sunitinib,²²⁶ and a similar result was found in ¹¹¹In-bevacizumab imaging for evaluation sorafenib.¹¹¹ In 70 evaluable lesions of 10 patients with metastatic renal cell carcinoma, the uptake of ⁸⁹Zr-bevacizumab (SUV_max) decreased by a mean of 9.1% (P < 0.0001) at 2 weeks and 23.4% (P < 0.0001) at 6 weeks after everolimus treatment. All 10 patients continued the treatment and had stable disease at 3 months.²²⁵ Similar data were found in patients with metastatic renal cell carcinomas treated with bevacizumab/interferon-α.²²⁶ In patients who received sunitinib from the same study, the uptake of ⁸⁹Zr-bevacizumab (SUV_max) decreased by 14.3% (P = 0.006) at 2 weeks but increased by 72.6% (P < 0.0001) at 6 weeks after treatment compared with that at baseline.²²⁶ These data indicate the role of ⁸⁹Zr-bevacizumab imaging in reflecting the biological effects of cancer-targeted therapy.

Evaluation of the success of HER2-targeted therapy depends on the precise determination of HER2 expression. Full-length monoclonal antibodies are usually labeled with long-half-life radionuclides, such as ⁶⁴Cu,^{227,228,229,230,231 89}Zr,^{232,233,234 111}In,^235,236 and ¹²⁴I.²³⁷ Besides, imaging tracers targeting HER2 by shorter fragments (antibody fragments [Fab or F(ab)2], nanobodies, or affibodies) have been introduced into clinical trials, such as ⁶⁸Ga-DOTA-F(ab0)2-trastuzumab,^{238 68}Ga-NOTA-2Rs15d,¹⁷¹ ABY-002 labeled with ¹¹¹In and ⁶⁸Ga,^{239 111}In-ABY-025,^{174 68}Ga-ABY-025,^{175 99}mTc-ADAPT6,²⁴⁰ and ⁹⁹mTc-(HE)3-G3,²⁴¹ to assess HER2 expression in breast carcinoma.

Most anti-HER2 probes exhibit the potential advantages of enabling the noninvasive and specific identification of HER2-positive tumors. In a study by Guo et al. ²⁴², PET imaging of ¹²⁴I-trastuzumab showed a difference in SUV_max (7.83 ± 0.55 vs. 1.75 ± 0.29, P < 0.0001) between HER2-positive and -negative lesions and recognized 18 out of 18 HER2-positive lesions in both primary and metastatic gastric cancer patients. In addition, Ulaner and colleagues published a series of articles about ⁸⁹Zr-trastuzumab^243,244 and ⁸⁹Zr-pertuzumab^245,246 to identify HER2-positive metastases in patients with HER2-negative primary breast cancer. Similar results were confirmed for ⁶⁴Cu-DOTA-trastuzumab^228,247,248 in breast cancer. Besides screening HER2-positive lesions, some novel molecular imaging probes were designed to monitor the response to anti-HER2 treatment directly.^249,250,251 In a preclinical study, ⁸⁹Zr-pertuzumab accurately detected changes in tumor volume from 243.80 ± 40.91 mm³ before T-DM1 therapy to 78.4 ± 40.43 mm³ after this therapy in mice bearing BT-474 tumors. In contrast to the findings with ⁸⁹Zr-pertuzumab, no apparent changes were observed in ¹⁸F-FDG.²⁴⁹ Similar preclinical results were obtained with ⁹⁹mTc-HYNIC-H10F, which can assess trastuzumab response at the earlier stage of day 4 post-treatment.²⁵⁰ In addition, in 2021, a study reported that ⁶⁴Cu-DOTA-trastuzumab could predict the response of metastatic breast cancer patients receiving T-DM1.²⁵² When compared with those in nonresponding patients, higher minimum SUV_max (5.6 vs. 2.8, P < 0.02) at day 1, higher average SUV_max (8.5 vs. 5.4, P < 0.05), and higher minimum SUV_max (8.1 vs. 3.2, P < 0.01) at day 2 were found in responding patients.

For trastuzumab-related cardiotoxicity, Perik et al. found that myocardial ¹¹¹In-DTPA-trastuzumab uptake was not associated with cardiotoxicity in all 15 evaluable HER2-positive metastatic breast cancer patients.²⁵³ Next, they discovered that myocardial ¹¹¹In-DTPA-trastuzumab uptake was observed in 50% of anthracycline-treated patients without symptomatic cardiac dysfunction, while none was found in non-anthracycline-related heart failure patients. They thus considered that ¹¹¹In-DTPA-trastuzumab potentially recognized patients susceptible to trastuzumab-related cardiotoxicity.²⁵⁴

For EGFR-targeted therapy, cetuximab, panitumumab, and their analogs were labeled with different radionuclides. Most of the tracers, such as ¹⁸F-FBEM-cEGF (ligand),^{255 64}Cu-panitumumab,^{256 111}In-cetuximab,^{257,258 64}Cu-cetuximab,^{258 89}Zr-cetuximab,^{259 111}In-cetuximab-F(ab’)2 (antibody fragment),^{260 64}Cu-cetuximab-F(ab’)2 (antibody fragment),²⁶¹ and ⁸⁹Zr-DFO-ZEGFR:2377 (affibody),²⁶² were designed to image the expression of EGFR in vivo. In 10 advanced colorectal cancer (mCRC) patients without K-RAS mutation, ⁸⁹Zr-cetuximab uptake was found in 6 patients, 4 of whom benefited from cetuximab treatment. Disease progression was detected in 3 of the remaining 4 patients without uptake of ⁸⁹Zr-cetuximab. The results suggested that ⁸⁹Zr-cetuximab uptake is correlated with response, but this warrants further clinical validation.²⁶³ Besides, N-(3-chloro-4-fluorophenyl)-7-(2-(2-(2-(2-18F-fluoroethoxy) ethoxy) ethoxy) ethoxy)-6-methoxyquinazolin-4-amine (¹⁸F-MPG) precisely quantified EGFR-activating mutation status, meanwhile monitored the response to EGFR-TKIs in NSCLC patients.²⁶⁴

To predict the treatment response of CDK4/6 inhibitors combined with endocrine therapy in ER+/HER2-metastatic breast cancer (MBC) patients, ¹⁸F-FES-PET/CT was undergone. 90% ¹⁸F-FES-negative patients developed progressive disease (PD), while only 8.7% ¹⁸F-FES-positive patients had PD. ¹⁸F-FES-positive patients had longer PFS than ¹⁸F-FES-negative patients (23.6 months vs. 2.4 months, P < 0.001).²⁶⁵

Immunotherapy is associated with pseudo-progression, which limits the application of conventional anatomically based imaging modalities for treatment evaluation. Recently, some studies have demonstrated the use of radiolabeled PD-L1 antibodies (⁶⁸Ga-NOTA-Nb10989, ⁸⁹Zr-atezolizumab, ⁸⁹Zr-durvalumab)^168,266,267 and PD-1 antibodies (⁸⁹Zr-pembrolizumab, ⁸⁹Zr-nivolumab)^268,269,270 to assess the efficacy of PD-1/PD-L1 blockade in cancer patients. The first-in-human whole-body PD-L1 imaging study by Niemeijer et al. used ¹⁸F-BMS-986192 and ⁸⁹Zr-nivolumab in 13 NSCLC patients with nivolumab treatment²⁶⁹(see Fig. 4). They showed that median SUV_peak values of ¹⁸F-BMS-986192 (6.5 vs. 3.2, P = 0.03) and ⁸⁹Zr-nivolumab (6.4 vs. 3.9, P = 0.019) were higher in responding lesions than in nonresponding ones. Similar results were obtained for using ⁸⁹Zr-atezolizumab to assess atezolizumab response in 22 patients, including 9 bladder cancer patients, 9 NSCLC patients, and 4 triple-negative breast cancer patients.¹⁶⁸

Fig. 4

Tracer uptake and correlation with PD1/PDL1 treatment response. a Study design. b Immunohistochemical staining of PD-L1 in patient 2. Biopsy of the tumor in the left lower lobe. PD-L1 expression is expressed in 95% of the tumor cells. Scale bar, 100 µm. c Immunohistochemical staining of PD-1 in patient 2. PD-1 expression in aggregates was scored as IC1. Scale bar, 100 µm. d ¹⁸F-FDG PET (225 MBq) (¹⁸F-FDG PET scan images of both patients were used from archival PET scans) demonstrates high glucose metabolism of tumors in both lungs and mediastinal lymph nodes. ¹⁸F-BMS-986192 PET (145.7 MBq, imaging time point 1 h post-injection (p.i.)) and ⁸⁹Zr-labeled Nivolumab PET (37.09 MBq, 162 h p.i.) demonstrate heterogeneous tracer uptake within and between tumors. e Patient 3 with tumor PD-L1 expression < 1%: ¹⁸F-FDG PET (268 MBq) (18F-FDG PET scan images of both patients were used from archival PET scans) demonstrates high glucose metabolism in the left-sided tumor. ¹⁸F -BMS-986192 PET (214.62 MBq, 1 h p.i.) demonstrates low tumor tracer uptake. ⁸⁹Zr-labeled Nivolumab PET (37.27 MBq, 160 h p.i.) demonstrates heterogeneous tracer uptake in the tumor. f SUV_peak of the ¹⁸F-BMS-986192 tracer is higher in responding lesions than non-responding lesions (comparison of lesions with a diameter of 20 mm or more). The p-value is 0.02, as determined by the Mann–Whitney U-test. SUV_peak of the ⁸⁹Zr-nivolumab tracer is numerically higher in responding lesions (comparison of lesions with a diameter of 20 mm or more). The p-value is 0.019, as determined by the Mann–Whitney U-test. For all the boxplots, the lower edge of the box represents the first quartile, and the upper edge represents the third quartile. The horizontal line inside the box indicates the median. Whiskers identify the minimum and the maximum value. (Reproduced from Niemeijer, A. N. et al. Whole body PD-1 and PD-L1 positron emission tomography in patients with non-small-cell lung cancer. Nat. Commun. 9, 4664 (2018))

Therefore, ¹⁸F-FDG PET is potentially useful for the early evaluation of therapeutic response. Tumor-specific nuclear imaging holds promise to assess the target expression and to predict the targeted treatment response early.

Prognosis prediction

Radionuclide imaging can also predict the prognosis of patients prescribed targeted therapy. As the most widely used tracer in oncology, ¹⁸F-FDG has been investigated in the early prediction of outcomes after targeted treatment, including that with anti-EGFR agents such as the monoclonal antibody cetuximab.²⁷¹ It was reported that an increase in peak tumor metabolism at the end of the first week of therapy implied poor PFS (P = 0.001) and OS (P < 0.001) at the end of the first week of third-line cetuximab-based therapy in metastatic colorectal cancer patients. Similar results were also found for the EGFR-TKIs erlotinib and gefitinib,^{272,273,274,275} the antiangiogenic agent bevacizumab,^{202,276,277,278} HER2-targeted therapy with trastuzumab,²⁷⁹ multiple TKIs such as regorafenib,^280,281 the CDK4/6 kinase inhibitors ribociclib, palbociclib, and abemcaciclib,^203,282,283 and the mTOR inhibitor everolimus,^204,284 as well as the ICIs nivolumab and pembrolizumab.^285,286 In a meta-analysis,²⁸⁷ 865 participants from 26 studies receiving inspection of FDG-PET or FLT-PET were included. In comparison with the PET nonresponsive group, the PET responsive group showed a decline in SUV_max, which was related to prolonged PFS (HR = 0.41, P < 0.00001), OS (HR = 0.52, P < 0.00001), and time to progression (TTP) (HR = 0.30, P = 0.003). However, some other clinical trials obtained inconsistent results on this issue. Rinzivillo et al. found no significant difference in clinical outcomes between ¹⁸F-FDG-PET-positive and -negative groups in patients with advanced neuroendocrine neoplasia receiving everolimus therapy (median PFS of 24 and 18 months, respectively; P = 0.337).²⁸⁴ In general, the prognostic value of ¹⁸F-FDG-PET in patients who received targeted therapy is inconclusive. The prediction based on ¹⁸F-FDG-PET imaging may be cancer-specific or treatment-specific.

To clarify the accuracy of PET or SPECT for predicting prognosis, especially in patients who have simultaneously received two or more combined targeted therapies or ICIs, a few specific radiotracers are under pre-clinical and clinical investigations. Rainer et al. investigated the predictive value of ¹²³I-VEGF165 scintigraphy in patients with glioma.²¹³ Their results showed that ¹²³I-VEGF165 may provide relevant prognostic information in glioma, as patients with a tumor-to-normal brain uptake ratio (T/N ratio) of <1.32 showed significantly longer survival (2680 days vs. 295 days; P < 0.05). Similar results were seen in a grade IV glioma subgroup, as patients with a T/N ratio < 1.75 had longer OS (720 vs. 183 days; P < 0.05).^{288 89}Zr-bevacizumab²²⁶ was used to predict differential effects of antiangiogenic treatment (bevacizumab with interferon-α or sunitinib) in metastatic RCC (mRCC). High baseline tumor SUV_max before antiangiogenic therapy in the three most intense lesions was correlated with longer time to disease progression (89.7 vs. 23.0 weeks; HR = 0.22; P = 0.050).

For HER2-targeted therapy, in 2016, Gebhart et al. ²⁵¹ conducted a prospective clinical trial (ZEPHIR study) using ⁸⁹Zr-trastuzumab PET/CT and FDG PET/CT to predict the efficacy of T-DMl therapy in patients with HER2-positive metastatic breast cancer. The results showed that, among the 55 evaluable patients, the negative predictive value was 100% when combining ⁸⁹Zr-trastuzumab-PET with early FDG-PET response after one cycle of T-DM1, which can predict response to T-DM1 and differentiate patients with a median time to treatment failure (TTF) of 2.8 months from those with a TTF of 15 months.

For EGFR-targeted therapy, even though FDG-PET/CT is better than CT at assessing the benefit of cetuximab in incurable squamous cell carcinoma of the head and neck,²⁸⁹ no relationship was identified between uptake on ⁸⁹Zr-cetuximab PET/CT and PFS (3.6 vs. 5.7 months, P = 0.15) or OS (7.1 vs. 9.4 months, P = 0.29) in patients with RAS wild-type advanced colorectal cancer receiving cetuximab monotherapy.

For immunotherapy, it was reported that high tumor uptake of ⁸⁹Zr-atezolizumab was correlated with a better response to atezolizumab treatment. Those with low uptake were more likely to progress or die, whereas PD-L1 IHC failed to predict the treatment outcome.¹⁶⁸ Similar results were seen for ⁸⁹Zr-pembrolizumab.²⁶⁸

Taking together, immuno-PET imaging, as a non-invasive method for the early detection of tumor receptor blocking by anti-cancer targeted drugs, may serve as an effective technique in predicting patient prognosis.

MRI imaging

Regarding the clinical application of MRI in cancer-targeted therapy, the focus has been particularly placed on three aspects: (1) predicting the response to targeted therapy by using information obtained by MRI performed pretreatment; (2) evaluating the treatment response to targeted therapy by analyzing the changes in MRI parameters; and (3) predicting patient outcome in those receiving targeted therapy. Data on the clinical application of MRI in cancer-targeted therapy are summarized in Table 3.

Table 3 Clinical application of MRI in cancer-targeted therapy

DCE-MRI and DSC-MRI

Among the different MRI techniques, DCE-MRI is the approach most commonly studied in evaluating treatment response to targeted therapies. In an early study of 21 patients with inflammatory and locally advanced breast cancer treated with bevacizumab, the DCE-MRI parameters K^trans, V_e, and K_ep significantly decreased compared with those at baseline after one cycle of bevacizumab.²⁹⁰ This was accompanied by a decrease in the tumor expression of p-VEGFR2 and an increase in tumor apoptosis, as evaluated by the TUNEL assay.²⁹⁰ Similar reductions in K^trans, V_e, and K_ep from baseline were observed in 70 patients with early breast cancer who received one cycle of bevacizumab as neoadjuvant therapy.²⁹¹ Besides, the changes in DCE-MRI parameters were significantly correlated to the changes in the SUV of FLT-PET imaging.²⁹¹ In patients with NSCLC under gefitinib or erlotinib therapy, K^trans, V_e, and V_p decreased significantly at day 7 post-treatment.²⁹² These results indicate that the changes in DCE-MRI parameters reflect the biological effects of bevacizumab or EGFR tyrosine kinase inhibitors on tumor cells and may serve as early noninvasive imaging biomarkers for evaluating the response to VEGFR-targeted agents.

Regarding anti-HER2 therapy, in a preclinical study, Syed and coworkers found that, in a mouse model with a HER2-positive tumor, both vascular heterogeneity and cellularity heterogeneity increased after trastuzumab treatment, as indicated by increases in the mean K–S distance for the K^trans distribution and V_e distribution, respectively.²⁹³ An increase in cellularity heterogeneity in trastuzumab-treated tumors is expected with increased tumor cell death. Therefore, the changes in these MRI parameters could potentially be used for evaluating the treatment efficacy of trastuzumab. In a study with 51 HER2-positive breast cancer patients who received neoadjuvant HER2-targeted therapy, the pretreatment initial enhancement ratio, defined as the percentage signal increase relative to the baseline at the first postcontrast acquisition, was not associated with tumor pathological complete response after treatment.²⁹⁴ Therefore, the changes in the DCE-MRI parameters of tumors after treatment may better reflect the impact of trastuzumab on tumors and serve as a more helpful tool for evaluating treatment efficacy.

The data from DCE-MRI imaging have also been shown to correlate with patient prognosis. For example, in a study of 22 patients with breast cancer that had metastasized to the brain, small reductions in K^trans and Peak in DCE-MRI scanning 3 weeks after bevacizumab therapy were independently correlated with shorter central nervous system-specific progression-free survival (PFS) and shorter overall survival (OS), respectively.²⁹⁵ Similar results were observed in patients with advanced hepatocellular carcinoma. Significant early reductions in K^trans, Peak, and AUC in the tumor were associated with longer PFS and OS than in those with smaller reductions.²⁹⁶ In addition, in patients with metastatic colorectal cancer, a drop in KEF, derived from K^trans and enhancing fraction (EF) (K^trans × EF), 15 days after regorafenib treatment was correlated with reduced CD31 expression (a marker of vascular density) in the tumor tissue, indicating the biological effect of the treatment.²⁹⁷ Moreover, patients with a >70% reduction in KEF had a higher disease control rate and longer PFS and OS than the remaining patients.²⁹⁷ The post-treatment contrast-enhancing volumetric change has been demonstrated to be prognostic. In patients with advanced-stage hepatocellular carcinoma treated with sorafenib, the difference in the enhancing lesion volume after treatment could discriminate patients with tumor progression from those with tumor control. Furthermore, this patient classification was proven to predict the prognosis as defined by OS independently.²⁹⁸ The similar predictive value of post-treatment enhancing volumetric change was confirmed in patients with recurrent glioblastoma receiving cabozantinib.²⁹⁹ The pretreatment contrast-enhancing volume could also be predictive. In post hoc analysis of the randomized ARTE trial, in which patients newly diagnosed with glioblastoma were treated with radiotherapy with or without bevacizumab, larger pretreatment contrast-enhancing volume was associated with inferior OS in both treatment arms.³⁰⁰ This result was supported by data from the ACRIN 6677/RTOG 0625 trial and the EORTC 26101 trial, which included patients with recurrent glioblastoma treated with bevacizumab and chemotherapy.^301,302 In addition, in both of these trials, early decreases in the relative cerebral blood volume (rCBV), derived from DSC-MRI, after treatment was associated with improved OS.^302,303 Moreover, in pediatric patients with supratentorial high-grade glioma treated with radiotherapy plus erlotinib, the CBV ratio of tumor to normal brain tissue remained relatively constant after treatment.³⁰⁴ Nevertheless, patients with a CBV ratio above 1.15 at 8 weeks after treatment had a shorter time to death than the other patients.³⁰⁴ Another study evaluated the predictive value of baseline rCBV before bevacizumab treatment in the ACRIN 6686 trial. Patients with newly diagnosed glioblastoma who had high pretreatment rCBV demonstrated improved OS in the bevacizumab-treated group compared with that in the placebo group.³⁰⁵ These findings indicate that monitoring the change in the blood volume of tumors may be more critical than merely learning about their absolute value in evaluating their prognostic significance. Contradictory data regarding the prognostic value of K^trans and rCBV have been reported in the literature, with the changes in these variables being reported not to be predictive for OS in patients newly diagnosed with glioblastoma.³⁰⁵ Moreover, in patients with NSCLC who were treated with bevacizumab and erlotinib, K^trans was not associated with PFS. However, an increase in K^trans heterogeneity 3 weeks after treatment was found to be associated with worse PFS.²⁰⁰ This discrepancy may have been due to the small number of patients in most of the studies and the different approaches used for defining the changes in K^trans among the studies.^295,305

Diffusion-weighted imaging (DWI)

DWI is another commonly used MRI technique for evaluating the response to anticancer treatments. Tang and colleagues demonstrated that in patients with gastrointestinal stromal tumor (GIST) treated by neoadjuvant imatinib or sunitinib, the percentage change in ADC 2 weeks after therapy differed significantly between responders and nonresponders (increase by 30% in responders vs. an increase by 1% in nonresponders, P < 0.001).³⁰⁶ Additionally, the predictive value of pretreatment ADC was confirmed in patients with unresectable intrahepatic cholangiocarcinoma (ICC) who received first-line systemic therapy with lenvatinib plus PD1 antibody.³⁰⁷ Another DWI-related parameter, the pretreatment fractional order parameter β (which correlates with intravoxel tissue heterogeneity), was proven to be predictive of treatment response in another study of GIST patients treated with second-line sunitinib.³⁰⁸

The prognostic value of DWI-related parameters has been explored in patients with multiple cancer types who received targeted therapies, such as those with unresectable ICC receiving lenvatinib plus PD1 antibody,³⁰⁷ recurrent glioblastomas receiving anti-VEGF monotherapy,³⁰⁹ colorectal liver metastases receiving bevacizumab,³¹⁰ and NSCLC brain metastases treated with whole-brain radiotherapy and gefitinib.³¹¹ In general, a higher baseline or post-treatment percentage change of ADC was associated with improved patient outcomes.^{300,307,309,310,311,312,313,314,315} These findings indicate that the pretreatment ADC value or post-treatment percentage change of ADC could accurately reflect the therapeutic efficacy of tumor-targeted therapies and predict patient survival. Therefore, the monitoring of ADC could potentially support the optimization of strategies in anticancer treatment.³¹³

Magnetic resonance spectroscopy

The most common metabolites detected by MRS are choline (Cho), lipids, N-acetyl aspartate (NAA), creatine/phosphocreatine (Cr), lactate, and glutamine.³¹⁶ In cancer tissues, the relative concentrations of these compounds are abnormal; for example, in brain tumors, Cho is generally increased, and NAA is decreased compared with the levels in normal brain tissue.^94,317 Thus, MRS has been commonly used for diagnosing brain tumors and evaluating the therapeutic response. The RTOG 0625/ACRIN 6677 trial, which included patients with recurrent glioblastoma treated with bevacizumab plus chemotherapy, demonstrated that the NAA/Cho level increased and the Cho/Cr level decreased within the enhancing tumor at 2 weeks after treatment compared with the pretreatment levels, indicating potential treatment efficacy.³¹⁷ Similarly, these metabolites demonstrated predictive value in patients with recurrent malignant glioma treated with tamoxifen.^318,319 In a critical preclinical study, Ros et al. showed that MRS could rapidly assess treatment response to PI3K inhibition in ER⁺ breast cancer mouse models by detecting the changes in lactate and pyruvate levels.³²⁰ Therefore, this imaging method could help to identify patients who would benefit from current treatments and design new drug combination strategies to counteract treatment resistance.

Chemical exchange saturation transfer

The application of CEST in evaluating treatment response is still in its infancy. In a preclinical study of a human mantle cell lymphoma xenograft model, the acidoCEST technique was used to monitor changes in the tumor extracellular pH (pHe) in response to an mTOR inhibitor, everolimus.³²¹ This study showed a significant increase in tumor pHe within 1 day of initiating treatment. Subsequently, acidoCEST MRI identified a decrease in tumor pHe 7 days after initiating treatment.³²¹ This contrasts with the findings of an untreated control group from another study performed by the same research group, which showed a consistent decrease in tumor pHe in the same tumor xenograft model.³²¹ This reflects the reduction in lactate production due to the inhibition of cellular metabolism by everolimus, as demonstrated in in vitro experiments.³²¹ The clinical application of CEST has been explored in small studies of patients with recurrent glioblastoma treated with bevacizumab.^322,323 Both pH-weighted amine CEST echoplanar imaging³²² and APT imaging³²³ were potentially useful for predicting the treatment response to bevacizumab and the PFS of patients. Moreover, residual or emerging regions of acidity, as assessed by pH-weighted amine CEST echoplanar imaging, may colocalize to the site of tumor recurrence, which may provide important information for site-specific treatment (see Fig. 5).³²²

Fig. 5

Three patient examples (a: patient #4, b: patient #8, c: patient #11) are demonstrated, with anatomic images (FLAIR and post-contrast T1-weighted images, T1 + C) and pH-weighted CEST-EPI images of MTR_asym at 3.0 ppm at three-time points: pre-bevacizumab (baseline), post-bevacizumab (follow-up), and the time of tumor recurrence. Baseline T2 lesion ROIs used for volume and median MTR_asym calculation are outlined in red (on FLAIR and post-contrast T1-weighted images) and black (on MTR_asym images). The red arrows demonstrate co-localization of residual or newly emerging areas of high acidity (MTR_asym at 3 ppm) at the post-treatment time point and the corresponding location of tumor recurrence ~2 months later. CEST-EPI: chemical exchange saturation transfer echoplanar imaging. ROIs: region of interest. FLAIR: fluid-attenuated inversion recovery. (Reproduced from Yao, J. et al. pH-weighted amine chemical exchange saturation transfer echoplanar imaging (CEST-EPI) as a potential early biomarker for bevacizumab failure in recurrent glioblastoma. J. Neurooncol. 142, 587–595 (2019).)

Other techniques

Several other novel techniques have been explored for evaluating the efficacy of treatment in cancer-targeted therapies, such as intravoxel incoherent motion (IVIM) MRI,^324,325 vascular architecture mapping (VAM) MRI,³²⁶ arterial spin-labeled (ASL) MRI,³²⁷ and magnetic resonance elastography (MRE).³²⁸ In general, these techniques demonstrated the potential to predict the treatment response or long-term outcome of patients. Additionally, MRI biomarkers of the tumor microenvironment, for example, necrosis, hypoxia with/without neovascularization, oxidative phosphorylation, and aerobic glycolysis, are predictive of the treatment response to bevacizumab in patients with glioblastoma.³²⁹ Gadolinium ethoxybenzyl diethylenetriamine (Gd-EOB-DTPA; a liver-specific contrast agent)-enhanced MRI is an imaging biomarker of OATP1B3, a transporter of Gd-EOB-DTPA.³³⁰ The expression of OATP1B3 is induced by Wnt/β-catenin mutation. Therefore, EOB-MRI is also considered an imaging biomarker of Wnt/b-catenin mutation/activation, which has been demonstrated to be an important mechanism of resistance to ICIs.³³⁰ In patients with hepatocellular carcinoma treated with ICIs, those with tumors with hyperintensity in EOB-MRI had significantly shorter PFS than patients with hypointense tumors.^331,332 Since these studies involved early trials with a small number of patients, the definitive predictive and prognostic value of these new techniques needs to be validated in future studies.

Thus, the MRI parameters, especially their change after treatment, are sensitive to the biological changes induced by cancer-targeted therapies. Together with their radiation-free nature, these MRI techniques are attractive tools in clinical practice for early treatment response assessment and patient prognosis prediction. Their clinical significance in improving patients’ outcomes is worthy of further evaluation in future prospective studies.

Optical imaging

The evidence on the application of optical imaging for predicting the efficacy of cancer-targeted treatment is from preclinical research but not human studies. BLI has been widely used in the preclinical setting for cancer detection, monitoring disease progression, and assessing the efficacy of anticancer treatment in vivo. The rapid and quantitative assessment of response to cancer treatment by this technique has accelerated drug discovery and development. However, a comprehensive review of the application of this technique in drug development is outside the scope of this paper. Here, we only present some examples to demonstrate how this technique can be used for evaluating the efficacy of cancer-targeted therapy.

Guo and colleagues developed genetically engineered bioluminescent reporters that reflected the G₁ phase alternation of the cell cycle.³³³ In vitro and in vivo experiments, this reporter system was shown to monitor G₁ phase arrest caused by a clinically used CDK4/6 inhibitor, palbociclib.³³³ In another preclinical study of a syngeneic murine triple-negative breast cancer model subjected to PD-1 inhibition, BLI could monitor the volume change of luciferase-tagged murine 4T1 tumors after treatment.³³⁴ Although minimal data from the literature demonstrated the use of FLI for assessing the efficacy of anticancer treatment, several FLI probes have been developed for monitoring the expression of HER2,^335,336,337 VEGF/VEGFR,^337,338 and EGFR.^339,340,341 These probes have the potential for the early detection of blocking of these receptors and may be used for early assessment of the efficacy of targeted treatment. Additionally, in a recently published study, Gao and colleagues developed an FLI probe for the real-time monitoring of CDK4 activity. In a hormone receptor-positive/HER2-negative breast cancer xenograft model, they demonstrated that the probe could reflect the therapeutic efficacy of palbociclib before an apparent change in the tumor size (see Fig. 6).³⁴² FLI probes have also been constructed for noninvasive, preclinical in vivo evaluation of the efficacy of cancer immunotherapeutics by detecting the presence of either the immune activation-related biomarker granzyme B³⁴³ or CD8⁺ cytotoxic T lymphocytes.³⁴⁴

Fig. 6

CPP30-Lipo/CDKACT4 reports the CDK4 inhibitor pharmacodynamics in vivo. a The chemical structures of molecular probes of CPP30-Lipo/CDKA4. b The activation of CPP30-Lipo/CDKA4 within cells. c Fluorescence images of mice bearing MCF-7 tumor injected with CPP30-Lipo/CDKACT4 before and after gavage of sterile water for 7 days, or before and after treatment with 150 mg/kg palbociclib daily for 7 days. The fluorescence signals were measured in radiance counts per cm² per second per steradian (p/s/cm²/sr) (left). Magnetic resonance imaging (MRI) of mice in (right). (Reproduced from Gao, Y. Y. et al. In vivo visualization of fluorescence reflecting CDK4 activity in a breast cancer mouse model. MedComm (2020) 3, e136 (2022).)

Interestingly, the clinical application of optical imaging techniques for evaluating the efficacy of cancer-targeted treatment has been explored using a diffuse optical tomography breast imaging system (DOTBIS).³⁴⁵ In a proof-of-principle study with seven postmenopausal women with early-stage breast cancer who received pre-surgical treatment with an AKT inhibitor (MK-2206) or an aromatase inhibitor (exemestane or letrozole), DOTBIS was performed on the tumor at baseline and post-therapy. Consistent decreases in DOTBIS-measured total hemoglobin, oxyhemoglobin, deoxyhemoglobin, and water fraction were observed in the tumor after treatment.³⁴⁵ This supports the value of further investigation of DOTBIS as a potential tool for assessing the response to cancer-targeted therapies in early-stage breast cancer.

In general, the exploration of optical imaging for treatment efficacy evaluation is in the very early stage. The clinical translation of BLI and FLI is challenging mainly due to their limited tissue penetration and the need for luciferase gene transfection in BLI. On the contrary, the clinical translation of DOTBIS is more feasible as it is a cheap, user-friendly, and repeatable technique.

Photoacoustic imaging (PAI)

As PAI can use endogenous contrast medium, such as hemoglobin, for real-time and noninvasive imaging, it is well suited for delineating the anatomy of the vasculature and evaluating tumor responses to antiangiogenic therapy.³⁴⁶ In a study by Yang et al. ³⁴⁷, PAI was used to measure early tumor response in a breast cancer mouse model treated with bevacizumab by clearly identifying the vessels surrounding tumors. By quantitative analysis, PAI parameters including MAP 760, MAP 840, hemoglobin (HbT), and deoxyhemoglobin (HbR) were shown to be significantly reduced 5 days after both high-dose and low-dose bevacizumab treatments compared with the levels in the control group, while no noticeable change in tumor volume was found. In the high-dose group, hypoxia showed negative correlations with these four parameters and CD31 (a marker of the maturation state) showed positive correlations with HbT, HbR, and MAP 760, while VMI (another marker of the maturation state) showed positive correlations with MAP 760 and HbR. Similar conclusions were reached by using photoacoustic tomography (PAT) in breast cancer mouse models treated with bevacizumab³⁴⁸ and an ovarian tumor mouse model treated with Trebananib.³⁴⁹ Moreover, Pham et al. ³⁵⁰ used contrast-enhanced ultrasound (CEUS) and PAI to perform the preclinical evaluation of the efficacy of bevacizumab in combination with CRLX101 (an investigational nanoparticle–drug conjugate). Therefore, PAI has been used as a noninvasive method in preclinical models to measure the effect of antiangiogenic therapy by visualizing vascular regression, normalization, and tumor hypoxia. Its clinical translation in cancer-targeted treatment efficacy evaluation is worthy of exploring, given the fact that PAI has been employed for the early detection of multiple cancers, including breast cancer, melanoma, and prostate cancer.³⁵¹

Multimodal imaging

Each imaging technique discussed above is associated with certain limitations.³⁵² For example, PET/SPECT are associated with poor spatial resolution and risks posed by exposure to radiation, while MRI has relatively low specificity and a long imaging time. In contrast, optical imaging has low spatial resolution and a small penetration depth. Meanwhile, US imaging has poor resolution and subjective results dependent on the particular operator. To overcome these limitations, researchers have attempted to fuse two or more different imaging techniques to create a new imaging mode, also known as multimodal molecular imaging, to obtain more consistent and accurate information.

Currently, most multimodal imaging techniques feature dual modes, involving optical imaging combined with MRI, PET, or SPECT, or PET and SPECT combined with CT or MRI, among others. Multimodal molecular imaging has been utilized in preclinical and clinical research for early diagnosis, disease staging, assessment of therapeutic response, surgical navigation, and prognosis evaluation.

In studies on the application of multimodal molecular imaging in predicting the efficacy of targeted therapy, the focus has mainly been placed on antiangiogenic drugs in preclinical studies. Using a novel metal-based imaging probe, Fe₃O₄-DMSA-SMCC-BCZM-^99mTc, with the monoclonal antibody bevacizumab radiolabeled with ^99mTc for dual-modality SPECT/MR imaging of angiogenesis by targeting VEGF-A was reported.³⁵³ This approach could be utilized to evaluate the efficacy of potential antiangiogenic drugs. Similar results were found for NIR830-bevacizumab-IONPs with MR and optical imaging.³⁵⁴ In a mouse model, Chen et al. detected cancer in vivo using multimodal imaging with photoacoustic and computed tomography (CT), which targeted epidermal growth factor receptors (EGFRs) and ErbB2 and may be used as a new platform for evaluating the response to EGFR- and HER2-targeted therapy.³⁵⁵

Conclusion and future perspectives

This paper reviews studies performed to date on the application of molecular imaging for early prediction of the efficacy of cancer-targeted therapy. Conventional predictive biomarkers for targeted therapy mainly rely on invasive tissue biopsy and subsequent pathological analysis. Since only a limited amount of tissue can be biopsied, as well as the issue of tumor heterogeneity, the conventional histopathological biomarkers obtained from a single lesion are not always predictive of the response to targeted therapy, especially for patients with multiple metastatic tumors. For post-therapy assessment of the efficacy of targeted treatment, the current gold standard is the use of the RECIST criteria, a method based on the change in the tumor size. However, such change often occurs over weeks, or even months, after treatment initiation, which rules out the possibility of detecting treatment resistance soon after it develops.

Against the above background, the novel molecular imaging technique demonstrates more potential than conventional imaging techniques for cancer-targeted therapy in the following aspects: (1) It provides whole-body imaging, such as by immuno-PET,¹⁶⁸ which better demonstrates the intra-tumoral and inter-tumoral heterogeneity of targeted molecule expression. This enables the prediction of targeted treatment efficacy at the lesion level.¹⁶⁸ (2) It enables the imaging of early changes (usually several days after treatment) in the functional status of a tumor, which could reflect the response to targeted therapy and enable early detection of treatment resistance and prediction of long-term efficacy. (3) It allows noninvasive monitoring of the changes in the expression of targeted molecules, which is essential for optimizing the treatment strategy during therapy.

To date, almost all studies reporting the application of molecular imaging in cancer-targeted therapy were in early-stage clinical trials with a small number of patients. Although these studies are essential for obtaining preliminary findings on the performance of these advanced imaging techniques for evaluating the efficacy of targeted treatment, their definitive value needs to be confirmed in future studies with larger sample sizes. There is also heterogeneity in the data acquisition among the applied approaches, such as the scanning protocol and timing of scanning, strategy for data analysis, and study endpoints. This makes it difficult to directly compare the data among studies and draw definitive conclusions on the performance of any technique or any parameter for evaluating the efficacy of targeted treatment. There is an urgent need to develop a uniform scanning protocol, data analysis strategy, and study endpoints for each imaging technique, as this is the only way to increase uniformity among studies. This would in turn allow direct data comparison and pooled data analysis and should eventually accelerate the clinical application of molecular imaging for evaluating the efficacy of targeted therapy.

At present, ¹⁸F-FDG PET-CT, DCE-MRI, and DWI-MRI are the most commonly used techniques for assessing cancer-targeted therapy in clinical practice. These techniques also hold great promise as they enable noninvasive whole-body imaging with deep tissue penetration and high sensitivity for detecting the biological changes after targeted treatment, such as changes in the metabolic activity, perfusion, and diffusion of tumors. Although immune-PET provides target-specific information on tumors and maybe a more helpful tool for evaluating the efficacy of targeted treatment. VEGF, HER2, EGFR, ER, and PD-1/PD-L1 are the most commonly studied targets in this field. Immuno-PET binding to these targets has been applied in early clinical trials to evaluate the anti-cancer targeted treatment efficacy and to predict patient prognosis. Early results demonstrated excellent potential translation of these tools to clinical practice. However, clinical translation of new radiotracers is difficult in many countries, including China, due to regulatory restrictions. Multimodality molecular imaging could be a promising approach to improve the predictive accuracy in evaluating the efficacy of targeted treatment.^{300,356,357,358} The development of this approach is an essential task in future work.

Along with the increasing use of molecular imaging in cancer management, artificial intelligence (AI) techniques have the potential for several related applications, such as automated interpretation of images and predicting treatment efficacy and patient survival. Information derived from AI may help select the most appropriate treatment for patients and thus improve patient care. An early attempt toward this goal has been reported by Mu and colleagues.³⁵⁹ They found that ¹⁸F-FDG PET-CT images could be used to predict the EGFR mutation status using the AI technique of deep learning. A higher EGFR deep learning score (EGFR-DLS) was associated with longer PFS in patients treated with EGFR-TKIs, and a shorter PFS in patients receiving immune checkpoint inhibitors.³⁵⁹ The utilization of AI to evaluate the efficacy of targeted treatment and guide clinical decision-making regarding appropriate treatment is expected to increase soon.

Targeted therapy has been one of the cornerstones in treating cancer patients for decades. The rapid development of molecular imaging should improve the early prediction of treatment response to allow early adaptation of patient management and eventually improve patient outcomes. From a perspective point of view, molecular imaging approaches will provide valuable tools to optimize the dosing schedule, determine therapeutic regimes, and monitor the therapeutic response to guide the change in treatment protocol timely if resistance happens. These are important in assisting the clinical decision-making of multidisciplinary treatment, including targeted therapy.