Since one century, a dramatic technological and scientific revolution has occurred in the field of cancer diagnosis. The concept of targeted diagnosis is closely associated to the concept of targeted therapy. This concept of targeted strategy was the dogma of Paul Ehrlich [1] and the magic bullet quest remains a living paradigm for modern cancer research. For over 70 years, many generations of medical oncologists have tried cancer immunotherapy to diagnose or fight cancer and the immune system stimulation was always very disappointing [2]. In 1978, functional imaging has given great hopes when for the first time David Goldenberg published promising results using radiolabeled antibodies to carcinoembryonic antigen (CEA) for the detection of diverse cancers by scintigraphy. Murine polyclonal antibodies against CEA or tumor-associated glycoprotein-72 (TAG-72) were produced, radiolabeled and used for imaging diagnosis [3, 4]. As the detection was achieved with poor resolution cameras, immunoimaging in cancer detection was limited to tumors up to 15 mm (Fig. 1). The exceptional advances in chemistry, immunology and genetic research as well in functional imaging have actually sped up cancer tracer research leading to molecular fusion imaging, the natural evolution of functional imaging.

Fig. 1
figure 1

Radioimmunoscintigraphy (1987) with intact murine B72.3 mAb to TAG-72 labeled with indium 111. Woman treated for an ovarian cancer presenting with an increasing level of CA125 tumoral marker and a totally normal staging. Detection performed 48 h after IV infusion of 111 MBq of radioconjugate in SPET mode with a FBP without attenuation correction. Transaxial (a), sagittal (b) and coronal (c) section on the head exhibit an abnormal mAb uptake related to a brain metastasis of the primary tumor

Full size image

Radiolabeled antibodies for molecular imaging

The very first-generation of antibodies used as imaging agents was derived from immunogenic mouse polyclonal antibodies. Rapidly, monoclonal antibodies (mAbs) were a substitute for polyclonal antibodies as well chimeric or humanized antibodies have become a substitute for murine antibodies, strongly immunogenic [5]. Human antibodies were obtained from a patient with a colonic cancer and participating in a trial of active specific immunotherapy. Fortunately, humanization of murine antibodies has become routine and totally human antibodies can be produced by immunization of mice genetically engineered to carry germ line human immunoglobulin genes. Recent progress in biology and genetics have provided extensive information on tumor cell particularly on surface receptors, ligands, adhesion molecules, proteases and proteins responsible for differentiation and activation, all information which represent potential targets for molecular imaging with antibodies [6]. However, as antibodies are directed against cell surface targets, their use for immunoimaging is restricted to the assessment of cell surface markers. Nevertheless the clinical efficacy of mAbs such as trastuzumab (anti-Her2/neu) or rituximab (anti-CD20) is a proof for the ability of cell surface target antibodies to block critical biologic targets or signaling pathways in patients. They can be potentially used as immunoimaging agents.

Radio labeled intact antibodies images have to be taken 2 or 3 days after the infusion to enhance the contrast between the tumoral uptake and the circulating activity (tumor to background ration) and require radionuclide with half-life (from several hours to days). In order to have a faster blood clearance and to enhance image quality, Fab, Fab′ or F(ab′)2 fragments obtained by enzymatic digestion of antibodies was suggested. These fragments radiolabeled with radionuclides like technetium-99m were more suitable for scintigraphy (Fig. 2).

Fig. 2
figure 2

Radioimmunoscintigraphy (1990) with F(ab′)2 fragment of murine mAb to CEA labeled with technetium 99m. Man treated for a colonic cancer presenting with an increasing level of CEA tumoral marker and a totally normal staging. Detection performed 12 h after IV infusion of 185 MBq of radioconjugate in SPET mode with a FBP without attenuation correction. Transaxial (a), sagittal (b) and coronal (c) section on the abdomen exhibit an abnormal mAb uptake (white arrow) related to a liver metastasis of the primary tumor

Full size image

Research and development of antibodies for molecular imaging have led to impressive progress. The removal of the Fc region was really determinant resulting in an accelerated blood clearance and a smaller size (molecular weight under 60 kDa). Single-chain variable fragments (scFvs) became the basis for engineered antibody fragments that retain full binding specificity. Dimers, trimers and multimers of scFv (diabodies, triabodies, and tetrabodies and tandem diabodies) were built [7, 8]. Minibodies resulting from the fusion of scFvs to additional antibody domains such as the closely related small immunoproteins and the larger scFv–Fc fusion proteins were also engineered [9]. All these engineered antibodies typically retain the high affinity and specificity of the parental antibody and their faster blood clearance is directly format-dependent. Diabodies and bivalent minibodies exhibit terminal half-lives in the range of 3–8 h with tumoral uptake levels ranging from 7 to 20 % of the injected dose per gram of tumor in murine models. These properties allow high-contrast single photon emission tomography (SPET) or positron emission tomography (PET) images within a few hours after tracer injection [10]. Recently a high affinity, monovalent Her2-specific affibodies had demonstrated an impressive tumoral targeting in preclinical models [8]. As compared to the results obtained with radiolabeled intact antibodies and Fab or F(ab′)2 fragments, immunoimaging is moving in a new era where display technologies (phage, yeast or ribosome) will be used for identifying human antibodies according to the desired binding properties and small proteins engineering based on fibronectin domains, or protein A domains (Affibodies; Affibody AB) will provide binding molecules for molecular targeting [9, 11].

Radionuclides for immunoimaging

Since the beginning of radioimmunoimaging, iodine-131, indium-111, iodine-123 and technetium-99m were mainly used for scintigraphy (planar and/or SPET imaging). Several years ago, we have carried out a prospective trial on 31 consecutive patients presenting with a breast cancer [12]. A breast radioimmunolymphoscintigraphy (RILS) using a human mAb (LiLo-16.88 labeled with indium-111) reacting with a cytoplasmic antigen (CTAA 16.88) was performed prior to surgery (Fig. 3). Planar and SPET images were successively obtained at 4, 24 and 48 h after a peri-areolar subcutaneous injection. 370 lymph nodes were assessed by ex vivo scintigraphy and correlated to pathology and immunohistochemistry. The positive predictive value for CTAA 16.88 positive nodes including both metastatic and hyperplastic nodes was 90 % in stages 0–IIB. Four positive nodes with follicular hyperplasia contained micrometastases. The negative predictive value was 93 %. These results were suggesting that RILS may be clinically useful and selectively limit the extent of the surgical procedure to positive lymph nodes. This study was achieved using a gamma camera with a limited resolution and failed to detect small abnormal lymph nodes under a 15 mm diameter. With the new generation of hybrid systems including SPET and PET/CT (computed tomography) or MRI gathering molecular and anatomical imaging and resulting in a dramatic enhancement in image quality, it should be of interest to compare the results obtained with non specific tracers like colloids to specific radiolabeled antibodies in the sentinel node detection. The advantage of hybrid gamma camera was demonstrated in a pilot study in refractory Hodgkin disease using a rabbit mAb against ferritin labeled with indium-111 for imaging and yttrium-90 for therapy [13]. Images were more precise and a relative quantification of the mAb uptake was possible. Today PET/CT scanners using 2-deoxy-2[18F]-fluoro-d-deoxyglucose (18FDG) are widely used in nuclear medicine departments and time for the development of novel molecular PET tracers has come. Among positron-emitting radionuclides several of them are of interest: copper-64 (64Cu-12.7h), yttrium-86 (86Y-14.7h), bromine-76 (76Br-16.0h) for engineered antibodies, zirconium-89 (89Zr-78.4h) and iodine-124 (124I-100.2h) for intact antibodies. A next step for molecular imaging is to evaluate target expression and avidity before an antibody therapy. This is becoming possible with quantitative immuno-PET and the availability of radionuclide pairs, such as 124I/131I, 64Cu/67Cu and 86Y/90Y which can be used either for imaging or therapy purposes. A significant impact upon the patients management of 18FDG-PET imaging-based strategies is now demonstrated [14, 15]. Therefore, the development and selection of the antibodies with highest affinity and specificity is critical to its efficacy in oncology may be based on immuno-PET imaging [16]. A large number of preclinical and clinical studies are in progress. A good example is the work achieved by the group of the Memorial Sloan-Kettering Cancer Center. Since the late 1980s, they have developed a specific antibody G250 which recognizes the tumor-associated protein carbonic anhydrase IX expressed in clear cell renal cancer. This antibody was labeled with iodine-131 and used for radioimmunodetection and radioimmunotherapy [17]. No reliable quantification was possible. Recently, the same group has developed a chimeric antibody cG250, which can be labeled with iodine-124 [18]. PET/CT was used to detect and quantify 26 renal masses before surgery. Immunochemistry, autoradiography and gamma-counting of tumor tissue were performed on surgical specimens. All patients with a clear cell renal cancer were positive. A significant correlation between tumor uptake estimated by in vivo and in vitro methods (p = 0.000001) demonstrated that PET quantitative parameters provide quantitative pharmacokinetic data [19].

Fig. 3
figure 3

RILS (1991) with intact totally human mAb-16.88 to CTAA labeled with indium 111. Woman with a breast cancer planned for mastectomy and lymph node dissection (LN). Detection performed 4, 24 and 48 h after periareolar infusion of 111 MBq of radioconjugate in planar and SPET mode with a FBP without attenuation correction. a Planar detection: slow lymphatic drainage of the radioconjugate and visualization of positive lymph nodes from 24 h and after. b SPET detection: visualization of four spots corresponding to an abnormal uptake in the lymph nodes. c Ex vivo scintigraphy: positive tumoral uptake and detection of seven abnormal lymph nodes (six metastases)

Full size image

Progress in image quality

The lack of sensitivity of planar and SPET detectors were giving images with poor resolution and attenuation correction was suggested to enhance image quality. In the 1990s, attenuation correction was performed with radionuclide-based transmission systems using germanium-68 or cesium-137 (total or segmentary). Since the development of hybrid system, attenuation map are now calculated from the CT avoiding unnecessary dose delivering and result in a sharpest contrast with a higher spatial resolution. Iterative reconstruction algorithms using the ordered subsets expectation maximization (OSEM) technique has replaced the filtered back projection (FBP) previously used. The comparison between the FBP using a Butterworth filter (5; 0.3), iteration without (IRNC) and with attenuation correction (IRAC) was performed on 25 patients [20]. Three different set of images were obtained for each reconstruction software. FBP was too noisy generating a lot of false positive results. IRAC allowed to obtain the best image quality demonstrating some small abnormal foci not detected by the FBP method. This is confirming the benefits of using CT images for attenuation correction of SPET or PET data. Due to the significant improvement in image quality, OSEM has replaced all other analytic methods. Advancements in computing power and algorithm design have resulted in new and more accurate models being utilized for reconstruction algorithms. This includes improvements in components of the system model based on the detector geometry [21], fully 3D scatter estimates [22], and low noise random estimates [23]. A new algorithm, the SharpIR® algorithm (GE Healthcare), has been developed to improve the vue point high definition algorithm (3D OSEM IR). The PET detector response which is depending on sampling width, detector geometry and co linearity effects was included as a function of position in the scanner. By incorporating this information into the PET iterative reconstruction software, the contrast was significantly enhanced and a significant reduction in noise was obtained [24]. Images obtained in patients with the SharpIR® algorithm are significantly better and more precise (Fig. 4).

Fig. 4
figure 4

PET/CT scan in the staging of left breast cancer. Detection performed 1 h after the IV infusion of 185 MBq of 18F-FDG. a Iterative reconstruction using the OSEM technique with attenuation correction: visualization of axillary lymph nodes. b Reconstruction using the SharpIR® algorithm. Images of axillary lymph nodes are more precise

Full size image

Conclusion

Since 1928, when the Aschheim–Zondek test [25] was developed for either the diagnosis of pregnancy or cancer of the testis, progresses in biology (immunoassays, genetic and biopathology) have been impressive. A large number of new biomarkers have been identified and they are widely used to characterize tumors. It is well known that tumor markers can be detected in the blood a long time before lesion detection however very few of them are used for therapeutic decision or as an evaluable target to assess the therapeutic efficacy. On the same way, the spatial resolution and sensitivity of molecular imaging has also progressed with new detectors and new software. In spite of a better spatial resolution, anatomical imaging modalities fails to disclose small tumour lesions that have not yet caused morphologic alterations or to characterize them. Molecular imaging using radionuclides is challenging to detect lesions earlier before their anatomical visualization, and particularly with hybrid systems that provide clinically relevant information on a synergic way. The association of molecular biology and hybrid imaging is opening a new era as potential molecular imaging biomarkers in oncologic applications. Fused images are determinant in one out of four patients (24.6 %) particularly to dissociate physiological from pathological activities. In clinical oncology fused images are commonly used and may induce change in therapeutic choice in approximately 50 %. Moreover an occult metastatic disease suspected on the basis of a significant elevation of a tumour marker is demonstrated in 87.1 % of all cases and can be treated earlier. The association of biomarkers and molecular imaging is our next future. Targeted therapies with mAbs are widely used in oncology with efficacy for a small percentage of patients. The development of engineered antibody-based imaging agents is a stimulating challenge to verify targeting, to establish pharmacokinetics and to predict their clinical performance. Looking behind, the significant results obtained using engineered fragments labeled with radionuclides for SPET imaging are a relevant basis for the development of immuno-PET. With a wide range of positron-emitting radionuclides and a large number of engineered antibodies the use of immuno-PET with labeled cell surface biomarkers is quickly growing in preclinical and clinical studies [26]. In oncological applications, immuno-PET can play an important role at diagnosis, staging, for treatment selection, monitoring and during follow-up.