PET imaging with radiolabeled antibodies and tyrosine kinase inhibitors: immuno-PET and TKI-PET

Currently, 12 mAbs have been approved by the FDA for the treatment of cancer, all being intact mAbs [1]. Seven of the mAbs have been approved for the treatment of hematological malignancies, being rituximab, gemtuzumab ozogamicin, alemtuzumab, ibritumumab tiuxetan, tositumomab, ofatumumab, and brentuximab vedotin. Five mAbs have been approved for the therapy of solid tumors, and four of them interfere with signal transduction pathways by targeting growth factors or the extracellular domain of their receptors. Those mAbs comprise trastuzumab for the treatment of metastatic breast cancer; cetuximab, bevacizumab, and panitumumab for the treatment of colorectal cancer; and cetuximab and bevacizumab for the treatment of head and neck and non-small cell lung cancer. The fifth mAb, ipilumumab, has an immunostimulatory effect via cytotoxic T lymphocyte-associated antigen 4 (CTLA-4) directed against melanoma. Most naked mAbs can also act via other effector mechanisms than described above such as antibody-dependent cellular cytotoxicity, complement-dependent cellular cytotoxicity, or apoptosis induction. However, naked mAbs have limited efficacy on their own and should preferably be used in combination with chemo- or radiotherapy. Alternatively, mAbs can be loaded with toxic payloads like the radionuclides yttrium-90 or iodine-131 as in the case of ibritumumab tiuxetan and tositumomab, respectively, or with super toxic drugs as in the case of gemtuzumab ozogamycin and brentuximab vedotin. The use of supertoxic drugs is becoming increasingly popular, as illustrated by the approval of gemtuzumab ozogamycin and brentuximab vedotin (containing calicheamicin and auristatin as the supertoxic drug, respectively) and the development of the next generation anti-human epidermal growth factor receptor 2 (HER2) therapeutics such as trastuzumab-DM1 (trastuzumab coupled to the supertoxic drug mertansine) [7]. However, for highly toxic conjugates, selective tumor targeting is a must. Cross-reactivity of such supertoxic conjugates with normal tissues might result in unacceptable toxicity, as was recently demonstrated for the anti-CD44v6 conjugate bivatuzumab-DM1 [8].

In contrast to mAbs, TKIs are capable of entering the tumor cell where they compete for adenosine triphosphate (ATP) binding sites of transmembrane receptor tyrosine kinases, resulting in inhibition of signaling pathways. TKIs like gefitinib, erlotinib, and vemurafanib are monospecific and target just one tyrosine kinase, in this case epidermal growth factor receptor (EGFR), while all other FDA-approved TKIs are dual- or multispecific (see Table 1). Next to reversible TKIs, more recently, also irreversible TKIs are under investigation like the EGFR and HER2 inhibitors afatinib and neratinib [9, 10]. In contrast to mAbs, TKIs are fast kinetic molecules without immune-mediated effector functions and are not suitable for the delivery of toxic payloads to tumors.

The ability of PET to quantitatively image the distribution of radiolabeled drugs within the body makes this technique a valuable tool at several stages of drug development and application. From first-in-man clinical trials with new drugs, it is important to learn about the ideal drug dosing for optimal tumor targeting (e.g., saturation of receptors), the uptake in critical normal organs to anticipate toxicity, and the interpatient variation in pharmacokinetics and tumor targeting. Drug imaging might provide this information in an efficient and safe way with fewer patients treated at suboptimal doses. Pretreatment imaging with the drug of interest might also have added value for patient selection because it can be used to assess target expression and drug accumulation in all tumor lesions and normal tissues noninvasively, quantitatively, and even over time. This information might be particularly relevant for heterogeneous tumor types or when targeted drugs are combined with other treatment modalities like chemo- and radiotherapy, to find routes of maximum synergism. Ideally, anatomical information on tumor extension is obtained, like possibly with PET-CT and PET-MRI, to enable the assessment of homogeneity of tumor drug accumulation. Also, imaging during therapy is attractive to show that tumor targeting is effective and indeed results in antitumor effects as can be assessed by e.g., ¹⁸F-fluorodeoxyglucose PET. If a targeted drug is not effective in a particular patient, adaptive treatment can be considered by dose escalation or by choosing targeted drugs that inhibit compensatory pathways. Apart from applications in treatment planning, treatment monitoring, and response monitoring, imaging with radiolabeled targeted drugs like mAbs can also be used for diagnostic purposes and for better understanding in vivo biology (“immunohistochemistry in vivo”).

To enable visualization of a targeted drug with a PET camera, the drug should be labeled with a positron emitter in an inert way, i.e., that neither binding nor pharmacokinetic characteristics of the drug become altered. Moreover, the physical half-life of the positron emitter should be compatible with the residence time of the targeted drug in the body, which is typically several days for slow kinetic intact mAbs and a couple of hours for the fast kinetic small molecules like TKIs. Due to their large size of 150 kDa, it is quite easy to radiolabel mAbs for PET imaging in an inert way. Very recently, universal procedures were introduced for radiolabeling of intact mAbs with the long-lived positron emitters iodine-124 (¹²⁴I, t _½ = 100.3 h) [11] and zirconium-89 (⁸⁹Zr, t _½ = 78.4 h) [12], of which ⁸⁹Zr is particularly suitable in combination with internalizing mAbs. For radiolabeling of mAb fragments, which are more rapidly cleared from the body than intact mAbs, short-lived positron emitters became available like gallium-68 (t _½ = 1.13 h), copper-64 (t _½ = 12.7 h), yttrium-86 (t _½ = 14.7 h), and bromine-76 (t _½ = 16.2 h). Of these, the short-lived positron emitter gallium-68 (⁶⁸Ga) is of particular clinical interest because it can be obtained from a commercially available long life-span ⁶⁸Ge/⁶⁸Ga-generator (half-life 271 days), making it continuously available even for centers without a cyclotron and at reasonable costs. For some comprehensive reviews on positron emitters for immuno-PET, see references [13‐15].

Due to the fact that most therapeutic mAbs are internalizing intact mAbs, ⁸⁹Zr nowadays becomes most broadly applied. To make this happen, crucial achievements have been obtained in the production [12, 16, 17] and commercial availability of ⁸⁹Zr for clinical use [12], the development of a chelate for facile and stable coupling of ⁸⁹Zr to mAbs [18‐20], and the development of protocols for labeling of mAbs with ⁸⁹Zr in a current good manufacturing practice (cGMP) compliant way [19]. In Europe, it takes 3–4 months of preparation before a ⁸⁹Zr-labeled mAb can be evaluated in a clinical trial, including all the steps of setting up radiochemistry and quality controls in a cGMP compliant way, performing productions for validation of procedures, and taking regulatory hurdles.

During the past years, several preclinical and clinical ¹²⁴I- and ⁸⁹Zr-immuno-PET studies have been performed with intact mAbs, for example, with mAbs U36 (anti-CD44v6) [21], DN30 (anti-cMet) [22], G250 (anticarbonic anhydrase IX) [23‐25], L19-SIP (antifibronectin) [11], R1507 (anti-IGF-1R) [26], J591 and 7E11 (antiprostate-specific membrane antigen (PSMA)) [27, 28], TCR105 (anti-CD105) [29], cetuximab (anti-EGFR) [30, 31], inbritumumab tiuxetan (anti-CD20) [32], rituximab (anti-CD20) [33], bevacizumab (anti-vascular endothelial growth factor (VEGF)) [34, 35], and trastuzumab (anti-HER2) [36‐38].

In contrast to the radiolabeling of mAbs, radiolabeling of small molecules (<1 kDa) like TKIs is much more challenging and requires a drug-specific labeling strategy. In many cases, labeling with ¹¹C (t _½ = 20 min) will appear possible, while in some cases, the chemical structure of the drug will allow labeling with ¹⁸F (t _½ = 110 min). To increase chemical flexibility, radiolabeled TKI analogs are also used for PET imaging [5, 39]. However, to be sure that a PET tracer exactly represents the TKI of interest and to avoid extra regulatory hurdles, it is better to use labeling strategies that end up with exactly the original TKI chemical structure in which a cold carbon or fluorine atom has been substituted by ¹¹C or ¹⁸F, respectively.

While mAbs are normally metabolically stable in the blood circulation and therefore radioactivity uptake in organs will represent mAb uptake, this can be different for radiolabeled TKIs. In case of TKIs, radioactive metabolites might be formed, which do not represent the biodistribution of the original TKI. The arising deviations can be depending on the position of the radioactive atom in the TKI. To this end, it might be interesting to label a TKI at different positions if possible and to select the most suitable candidate tracer. Nevertheless, before starting clinical trials with radiolabeled TKIs, the evaluation of their in vivo stability and the identification of metabolites are crucial, especially to make the quantification of TKI biodistribution possible.

Noninvasive in vivo quantification of TKIs in TKI-PET studies requires tracer kinetic models that describe uptake, retention, and clearance of the tracer in tissue. For this, the input function, i.e., the time course of the tracer in arterial plasma, should be known. Routinely, the time course of radioactivity in the whole blood and plasma is measured continuously using an online withdrawal and detection system. However, plasma usually contains labeled metabolites. The proper input function for the kinetic model is the metabolite-corrected plasma curve, and therefore at set times, additional discrete samples (usually seven during a 60-min scan) are collected to correct for labeled metabolites.

During the past years, procedures for radiolabeling of several FDA-approved anticancer TKIs have been described, including ¹¹C-imatinib [40], ¹¹C-gefitinib [41‐43], ¹⁸F-gefitinib [44, 45], ¹⁸F-sunitinib [46], ¹¹C-erlotinib [47], ¹⁸F-lapatinib [48], and ¹¹C-sorafenib [49]; however, only few of these tracers have been evaluated in animal models and just one in a clinical trial of which the results have not been published yet.

Preclinical immuno-PET has been used for different kinds of applications. Many studies have been focusing on the quantitative evaluation of mAbs for their capacity to target tumors selectively and on the assessment of the expression of critical tumor targets like CD20 [32, 33], CD105 [29], CD44v6 [21], cMet [22], PSMA [27, 28], IGF-1R [26], EGFR [30, 31], HER2 [36‐38], and VEGF [34, 35]. In several of these studies, immuno-PET was evaluated for its suitability as scouting procedure prior to radioimmunotherapy to predict biodistribution and dosimetry of the corresponding therapeutic radioimmunoconjugates. For this purpose, the diagnostic radioimmunoconjugate should show a similar biodistribution as the therapeutic radioimmunoconjugate [50]. It was shown that ⁸⁹Zr-ibritumumab tiuxetan is suitable for the prediction of ⁹⁰Y-ibritumumab tiuxetan biodistribution [32], ⁸⁹Zr-cetuximab for ⁹⁰Y-cetuximab and ¹⁷⁷Lu-cetuximab biodistribution [30], and ¹²⁴I-L19-SIP for ¹³¹I-L19-SIP biodistribution [11]. In other studies, ⁸⁹Zr-immuno-PET was evaluated for imaging and quantification of “therapeutic effect sensors” [6]. These are proteins whose expression is modulated in response to the functional inhibition of a drug target. For example, HER2 and VEGF are known to be downregulated upon Hsp90 inhibition, and their downregulation could indeed be visualized in vivo by the use of PET with ⁸⁹Zr-trastuzumab [38] and ⁸⁹Zr-bevacizumab [35], respectively. Therapeutic response upon chemo-, hormone-, and radiotherapy was also evaluated in another ⁸⁹Zr-immuno-PET approach. Ruggiero et al. [28] used in their studies ⁸⁹Zr-7E11, a novel mAb directed against an intracellular epitope of PSMA. In vivo ⁸⁹Zr-7E11 can only bind to PSMA in case dead or dying cells are present. The authors showed that ⁸⁹Zr-7E11 can be used to monitor and quantify, with high specificity, tumor response upon therapy in PSMA-positive prostate tumors. In another approach, ⁸⁹Zr-immuno-PET appeared suitable for imaging of tumor hypoxia. To this end, ⁸⁹Zr-cG250 was used for detection of the membrane protein CAIX, which is involved in pH regulation and is upregulated in many tumors in case of hypoxia [25].

The disadvantage of using intact mAbs for confirmation of target expression with immuno-PET is their long residence time ranging from days to over weeks and the relatively low tumor-to-nontumor ratios at early time points after injection. In addition, the large size of a mAb molecule on one hand limits its diffusion from the vasculature into tumor, while on the other hand results in unspecific tumor uptake due to an enhanced permeability and retention effect. For this purpose, the use of mAb fragments or engineered proteins like diabodies and nanobodies might be an option. These fast kinetic antibody-based PET probes allow same-day imaging at low radiation burden for patients, to confirm target expression, as recently described by the group of Wu [51, 52].

Some recent landmark studies also illustrate the potential of immuno-PET in clinical settings. ¹²⁴I-immuno-PET was evaluated for in vivo profiling of renal cancer. Divgi et al. [24] used ¹²⁴I-cG250 to predict the presence of clear-cell renal carcinomas in 25 patients scheduled for surgical tumor resection. It might be informative to identify those renal cancer patients who have this aggressive type because of treatment decisions. Of the 16 clear-cell patients, 15 were identified accurately by immuno-PET, and all 9 nonclear-cell renal masses were negative for the tracer. This study illustrates how molecular imaging can contribute to personalized medicine.

The first clinical ⁸⁹Zr-immuno-PET study ever was reported in 2006 by Börjesson et al. [21]. This feasibility study with 20 head and neck cancer patients showed that ⁸⁹Zr-cmAb, directed against CD44v6, can be safely applied in patients and that it is a promising tool for PET detection of primary tumors as well as of metastases in the neck (Fig. 1). In this study, biodistribution, radiation dose, and quantification potential of immuno-PET were also assessed [53]. PET quantification of blood-pool activity in the left ventricle was in good agreement with sampled blood activity, except for heavy weight patients (>100 kg). The same accounts for the uptake in tumor tissue, where a good agreement was observed between PET-derived data and biopsy data. This suggests that patients with high and low mAb uptake can be differentiated, which might be important for the selection of patients with the highest chance of benefit from mAb therapy.

High quality images were obtained by Dijkers et al. [37] in an immuno-PET study with ⁸⁹Zr-trastuzumab in breast cancer patients. In this feasibility study with 14 patients, three different dose cohorts were evaluated: 10 or 50 mg for trastuzumab-naïve patients and 10 mg for patients on trastuzumab treatment. It was proven that the latter two performed equally. Although this study was not aiming for the comparison with conventional staging modalities or for assessing specificity and sensitivity, lesions with ⁸⁹Zr-trastuzumab uptake were generally in good agreement with CT, MRI, and bone scans. PET images showed a high spatial resolution and a good signal-to-noise ratio, which resulted in an image quality unapproachable by previous ¹¹¹In-trastuzumab SPECT scans. Excellent visualization of mAb uptake in HER2-positive lesions as well as in metastatic liver, lung, bone, and even brain HER2-positive lesions was observed (Fig. 2). ⁸⁹Zr-trastuzumab PET allowed the quantification of conjugate uptake in HER2-positive lesions, and it became clear that for some patients with extensive tumor load, no HER2 saturation occurred during trastuzumab therapy.

PET imaging might also contribute to better understanding of TKI activity, although preclinical in vivo proof of concept studies are scarce and thus far limited to ¹¹C-gefitinib, ¹⁸F-gefitinib, ¹¹C-erlotinib, and ¹¹C-sorafenib. The most appealing results have been obtained with erlotinib and gefitinib, which compete with ATP for the ATP-binding site on the EGFR, thereby preventing signal transduction leading to proliferation. Erlotinib and gefitinib can induce dramatic clinical responses but only in 4–10% of HNSCC patients and 10–15% of non-small-cell lung cancer (NSCLC) patients, when used as single agents [54]. Expression and mutation status of the EGFR have been associated with increased response [54]. It has been hypothesized that the presence of sensitizing mutations might increase the binding of the drug with its target. This might result in better drug retention within the tumor as well as in a more efficient inhibition of signaling through EGFR. However, for the assessment of EGFR expression and mutation status, a tumor biopsy has to be taken, which is not always possible like in NSCLC. Even when a biopsy is available, it is questionable whether this is sufficient to obtain a representative overview of the whole (often heterogeneous) tumor. Moreover, it is possible that expression and mutation status differ in primary tumor and metastatic lesions and change during the course of disease, for example, upon chemo- or radiotherapy. Taking this into account, it might be that PET imaging with the radiolabeled EGFR TKI itself gives a more comprehensive overview of EGFR receptor status and the interaction of the drug with this receptor. To test this possibility, Memon et al. [47] evaluated the uptake of ¹¹C-erlotinib in nude mice bearing lung cancer xenograft lines with a different sensitivity to erlotinib treatment and a different mutation status. In mice carrying the most sensitive xenograft line, a xenograft line with a mutation in EGFR, tumor uptake of ¹¹C-erlotinib was the highest, indicating that ¹¹C-erlotinib PET can indeed identify erlotinib sensitive tumors (Fig. 3).