Agents described in the Molecular Imaging and Contrast Agent Database for imaging carbonic anhydrase IX expression

Abstract

Carbonic anhydrase IX (CA IX) is selectively expressed in a range of hypoxic tumours and is a validated endogenous hypoxia marker with prognostic significance; hence, CA IX is of great interest as a molecular imaging target in oncology. In this review, we present an overview of the different imaging agents and imaging modalities that have been applied for the in vivo detection of CA IX. The imaging agents reviewed are all entries in the Molecular Imaging and Contrast Agent Database (MICAD) and comprise antibody, antibody fragments and small molecule imaging agents. The effectiveness of these agents for imaging CA IX in vivo gave variable performance; however, a number of agents proved very capable. As molecular imaging has become indispensable in current medical practice we anticipate that the clinical significance of CA IX will see continued development and improvements in imaging agents for targeting this enzyme.

• Antibody
• hypoxia
• oncology
• positron emission tomography
• sulfonamide

Molecular Imaging and Contrast Agent Database

The Molecular Imaging and Contrast Agent Database (MICAD) is a collection of data comprising molecular imaging and contrast agents that have in vivo animal or human data published in peer-reviewed scientific journals1,2. MICAD is part of the National Institute of Health (NIH) Common Fund and aims to promote research using molecular imaging probes and to facilitate the translation of preclinical findings to clinical practice1–3. As of March 2013 there were ∼1400 agents listed in MICAD. Of this, the majority were for positron emission tomography (PET) (41%), followed by single-photon emission computed tomography (SPECT) (29%), optical imaging (14%), magnetic resonance imaging (MRI) (11%), ultrasound (2%), multimodal imaging (1%), X-ray/computer tomography (X-ray/CT) (1%) and photoacoustic imaging (1%).

Cancers are traditionally staged and monitored using invasive methods, for example, biopsies or exploratory surgery, characterizing treatment response by a change in tumour volume4. These methods are limited, however, as biochemical changes happen much faster than morphological ones. Molecular imaging allows a repeated in vivo measurement of several critical molecular features such that molecular imaging has become indispensable in current medical practice4. Of the MICAD entries, ten agents target carbonic anhydrase IX (CA IX) (Table 1). CA IX is overexpressed in several solid cancer types in hypoxic conditions, while expression in normal tissues is minimal. CA IX is thus considered a marker for tumour hypoxia as it provides differentiation between normal and hypoxic tissue, allowing the opportunity for selective imaging of solid tumours. This article presents the ten CA IX imaging agents in MICAD and outlines their effectiveness for imaging CA IX in vivo.

Table 1. Agents listed in the MICAD that target CA IX.

Entry	MICAD agent name	Agent short name	Category	Imaging modality	Tumour model
1	^124I-chimeric mAb G2505	^124I-cG250	Antibody	PET	Human clinical trial ccRCC6 SK-RC-52 human RCC xenograft7
2	^125I-mAb8,9	^125I-mAb	Antibody	SPECT	HT-29 human colorectal xenograft
3	^89Zr-N-Succinyldesferalchimeric mAb G25010,11	^89Zr-Df-cG250	Antibody	PET	SK-RC-52 human RCC xenograft
4	^111In-DOTA-cG25012,13	^111In-DOTA-cG250	Antibody	SPECT	HT-29 human colorectal xenograft SK-RC-38 human RCC cell line
5	^111In-DOTA-Fab-cG25012,14	^111In-DOTA-cG250-Fab	Antibody	SPECT	HT-29 human colorectal xenograft SK-RC-38 human RCC cell line
6	^111In-DOTA-F(ab′)2-cG25012,15	^111In-DOTA-cG250-F(ab′)2	Antibody	SPECT	HT-29 human colorectal xenograft SK-RC-38 human RCC cell line
7	^89Zr-labelled N-suc-Df-conjugated anti-CA IX chimeric mAb cG250-F(ab′)2 fragments16,17	^89Zr-Df-cG250-F(ab′)2	Antibody	PET	SCCNij3 human head and neck xenograft
8	^177Lu-benzyl-diethylenetriamine pentaacetic acid-anti CA IX SIP A318,19	^177Lu-DTPA-A3-SIP	Antibody	SPECT	LS174T human colorectal xenograft
9	[^99mTc](CO)3 N-(pyrindin-2-yl-methyl)-N[2-4-sulfamoylphenyl)-ethyl]aminoethylacetate20,21	^99mTc-PS	Small molecule	SPECT	HT-29 human colorectal xenograft
10	Hyperpolarized ^13C-labelled bicarbonate (H^13)22,23	^13C-labelled bicarbonate	Small molecule	MRI MRS	EL4 murine lymphoma xenograft

CA IX as a tumour marker

CA IX is a dimeric zinc metalloenzyme that catalyses the reversible hydration of CO₂ to and H⁺ to play an important role in pH homeostasis24. Each CA IX monomer is comprised of an intracellular domain, a single transmembrane helix domain, an extracellular proteoglycan domain and an extracellular catalytic domain25,26. CA IX is overexpressed in several solid cancer types, including cervical, clear cell renal cell, pancreatic, breast, bladder, head and neck, soft tissue and non-small cell lung carcinoma8,12. In solid tumours, hypoxia inducible factor 1 (HIF-1) regulates a signalling cascade involving ∼100 genes, including CA IX7, to enable adaptive cellular functions to support tumour cell survival, proliferation and metastasis in hypoxia27. The extracellular pH (pH_e) of tumour cells is lower than in normal tissue owing to elevated metabolism leading to increased production and export of acidic metabolites such as lactic acid22. At a molecular level, CA IX overexpression leads to dramatically increased production of and H⁺ at the extracellular surface of hypoxic tumour cells. The is actively transported into the cell to buffer intracellular pH (pH_i) while the H⁺ contributes to acidosis of the extracellular space28 to support invasion and metastasis27. The proteoglycan domain makes CA IX unique among CAs, and is proposed to assist CA IX to function efficiently at a more acidic pH25,26.

Solid tumour masses are heterogeneous and show significant variation in oxygen concentrations18,29. These heterogeneously distributed areas of hypoxia are referred to as the “hypoxic fraction” and arise from the insufficient organization of blood vessels within the solid tumours30. Hypoxic tumours have a high metastatic potential associated with a more aggressive phenotype, this results in a resistance to radio- and chemotherapy and therefore a poor outcome for the patient4,13,29,31,32. Gray et al.33 reported that tumour cells lacking oxygen were resistant to death via radiation by 2.5- to 3-fold compared to fully oxygenated cells29,33. These results were obtained separately by other groups34. Hypoxia has been established as a key feature that can significantly influence tumour behaviour and predict a response to therapy4,29,32. There are several invasive and non-invasive methods available for the detection of tumour hypoxia, but these are not fully reliable4,32. These techniques either yield a variable diagnosis or have functional limitations due to the incomplete penetration of the tumours or resistance to some tumour types13,18,32. For example, hypoxia can be determined post-mortem in animal models given pimonidazole by intravenous administration 30 min prior to sacrifice. Pimonidazole is a 2-nitroimidazole compound that is reduced in hypoxia, the reduced form can react with and stain intracellular macromolecules in hypoxic regions to provide generic, but selective hypoxia detection18,35. Co-localization analysis between pimonidazole staining and ex vivo CA IX expression detected by monoclonal antibodies (mAbs) to CA IX has yielded conflicting results, this has been attributed to different levels of oxygen at which pimonidazole is reduced or CA IX overexpressed, respectively; local temporal fluctuations in hypoxia level; or non-hypoxia-related factors that affect CA IX expression16. For in vivo hypoxia measurements, an invasive oxygen electrode can be used but this is not practical for all cancer types and causes significant patient discomfort. CA IX offers several features that make it a promising clinically relevant candidate for diagnostic imaging29,31. Firstly, the active site domain is extracellular and can be targeted by imaging agents that do not have to transverse the cell membrane29, and secondly, it is expressed in few normal tissues but is overexpressed to high concentrations in hypoxic conditions with a long half-life7,31. It has been postulated that while some cell lines may subvert known hypoxia mechanisms and that CA IX is not overexpressed in all cancer cell types, a vast array of human solid tumours typically overexpress CA IX7,31.

Antibody and antibody fragment imaging agents that target CA IX

G250 and cG250 mAbs and antibody fragments, A3 and CC7 mAbs

mAbs are a homogenous population of immunoglobulins (e.g. IgG) directed against a single epitope36. G250 is a mAb that was isolated >20 years ago and the antigen identified much later as an epitope of CA IX6,37,38. cG250 is a chimeric form of G250 that was developed to be less immunogenic than G250 while retaining the CA IX antigen specificity (K_a = 4 × 10⁹M⁻¹)6,7,10. Antibody fragments have several advantages over parent antibodies in the IgG format for molecular imaging owing to altered physicochemical properties. This includes more rapid clearance from the blood and normal tissues, reduced residence time, a reduction in the radiation absorbed by the bone marrow, improved tumour penetration – particularly into hypoxic tumour regions, and potentially reduced immunogenicity39. The increased clearance of fragments has several caveats as the absolute tumour uptake may be lower while non-specific uptake by other organs may be higher12,39. In 2009, Neri and colleagues18 reported the generation and characterization of two high-affinity human mAbs specific to extracellular CA IX, named A3 and CC7. These are the first fully human mAbs with high affinity (nanomolar) for CA IX, importantly they do not inhibit CA IX activity and specificity to CA IX over other forms of CA was also confirmed18. Due to limitations in antibody diffusion within solid tumour masses40–42, A3 and CC7 antibodies in the small immunoprotein (SIP) format were prepared with a lower molecular weight (∼76 kDa) cf. A3 and CC7 when in the IgG format (∼150 kDa). Researchers have undertaken work using these mAbs and mAb fragments conjugated to a range of clinically used radionuclides for imaging CA IX using PET and SPECT (Table 2).

Table 2. PET and SPECT radionuclides combined with antibodies and antibody fragments for imaging CA IX.

Radionuclide	Half-life (days)	Reference
^124I	4.16	6,10
^125I	59.40	43
^131I	8.02	49
^111In	2.80	43
^89Zr	3.26	10
^177Lu	6.73	49
^99mTc	0.25	20

Directly radiolabelled CA IX antibodies

With direct radiolabelling the mAb structure remains virtually unchanged and retains high specificity for the target antigen. Direct iodination of mAbs with ¹³¹I, ¹²⁴I and ¹²⁵I is relatively easy to perform and provides high specific activities45. It is typically achieved via the iodogen (1,3,4,6-tetrachloro-3α,6α-diphenyl glycoluril) method, although other methods are available7,10,46–50. The radioactive iodine is provided via an aqueous solution of NaI47. Iodogen, a strong oxidising agent, is used to convert the I⁻ anion to an electrophilic I⁺ cation, this electrophile then reacts to insert an iodine ortho to the hydroxyl groups of tyrosine residues of the antibody (Figure 1)48. The iodogen reagent also limits oxidative damage to the protein during radiolabelling47.

Figure 1. Direct iodination of antibodies is typically achieved via the iodogen method to modify a tyrosine residue47,48.

¹³¹I has a higher intensity gamma ray when compared to ¹²⁴I and ¹²⁵I, it is also widely available, relatively inexpensive and was the first to be established as a radiolabelling agent for cG25044,45,49. ¹³¹I was however readily released from the tumour cells following internalization of ¹³¹I-cG250 and the abundant high energy gamma photons could affect normal tissues such as the bone marrow, limiting the dose of ¹³¹I-cG250 that could be used therapeutically49,51. Additionally, the high energy single-photon emission compromises image quality6. With these drawbacks, research moved to the direct iodination of cG250 using ¹²⁴I (Table 1, entry 1) and ¹²⁵I (Table 1, entry 2)6. ¹²⁴I, used for PET, has a half-life of ∼4 days, this allows both long term pharmacokinetics to be determined and for centralized production and wide distribution of ¹²⁴I radiolabelled antibody6. ¹²⁵, used for SPECT, I is considered a pure gamma emitter with a low radiation energy and medium range half-life of ∼60 days, whereas ¹³¹I is both a beta and gamma emitter with medium range energy and a short range half-life of ∼8 days52. ¹²⁵I gives a higher counting efficiency than ¹³¹I such that about half the amount of ¹²⁵I is needed for tracer purposes52.

Renal cell carcinoma (RCC) is the most common malignancy in the kidney and each year in USA ∼39 000 people are diagnosed with and ∼13 000 people die from RCC44. Approximately 80% of RCCs are of the more aggressive clear cell type (ccRCC), of these >95% express the G250/CA IX antigen on the cell surface6,7,37,39,44. In RCC, it has been shown that production of CA IX begins after the loss of the tumour suppressor gene VHL (Von Hippel–Lindau protein), whose product is integral to the HIF-1α pathway7,12,18. This upregulation of CA IX has been determined to be one of the most prominent accessible markers of RCC18. There is debate within the literature regarding the mechanism leading to CA IX overexpression in RCC, irrespective of mechanism, however, the high cell surface production of CA IX renders RCC an excellent model for investigating the properties of CA IX imaging agents.

¹²⁴I-cG250

Divgi et al.6 carried out the first phase I trial with patients with ccRCC using ¹²⁴I-cG250 (Table 1, entry 1)5 pre-operatively to predict ccRCC, the only human clinical trial in this review. The aim of this research was to determine if PET with ¹²⁴I-cG250 was able to discern ccRCC from other renal masses. For this clinical trial, 25 patients who were surgical candidates for kidney tumour resection, accrued via one surgeon, were administered with 185 MBq/10 mg of ¹²⁴I-cG250 over 20 min. PET and CT images were taken 3 h later with positive PET images being indicative of ccRCC. Both PET and CT modalities were used to unequivocally determine tumour location and structure. Out of the 25 patients imaged pre-surgery, 15 gave positive PET images. Following imaging, surgery determined that these 15 patients all had ccRCC. Of the 10 patients with negative PET images there was only one patient surgically determined to have ccRCC.

Animal model studies with ¹²⁴I-cG250 have been undertaken by Lawrentschuk et al.7 The purpose of this investigation was to identify if there was direct data that linked hypoxia and CA IX expression in RCC. They provided two hypotheses based on the loss of the VHL: (i) that CA IX will reflect hypoxia or (ii) that tumours subvert this pathway with CA IX expression unrelated to hypoxia. SK-RC-52 human RCC xenograft models were established in nude mice. At Day 0, 1, 3, 7 and 10 invasive oxygen electrodes were used to investigate intratumoural hypoxia directly before injection with 5 MBq of ¹²⁴I-cG250 (25 μg) in 0.2 mL phosphate-buffered saline followed by PET imaging. The PET results showed localization of ¹²⁴I-cG250 in the tumour as well as the cardiac blood pool. After imaging, the animals were sacrificed and their tumours and organs removed, weighed and biodistribution determined. To investigate CA IX expression, one mouse was taken at 1, 2, 3 and 5 day time points and autoradiography, histology and immunohistochemistry performed. They found no significant correlation between any oxygen parameter and the CA IX expression and concluded that in this model, the level of CA IX expression could not be clearly related to hypoxia with only a weak correlation between hypoxia and the cG250 uptake7.

¹²⁵I-mAb

Zhang and colleagues8 have described a ¹²⁵I-labelled mAb (¹²⁵I-mAb) (Table 1, entry 2)9 that recognized CA IX without cross-reactivity to other human CAs. A HT-29 xenograft model was established in athymic male mice and an in vivo binding study, biodistribution study and planar imaging were carried out with ¹²⁵I-mAb. To determine in vivo binding, the mice were injected with 3.70 MBq of ¹²⁵I-mAb and sacrificed 24 h later. Their blood was collected, the unbound ¹²⁵I-mAb removed and the activity quantified using a gamma counter. Radioactivity in the blood was high: the radioactivity was 5.5% that of the blood activity at 24 h post-injection (p.i.) which the authors attributed to a small amount of the tracer diffusing into blood cells. For the biodistribution study, 1.295 MBq of ¹²⁵I-mAb was injected intravenously into mice and the radioactivity of both normal tissues and tumour measured using a gamma counter at a range of time points8. Uptake of ¹²⁵I-mAb by the tumour was significantly higher than for normal tissues, 4.9 ± 1.2 %ID/g (percentage of the injected dose per gram of tissue) at 24 h p.i.8 SPECT imaging, mice were injected with 5.55 MBq of ¹²⁵I-mAb and after 48 h planar static images (duration 30 s) were acquired using a gamma camera. In parallel, a control study was also undertaken using a 100-fold molar excess of unlabelled mAb injected into the mouse 24 h prior to injection with ¹²⁵I-mAb. The results showed that the ¹²⁵I-mAb preferentially concentrated in the HT-29 tumour at 48 h p.i., however, it also accumulated in the thyroid, while in the control study unlabelled mAb blocked accumulation of ¹²⁵I-mAb in both the tumour and the thyroid. Although the signal to noise ratio was promising, the high levels of radioactivity in the blood make ¹²⁵I-mAb currently unsuitable for as a diagnostic tool8.

Indirectly radiolabelled CA IX antibodies and antibody fragments

Direct radiolabelling of antibodies with most other radionuclides is not possible and instead an indirect approach with a host-linker moiety is needed. In this indirect labelling method, the host coordinates to the radionuclide while the linker allows covalent attachment to the antibody or antibody fragment with minimal disruption to structure. A summary of host-linkers and their method of attachment to the CA IX antibody component are shown in Figure 2. Brouwers et al.10 (Table 1, entry 7) and Hoeben et al.16 (Table 1, entry 3) used a desferrioxamine (Df) backbone as host linker for ⁸⁹Zr incorporation. Carlin et al.12 employed an amide linkage to 1,4,7,10-tetraazacyclododecane-1,4,7,10-tetraacetic acid (DOTA) for ¹¹¹In incorporation (Table 1, entries 4, 5 and 6). Finally, Neri et al.18 conjugate diethylenetriaminepentaacetic acid (DTPA) via a thiourea linker for radiolabelling with ¹⁷⁷Lu (Table 1, entry 8).

Figure 2. Host-linker moieties that have been employed to indirectly radiolabel antibodies and antibody fragments targeting CA IX.

⁸⁹Zr-Df-cG250

To synthesize ⁸⁹Zr-Df-cG250 (Table 1, entry 3)11, Brouwers et al.10 conjugated the desferrioxamine (as the mesylate salt) host to cG250 via amide bond formation followed by radiolabelling with 165 MBq of ⁸⁹Zr (Figure 3). ⁸⁹Zr was selected as radionuclide because its half-life (3.26 days) is similar to the clearance time of the cG25010. PET imaging with ⁸⁹Zr-Df-cG250 was used in combination with radioimmunoscintigraphy (RIS) in a SK-RC-52 xenograft rat model10. Results were compared both to ¹⁸F-FDG (¹⁸F-fluorodeoxyglucose), the gold standard imaging agent for oncology with PET, and to ¹¹¹In-DTPA-cG250 for RIS10. Eight rats were injected with 20 MBq of ⁸⁹Zr-Df-cG250 and PET imaging undertaken at 5 min, 24, 48 and 72 h p.i. After imaging, the animals were sacrificed and their organs removed, weighed and the activity of the samples calculated. The PET images showed that ¹⁸F-FDG did not accumulate in the tumours, sufficiently for visualization, while at 24 h p.i. ⁸⁹Zr-Df-cG250 successfully labelled even small (∼100 mg) tumours. The ⁸⁹Zr-Df-cG250 PET images improved over time with optimal images taken at 72 h p.i.

Figure 3. General synthetic methodology for the production of radiolabelled mAbs using desferrioxamine mesylate (desferal, Df) as the host-linker moiety54.

The comparison of biodistribution acquired with ⁸⁹Zr-Df-cG250 to those acquired with ¹¹¹In-DTPA-cG250, allowed researchers to conclude that the tumour uptake of ⁸⁹Zr-Df-cG250 was identical to ¹¹¹In-DTPA-cG250 (5.0 ± 2.4 and 4.9 ± 2.9%ID/g, respectively) as was the blood levels at the 72 h time point (1.4 ± 0.4 and 1.7 ± 0.7%ID/g, respectively). The uptake of both antibodies was similar in other organs assessed, however, the authors state that in some cases, the PET images produced using ⁸⁹Zr-Df-cG250 were highly superior to the ¹¹¹In-Df-cG250 images and therefore this agent may be of value in the planning an monitoring of RCC patients10. ¹¹¹In-DTPA-cG250 is not listed as a CA IX targeting agent in MICAD, however, the synthesis and imaging properties in patients with RCC metastases has been reported and showed improved visualization and a higher activity compared with ¹³¹I-cG25045. Unlike ¹³¹I, ¹¹¹In is a residualizing radionuclide, that is, one that is retained inside a cell as the radioactive catabolites are trapped in the lysosomes45,53. Based on increasing evidence that cG250 can be internalized by CA IX expressing RCC the use of residualizing radionuclides was suggested as beneficial. Previous studies had shown that labelling cG250 with the residualizing nuclides ⁸⁸Y and ¹⁷⁷Lu had ∼3–4 times the activity of ¹²⁵I-cG250 at 7 days p.i.45 Due to the similarity between the RIS results for ¹¹¹In-DTPA-cG250 and ⁸⁹Zr-Df-cG250 in tumour uptake, blood level and also a similarity amongst other tissues, the authors suggest that ⁸⁹Zr could be re-classed as a residualizing radionuclide53.

¹¹¹In-DOTA-cG250, ¹¹¹In-DOTA-cG250-Fab and ¹¹¹In-DOTA-cG250-F(ab′)₂

To evaluate the potential of cG250 fragments as imaging agents, Carlin et al.12 compared the binding affinity for the CA IX antigen of two low molecular weight fragments of cG250, cG250-Fab and cG250-F(ab′)₂, with the parent cG250 (Table 1, entries 4, 5 and 6)13–15. The fragments were prepared via enzymatic degradation of the intact cG250 using the hydrolytic enzymes pepsin and papain to give cG250-F(ab′)₂ and cG250-Fab, respectively12,39. The F(ab′)₂ fragment is bivalent (two antigen binding sites), whereas the Fab fragment is monovalent (one antigen binding site)39. cG250, cG250-Fab and cG250-F(ab′)₂ were each conjugated to DOTA and then complexed to the residualizing radionuclide ¹¹¹In (Figure 4)55. The conjugation occurred in two steps. Firstly, N-hydroxysuccinimide (NHS) and 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide (EDC) are used to synthesize the active monoester of DOTA. Secondly, the DOTA active ester is linked to the mAb or fragment via amide bond formation through a side chain lysine residue on the mAb55. The stoichiometry of incorporated DOTA moieties, binding affinity (K_d) for CA IX expressed by SK-RC-38 cells and the saturation binding data (B_max) of the mAb and fragments varied (Table 3).

Figure 4. Synthetic approach for the production of radiolabelled mAbs and mAb fragments using DOTA as the host-linker moiety.

Table 3. Properties of DOTA-CA IX mAb and mAb fragment conjugates12.

mAb/fragment	Stoichiometry(molecules of DOTA per mAb)	Bmax (binding sites/cell)	Kd (nM)
^111In-DOTA-cG250	5.2 ± 0.2	120 000 ± 22 000	2.5
^111In-DOTA-cG250-Fab	1.9 ± 0.1	118 000 ± 21 000	14
^111In-DOTA-cG250-F(ab′)2	4.1 ± 0.2	114 000 ± 19 000	1.8

Athymic male mice with HT-29 tumours were injected with 1.6 MBq of each ¹¹¹In-DOTA-cG250, ¹¹¹In-DOTA-cG250-Fab and ¹¹¹In-DOTA-cG250-F(ab′)₂ for in vivo biodistribution studies. ¹¹¹In-DOTA-cG250 (tested at 2, 4 and 7 days p.i) showed a high uptake of radioactivity into the HT-29 tumour. At 2 days p.i. uptake (20.1 ± 4.8% ID/g) was similar to that obtained with two non-specific ¹¹¹In-labelled control antibodies, the specificity of the ¹¹¹In-DOTA-cG250 was however confirmed over the next 5 days as the tumour uptake increased, while uptake of the control antibodies decreased. Compared to ¹¹¹In-DOTA-cG250, the radiolabelled fragments were tested at 6 and 24 h p.i due to faster clearance and showed a lower absolute tumour uptake. For F(ab′)₂ uptake was 7.6 ± 1.4 and 9.3 ± 2.1%ID/g at 6 and 24 h p.i., respectively, while for Fab uptake was lower again (3.6 ± 1.3 and 3.5 ± 1.7%ID/g at 6 and 24 h p.i., respectively) due to a much faster clearance from circulation. Biodistribution was similar for all three ¹¹¹In-DOTA-antibodies across other organs. The tumour/blood (T/B) and tumour/muscle (T/M) values were compared for all three agents (Table 4). Both the F(ab′)₂ and the Fab fragments showed an improvement in the T/B and T/M ratios between 6 and 24 h p.i. with the T/B ratio for the Fab fragment increasing dramatically from 2.8 to 16.6, this ratio is higher than for the ¹¹¹In-DOTA-cG250 at 7 days p.i. (Table 4).

Table 4. T/B and T/M ratios for ¹¹¹In-DOTA-cG250, ¹¹¹In-DOTA-cG250-Fab and ¹¹¹In-DOTA-cG250-F(ab′)₂ in mice with HT-29 tumours12.

mAb/fragment	Time p.i.	T/B ratio	T/M ratio
^111In-DOTA-cG250	7 days	6.6	69
^111In-DOTA-cG250-Fab	24 h	16.6	6.7
^111In-DOTA-cG250-F(ab′)2	24 h	4.6	8.9

To correlate ¹¹¹In-DOTA-cG250 to regions of perfusion, hypoxia and CA IX expression, autoradiography images were compared to fluorescent images of the same tumour section stained with Hoechst 33342, pimonidazole and a CA IX fluorescent antibody at 2 days p.i. of ¹¹¹In-DOTA-cG250. The Hoescht stain revealed high perfusion around the rim of the tumour and disorganized perfusion at the centre, the anti-correlation of the endogenous CA IX. The regions of CA IX staining co-localized with pimidazole, showing that the CA IX expression was limited to the hypoxic regions but more importantly for this study, it was shown that although there was some targeting of CA IX with ¹¹¹In-DOTA-cG250 close to the blood vessels, the uptake of ¹¹¹In was not directly proportional to the endogenous CA IX. At 4 and 7 days p.i., the correlation of ¹¹¹In-DOTA-cG250 to CA IX improved.

Similarly, the ¹¹¹In-DOTA-fragment autoradiography images were compared to the Hoescht, pimidazole and endogenous CA IX fluorescent images at 6 and 24 h p.i. At 6 h p.i. the correlation of ¹¹¹In-F(ab′)₂-cG250 to CA IX is comparable to that of the intact cG250 after 2 days p.i. At 24 h, the ¹¹¹In matches almost perfectly to the endogenous CA IX image, demonstrating excellent targeting. Additionally, except for the later data points, there is almost an inverse correlation with Hoechst staining12. Due to the antibody fragments clearing faster into the blood, likewise, diffusion into the tumour is faster, allowing more rapid correlation with CA IX expression. The immunohistochemistry results for the F(ab′)₂ and Fab fragments showed a diffuse uptake in the core of the tumour, where the CA IX production is high. The Fab fragment had the fastest blood clearance and from digital analysis of the images, the authors concluded that the Fab fragment was the most rapid and reliable indicator for CA IX expression in hypoxic tumour regions. However, despite this conclusion, due to the lower absolute tumour uptake for both the fragments, and additional problems posed by the non-specific tracer uptake and poor perfusion, the authors recommend using the intact ¹¹¹In-DOTA-G250 at 7 days p.i. as this gave the most sensitive and accurate detection of CA IX12. The late images of the ¹¹¹In-DOTA-cG250 were comparable to the early images of the fragments but the higher tumour/non-tumour contrast of ¹¹¹In-DOTA-cG250 would reduce imaging times.

⁸⁹Zr-Df-cG250-F(ab′)₂

Hoeben et al.16 investigated Df-cG250-F(ab′)₂ fragments using ⁸⁹Zr as the radionuclide. They selected ⁸⁹Zr as its half-life (3.26 days) matched the pharmacokinetics of the fragment (Table 1, entry 7). The labelling chemistry for Df-cG250-F(ab′)₂ with ⁸⁹Zr to generate ⁸⁹Zr-Df-cG250-F(ab′)₂17 is similar to that presented in Figure 316. A cohort of BALB nu/nu mice subcutaneously implanted with SCCNij3 cell xenograft tumours, a head and neck tumour model, were used to qualitatively determine the intratumoural distribution of radiolabelled cG250 via autoradiography and compared to ¹¹¹In-DTPA-cG250. The preparation of ¹¹¹In-DTPA-cG250 is described by Troost et al.56 It was found that the autoradiography uptake signal of ⁸⁹Zr-Df-cG250-F(ab′)₂ was very similar to that of ¹¹¹In-DTPA-cG250 with all correlations over two time points p < 0.0001. The uptake of ¹¹¹In-DTPA-cG250 was much higher than ⁸⁹Zr-Df-cG250-F(ab′)₂ (25.2 ± 6.2 versus 2.68 ± 1.30%ID/g), however, this was expected due to the longer circulatory half-life of IgG16. That being said the ⁸⁹Zr-Df-cG250-F(ab′)₂ showed a faster accumulation in CA IX rich regions at 4 h p.i.

Ten mice were injected with 6.85 ± 0.55 MBq of ⁸⁹Zr-Df-cG250-F(ab′)₂, eight mice (two groups of n = 4) were examined via a small animal PET imaging scanner at 4 and 24 h p.i. At 30 min prior to PET imaging, mice were injected with 80 mg/kg of pimonidazole (hypoxia marker). One minute before sacrifice, mice were injected with Hoechst 33342 (perfusion marker). The tumours were clearly visible in the PET images at both 4 and 24 h p.i. The radioactive uptake of the ⁸⁹Zr-Df-cG250-F(ab′)₂ was compared to the positive area of the tumour stained for CA IX and pimonidazole in tumour sections. The standardized uptake value maximum (SUV_max) of ⁸⁹Zr-Df-cG250-F(ab′)₂ for tumour, muscle and tissue was determined (Table 5). The kidneys and liver displayed high uptake of ⁸⁹Zr-Df-cG250-F(ab′)₂, however, this is a common side effect of using F(ab′)₂ fragments: it is probable that the fragment can be reabsorbed, which may cause problems for imaging in the future.

Table 5. The SUV_max of ⁸⁹Zr-Df-cG250-F(ab′)₂ and %ID/g of tissue taken from BALB nu/nu mice subcutaneously implanted with SCCNij3 cell xenograft tumours at 4 and 24 h p.i.16.

	4 h p.i.		24 h p.i.
Tissue type	SUVmax	%ID/g	SUVmax	%ID/g
Tumour	1.65 ± 0.26	3.71 ± 0.97	0.57 ± 0.32	1.66 ± 0.48
Muscles	0.38 ± 0.05	–	0.14 ± 0.00	–
Blood	–	4.34 ± 0.27	–	0.20 ± 0.06

A significant correlation (p < 0.0001, r = 0.57–0.74) between the CA IX staining and radioactivity uptake was observed at both time points16. Similarly, there was a significant correlation (p < 0.0001, r = 0.46–0.68) between pimonidazole (hypoxia) staining and tracer uptake. A strong positive correlation (p = 0.0067, r = 0.93) was also observed between SUV_max on the PET images and the measured tumour uptake found from the analysis of tissue (%ID/g), this correlation is similar to that found by Lawrentschuk et al7. for ¹²⁴I-cG250 in a SK-RC-52 model. This study concluded that ⁸⁹Zr-Df-cG250-F(ab′)₂ could be useful in determining CA IX-positive hypoxic areas for head and neck carcinomas via demonstrating a significant spatial correlation of ⁸⁹Zr-Df-cG250-F(ab′)₂ and CA IX expression, however, the authors state that further research is required.

¹⁷⁷Lu-DTPA-A3-SIP and ¹⁷⁷Lu-DTPA-CC7-SIP

Radiolabelled A3-SIP and CC7-SIP antibodies were prepared by covalent modification of a side chain lysine residue of the SIP by reaction with p-isothiocyanate benzyl DTPA (p-SCN-Bn-DTPA) to form a thiourea link, followed by coordination of ¹⁷⁷Lu radionuclide to the DTPA host (Figure 5)18,57. Mass spectrometry analysis showed a stoichiometry of ∼1.8 DTPA molecules per antibody molecule18. The biodistribution of ¹⁷⁷Lu-DTPA-A3-SIP and ¹⁷⁷Lu-DTPA-CC7-SIP (0.259–0.407 MBq) (Table 1, entry 8)19 was evaluated following injection into LS174T colorectal xenograft models (n = 4) and compared to ¹⁷⁷Lu-DTPA-SIP-L19 (a clinical stage antibody that targets neo-vasculature structures). The mice were sacrificed 24 h after SIP administration, the organs removed, weighed and the radioactivity counted. Very high kidney and high liver uptake values was observed for both ¹⁷⁷Lu-DTPA-A3-SIP and ¹⁷⁷Lu-DTPA-CC7-SIP. ¹⁷⁷Lu-DTPA-A3-SIP showed a ∼4-fold lower tumour accumulation compared to ¹⁷⁷Lu-DTPA-SIP-L19 accumulation (2.4 versus 9.3 %ID/g, respectively), however, had an improved T/B ratio (16.7 versus 5.8, respectively). Biodistribution data for ¹⁷⁷Lu-DTPA-CC7-SIP was not presented. This study concluded that simultaneous use of vascular-targeting and hypoxia-targeting antibodies may be desirable to achieve a more homogenous distribution and detection of tumours.

Figure 5. Synthetic approach for the production of radiolabelled antibodies SIP format using p-SCN-Bn-DTPA as the host-linker moiety.

Small molecule imaging agents that target CA IX

Two of the ten agents in MICAD that target CA IX are small molecules.

^99mTc-PS

Almost all reported CA small molecule ligands comprise a zinc binding group in their structure of which the primary sulfonamide moiety (−SO₂NH₂) is the premier example. The primary sulfonamide anion coordinates to the active site Zn²⁺ of CAs, acting as an “anchor” on to which variable “tail” groups, selected to incorporate desirable properties, can be tethered54. An important finding by Dubois et al.30 was that the binding of sulfonamide based ligands to cell surface CA IX occurs only under hypoxic conditions. Akurathi et al.20 synthesized a ^99mTc-labelled small molecule sulfonamide compound, ^99mTc-PS (Table 1, entry 9)21, with the aim of imaging CA IX expression to determine regional tumour hypoxia using SPECT. ^99mTc is the most widely used radionuclide for diagnostic nuclear medicine (>90%)58. It can be generated at low cost, it has a suitable half-life (6 h) and its gamma ray emission energy is not accompanied by beta emissions, allowing a more precise alignment of imaging detectors58,59. Additionally, its emission energy is 141 keV, enough to be detected by gamma cameras but low enough to not carry out significant damage to the body20,58. 4-(2-Aminoethyl)benzenesulfonamide (AEBS) was chosen as the core structure of ^99mTc-PS as it has a high affinity for CA IX (K_i = 33 nM) and the other cancer associated CA (CA XII, K_i = 3.2 nM) but low affinity for off-target cytoplasmic CA I and CA II and membrane-bound CA IV (K_i = 21 μM, 160 and 2.4 μM, respectively). ^99mTc requires a chelating agent, so [^99mTc(H₂O)₃(CO)₃]⁺ (prepared using an Isolink™ kit) together with N-2-picolyl-N-acetic acid were used in the synthesis of ^99mTc-PS as the [^99mTc(H₂O)₃(CO)₃]⁺ can exchange the three water molecules with the amine, aromatic N-heterocycle and carboxylate donor of N-2-picolyl-N-acetic acid, while the latter is also able to covalently attach to the AEBS small molecule through amide bond formation (Figure 6). A rhenium analogue of this sulfonamide ligand, Re–PS, was synthesized as a “cold” control compound for ^99mTc-PS to enable the indirect study of the affinity of ^99mTc-PS for CA I, II, IX and XII under in vitro conditions20. The in vitro inhibition data for Re–PS was compared to AEBS, it is worthwhile to note that the K_i of Re–PS for CA IX is similar to that for AEBS while it is more potent than AEBS for CA XII.

Figure 6. Synthesis of the small molecule CA IX imaging agent ^99mTc-PS and control compound Re–PS.

The stability of ^99mTc-PS in whole blood and plasma of mice was analysed at 1 h p.i. In plasma 65% of the radioactivity was in intact ^99mTc-PS while 35% was present as a polar radiometabolite. Biodistribution studies with ^99mTc-PS were then carried out in mice bearing HT-29 colorectal tumours at 0.5, 1, 2 or 4 h p.i. Accumulation was initially highest in the liver and it was proposed that this may be due to an active transport mechanism on the liver surface. The tracer is moderately hydrophilic resulting in liver clearance, hence at 2 and 4 h p.i. ∼18% ID/g was found in the intestines, indicating clearance to the hepatobiliary pathway. The tracer was also found initially in a significant quantity in the kidneys (36–42 %ID/g at 0.5 and 1 h p.i.), an indication that the tracer may be binding to CA XII or XIV, which are highly expressed on the membrane of the kidneys. Control experiments using AEBS as a blocking agent were used to evaluate this theory and it was concluded that the tracer uptake in the kidneys was not related to interaction with CA XII or XIV. A minimal amount of ^99mTc-PS was shown to diffuse into red blood cells (0.1 %ID/g at 1 h p.i.) and is assumed to bind to either red blood cell CA I or CA II20. Disappointingly, the tumour uptake of ^99mTc-PS was low at each of the four tested time points (0.1 ± 0.1 %ID/g at 0.5 h then 0.0 ± 0.0 thereafter).

¹³C-labelled bicarbonate

Gallagher et al.22 designed hyperpolarized H¹³ as a MRI imaging agent to investigate the redox properties of CA, potentially including CA IX, but not specifically for CA IX (Table 1, entry 10). Hyperpolarized H¹³ was developed on the basis of dynamic nuclear polarization (DNP), a technique that increases the sensitivity of solution-state magnetic resonance spectroscopy (MRS). By injection of hyperpolarized H¹³, their aim was to produce a high resolution pH map of a transverse slice through an animal that may offer a new way of detecting disease and monitoring treatment response. pH changes can indicate inflammation, ischaemia or renal disease amongst other pathological conditions, however, there is currently no clinical tool to measure in vivo pH. Using the Henderson–Hasselbalch equation, the tissue pH can be determined from the measurements of the concentration ratio between H¹³ and ¹³CO_2, assuming that the pKa of bicarbonate is constant at 6.1723. The replacement of ¹²C with ¹³C does not alter the biochemistry of the molecule while it provides a >10 000-fold enhancement of the ¹³C MRS signals thus allowing changes in pH to be monitored in vivo23.

After validation of the MRI method for pH determination, hyperpolarized H¹³ (200 μL, 40 mg/kg) was injected into three groups of mice with EL4 murine lymphoma tumours. Group 1 were untreated controls, group 2 were given sodium bicarbonate in the drinking water to increase the tissue pH and group 3 were given an ammonium chloride gavage to decrease the tissue pH22. For comparison, several mice were injected with the pH sensitive molecule 3-aminopropylphosphonate (3-APP) so that ³¹P spectra could be obtained for comparison. 3-APP is a non-toxic MRS-visible extracellular marker that is relatively unaffected by temperature or ionic strength, the chemical shift frequency of 3-APP is sensitive to pH allowing the determination of pH in vivo60. ¹³C and ³¹P MRS imaging at 9.4 T were performed22. The spatial distribution of ¹³CO₂ and H¹³ in the transverse slice was determined. It was shown that the highest concentration of ¹³CO₂ was in the tumour mass compared to H¹³ which was highest in the aorta. Although minimal differences were shown in signal intensity between muscle/tumour tissue, a pH map was successfully calculated and a low pH value for the tumour mass was shown, demonstrating that it was possible to image tissue pH noninvasively in vivo. Comparing the untreated mice (tumour pH 6.71), NaHCO₃ treated mice (tumour pH 7.02) with NH₄Cl (tumour pH 6.47) suggests that the H¹³ measurement reflects more of the pH_e. The authors suggest this could be due to the acquisition of spectra immediately after H¹³ is injected and that normally there is a higher concentration of bicarbonate in the extracellular environment. A limitation of the method was also highlighted, namely small differences in the rate at which H¹³ and ¹³CO₂ lose polarization will lead to the calculated pH varying over time. The T₁’s were found to be 10.1 ± 2.9 s (n = 9) for H¹³ and 9.8 ± 2.5 s for ¹³CO₂ (n = 7) with the presence of CA increasing this exchange.

Properties of antibodies and small molecules as imaging agents

Although antibodies provide excellent specificity towards their antigen, there are several factors that make tumour targeting with antibodies difficult. Solid tumours are heterogeneous masses and a common limitation of antibody-based imaging agents is slow or incomplete tissue distribution, especially to areas with poor vasculature31 as the drugs cannot completely penetrate the tissue61. The contributing factors to this include molecular size (IgG ∼ 150 KDa), the antigen barrier and the high tumour interstitial pressure. Additionally, slow plasma clearance maintains high drug levels which complicate imaging and therapy61 and present a high radiation dose risk to the bone marrow13. Rudnick and Adams have discussed in depth the interaction between the affinity, avidity and efficacy of antibodies in tumour targeting62. Although not the focus of this article, the critical parameters they specify that affect a binding site are: antigen density, mAb internalization and metabolism and the mAb binding affinity62.

Antibody fragments may provide advantages over the intact antibodies. Their reduced molecular size (e.g. cG250-F(ab′)₂ fragments ∼100 kDa compared to intact cG250 ∼150 kDa) may aid in tissue penetration and their reduced clearance time from the blood can result in higher tumour uptake compared to normal tissues. However, fragments may have increased renal uptake as they are reabsorbed by the renal tubular cells and may lead to radiation-induced nephrotoxicity16,39. Similarly, scFv (single-chain variable) fragments of ∼25 kDa may have excellent tumour penetration but their rapid clearance from the blood may be advantageous or disadvantageous. Other disadvantages include that they are monovalent and there is a possibility of aggregation63,64. SIPs consist of two scFv fragments held via a hinge, making them bivalent, and their size of ∼80 kDa means that they have a slower clearance time63,64. They have an increased total affinity and have reduced immunogenicity63,64. However, they have no effector functions and can be difficult to generate as the synthesis involves several steps18,63.

Aromatic sulfonamides are the classical small molecule CA inhibitors that bind to the active site of CA65. The CA active sites are extremely tolerant to diverse and variable structures appended to the aromatic sulfonamide moiety66. It is possible to vary considerably the size, shape, charge, polarity, hydrophobicity etc. and still have nanomolar CA inhibition. This tolerance allows finetuning of the biopharmaceutical and toxicological properties of the inhibitor via structural manipulations66. Additionally, small molecules may be advantageous in their cost and ease of manufacture as well as their improved ADME (adsorption, distribution, metabolism and excretion) and DMPK (drug metabolism and pharmacokinetics) properties when compared to antibodies66. However, sensitivity, degradation, toxicity and clearance may all be considerably problematic with these compounds as well as specificity for CA IX over other CAs. To overcome these challenges the small molecule has to be designed very carefully to take the limitations into account.

Imaging via PET is the most common imaging modality used with the agents described here. PET is advantageous in that it allows the absolute quantitation of tracer uptake unlike determination of uptake via histological sections which is highly subjective12. As well as allowing the possibility of serial imaging, it is possible to image the whole tumour volume, instead of biopsy or tissue specimens12. PET displays a much higher sensitivity than SPECT (∼2–3 times the order of magnitude). Using SPECT, however, has the potential to increase the observational time window due to the longer half-life of the emitters used67. MRI has superior soft tissue contrast resolution, multiplanar image acquisition and functional imaging capabilities68. Of note is that only one of these agents has been designed for MRI. With the growth of MRI in hospitals and clinics, it is likely that more agents will be developed to use this modality which provides high resolution and superior soft tissue contrast. However, the ultimate deciding factor in the choice of imaging modality is the biological considerations of the system67. Rahmim and Zaidi67 warn that in the biological context, generalized comparisons between modalities are not appropriate and instead they need to be performed on a case-by-case basis so that the correct imaging modality is selected.

The development of agents for in vivo imaging of CA IX extends well beyond those agents listed in MICAD. Recent examples include small molecule ¹⁸F labelled PET agents69, and an extensive comparison of ¹²⁴I-cG250 with ⁸⁹Zr-Df-cG250 in murine models with CA IX expressing ccRCC zenografts (SK-RC-52 or NU-12) using PET70. The biodistribution results demonstrated very high tumour uptake. In the NU-12 tumour model uptake of the two agents was similar (48.7 ± 15.2%ID/g versus 32.0 ± 22.9%ID/g, respectively, p = 0.257), while in the SK-RC-52 tumour model the uptake of ⁸⁹Zr-Df-cG250 was significantly higher than ¹²⁴I-cG250 (114.7 ± 25.2%ID/g versus 38.2 ± 18.3%ID/g, respectively, p = 0.029). This high uptake gave excellent PET image contrast, for example small intraperitoneal lesions (7 mm³) were visualized with ⁸⁹Zr-Df-cG250 and were readily discriminated from the liver and spleen. The results from this study suggest that PET in RCC patients may be more sensitive using ⁸⁹Zr-Df-cG250 instead of ¹²⁴I-cG25070.

Conclusions

Of the MICAD entries, ten agents target CA IX, an antigen overexpressed in several solid cancer types in hypoxic conditions with expression in normal tissues is minimal. CA IX is a challenging diagnostic target for non-invasive molecular imaging, especially given the similarities in the active site for all 12 catalytically active human CAs. It is however an extremely important target as the detection and monitoring of hypoxic tumours can inform clinical decisions surrounding cancer therapy. Continued development of CA IX imaging agents will ideally balance patient safety, good tumour penetration and accurate localization of hypoxia. The ten MICAD agents described herein, highlight the scope and remaining opportunity for development of improved agents targeting CA IX imaging in oncology.

Declaration of interest

The Australian Research Council (Grant Nos DP110100071 and FF110100185 to S.-A.P.), Griffith University (Scholarship to D.S.) and the (Australian) National Imaging Facility provided financial support.

The authors report no conflicts of interest. The authors alone are responsible for the content and writing of this article.