carcinoembryonic antigen: phase I/Il study with comparative biodistribution of intact and F(ab') antibodies

'Cancer Research Cwapaign Clinical Research Laboratory, Department of Clinical Oncology, Royal Free Hospital School of Medicine, London NW3 2PF, UK; 2Department of Cliniical Oncology, Uniersity Cotlege Londn Medical School, Riding House Street, London WIN 8AA, UK; 3Department of Radiotherapy, Charing Cross Hospital, Fulham Palace Road, London W6, UK; 4Department of Medical Physics, Royal Free Hospital School of Medicine, London NW3 2PF, UK.

Radioimmunotherapy (RIT) uses an antibody delivery system to target a tumour site with radiation. A number of beta-emitting radionuclides may be conjugated to antibody for this treatment. Iodine 131 (131') is valable for development of RIT because it has a beta emission of moderate energy capable of killing tumour cells over a range of up to 40 cell diameters and a gamma emission of energy 364 keV, which can be imaged with a gamma camera for quantification of biodistribution.
While it has been relatively easy to perform biodistribution studies in experimntal animals, there has not until now been a satisfactory way to obtain quantitative information about antibody distribution in man. We have developed a method for accurate quantification of 13I distribution which incorporates corrections for Compton scatter and attenuation (Green et al., 1990). We have used this method to compare two antibody preparations in the clinical development of RIT. RIT had produced good response rates in radiosensitive tumours such as lymphoma when large amounts of radionuclide are given (Kaminski et al., 1993;Press et al., 1993). However, common epitheial tumours such as colorectal and breast carcinoma have not yet been treated so successfully because of their greater radioresistance, which diminishes the therpeutic ratio. In spite of this, responses to therapy have been reported , and it is likely that moderate increases in efficency in delivery of antibodymediated delivery of radiation could establish radioimmunotherapy as a useful form of therapy for metastatic colorectal carcinoma.
Intact IgG antibodies with a molcular weight (MW) of 150 kilodaltons (kDa) may not penetrate well from blood through endothelium and extravascular tissues to the tumour (Yokota et al., 1992). It is proposed that F(ab% antibodies (MW 100 kDa) will achieve more effective penetration because of their smaler molcular size and that this will significantly improve the prospects for effective radioimmunotherapy of colorectal cancer. Previous studies in animal tumour models of colorectal carinoma have shown that F(ab% antibodies labelled with 13'I give superior tumour to blood ratios than is achieved in therapy with intact antibody (Wahl et al., 1983;Buchegger et al., 1990;Pedley et al., 1993). This hypothesis has been investigated in man by comparing the F(ab% fiagment with the intact version of the same antibody for RHT in patients with colorectal carcinoma.

PadeUs a h
Patients Ten patients with a raised serum CEA were given repeated injections of '31I-labelled anti_CEA intact IgG (A5B7) (serum CEA up to 622 sg 1-', median 117.5). The next nine patients were given '311-labelled anti-CEA A5B7 fragment F(ab'k (serum CEA up to 390ggl-1, median 79). To suppress the immune response to mouse IgG, cyclosporin A was given to all patients (Lermnann et al., 1988). All patients had unresectable, locally recurrent or metastatic tumours and performance status 0-2 (WHO 1979 criteria) and gave written, informed consent. The study was approved by ethics committee and covered by ARSAC licence.
The serum level of human anti-mouse IgG antibody was negative before therapy, assayed as described previously (Ledermann et al., 1988). A full blood count and renal, liver and thyroid function tests were performed at regular intervals. All patients had a negative intradermal test with loig of antibody prior to therapy. The thyroid was blocked with potassium iodide 180mg given orally 8 hourly for 14 days and potassium perchlorate 200mg 6 hourly for 4 days. Details of patients are given in Table . Anti-CEA antibody The mouse MAb (ASB7) was raised against CEA (Harwood et al., 1986). It was purified from supernatant culture by protein A chromatography and shown to be free from ag-  (555) RTx, radiotherapy; CTx, chemodtrapy, RIT, radioimmunotherapy.
gregates by fast protein liquid chromatography (FPLC) (Ledermann et al., 1988). Antibody production and preclinical toxicology were performed in accordance with the CRC Operation Manual (1986). The F(ab% fragment was produced from the intact MAb A5B7 anti-CEA by digestion with pepsin (Lamoyi & Nisonoff, 1983) and presented in a sterile (50mM phosphate) buffer and purified by protein A and gel filtration. Radioiodination was performed by the N-bromosuccinamide method (Adam, 1989). This method results in a labelling efficiency of 88-94% without loss of immunoreactivity.
The specific activity of radiolabelling was 0.11-0.19GBq per mg of A5B7. Details of the method of administration of '1'I-A5B7 to patients have been reported previously (Ledermann et al., 1991). Repeated doses of F(ab%2 or intact antibody of 1.2-5.5 GBq were given approximately 4 weekly.
At the start of the study 1.8 GBq was given, and this was escalated using a Fibonacci scale to determine the maximum tolerated dose. Further treatment was withheld if an intrdermal skrin test with 10 jag of antibody became positive, if there was more than a 4-fold increase in human IgG anti-mouse klvd in the blood or if there was evidence of disease progression.
Cyclosporin A Cyclosporin A (CsA) was given orally, 15mg kg' in two divided doses per day (Lekemann et al., 1991), to those patients with normal renal function, starting 2 days before the radiolabelled antibody was administered and continuing for a total of 14 days. Serum sampls were taken at intervals to measure the serum CsA level and serum creatim.
Tissue and blood data Serial gamma-camera data were collected from 0 to 384 h and blood data from 0 to 142 h after aministration of the antibody. Radioactivity in blood and urine was masured with an LKB Wizard (Pharmacia) gamma counter. Radioactivity in tumour and normal tissues was estimated usng an IGE Gemini 700 gamma camera. Serial singkl-photon emission computerised tomographic (SPECT) images were obtained of the thorax, abdomen and pelvis and were reconscted usng IGE filtered backprojection software. Images were thn corrected for Compton scatter and photon attenuation as described by Green et al. (1990). Estimates of serial radioactivity per unit mass (MBq kg-') in tissues post administration were made using region of interest (ROI) analysis on transaxial SPECT slices. The cumulative activity (MBq h kg-') delivered by the antibody was esimated from the area under the actvity (MBq kg-') vs time (h) curve usng the trapezoid rule. A simplifid estimation of beta dose (cGy) to each tissue from beta radiation contained in that tissue was made using the MIRD absorbed dose equation: usng S (mean dose per unit accumulated activity) = 0.3691 for '3'I (MIRD pamphkt No. 11, 1975).
Antibody distribution in tumour and normal tissue was determined by decay correcting the measured activity (MBq kg-') expressed as a percentage of the injected radioactivity and plotted against time. Mean patient distribution corresponding to median gamma-camera imaging times was estimated from the measured distribution data.
Antibody clearance was alculated by assuming a biphasic exponential curve fit to the serial tissue and blood data. The claanc phase is taken at greater than 24 h and data points after 24 h used to estimate the cearance half-life (t). Statistical analysis of the two patient groups was performed uing the non-parametnc Mann-Whitney U test.

Tunour response
Evaluation of response to RIT included comparison of preand post-treatment CT images of the tumour, assesment of radiographic or ultrasound images and serum tumour marker levels (CEA, CA19-9). Redsm Antibody localisation Figure la and b shows the distribution of intact antibody and F(ab% in blood and tumour reetively from 4.25 to 120 h post RIT adminitation. The mean percentage of the injected activity per kilogram in tumour at 4.25 h with antibody fragment is 8.2% compared with 4.4% for intact antibody. Increased early loclisation is associated with the patients receiving F(ab% compared with those receiving intact A5B7 (P<0.05). There was substantial variation in tumour loalisation between different patients. Figure 2 shows specific localisation in patient 3 (Table I) receiving A5B7 intact antibody. The maximum masured percentage of the injected activity per kilogram in tumour is 18.4% at 27 h post RIT.

Antibody clearance
The half-lives (h) for the clearance of the A5B7 intact antibody and F(ab') in tumour, blood, liver and lung are shown in Table II Fugwe 1 a, Distribution of A5B7 intact (-) and F(ab)2 fragment (0) in (a) blood and (b) in tumour. Tumour data were derived from serial gamma-camera imaging and blood data from gamma counting of venous blood samples.

Dosimetry
The cumulative doses to tumour, blood, liver and lung per unit of administered activity are shown (cGy MBal) in Table  III. There appears to be no significant differences (P>0.05) in the cumulative dose delivered for F(ab% or intact antibody. However, higher percentage injected activity per kilogram associated with F(ab')2 at 4.25 h will give higher initial dose rates to the tumour.

Toxicity
Toxicity was similar in both groups (Table IV). There was significant nausea and vomiting, together with mild abnormalities of liver and renal function in both groups, which was attributed to CsA. Myelosuppression was the significant dose-limiting toxicity, with the nadir of platelets and granulocytes occurring at 4-6 weeks. The maximum tolerated dose was 2.4 GBq m2. No patients were excluded because of positive intradermal testing with antibody.
Tumour response Responses were seen in both groups. One patient receiving the intact antibody showed a partial response in the size of lung metastases (Figure 3a). Complete resolution of liver metastases was seen in one patient receiving F(ab% 4 weeks after the first treatment, but the tumour regrew to > 50% of its original size after 8 weeks. CT scans of the liver tumour before and after RIT are shown (Figure 3b and c).
'3'1-labelled antibody to CEA has been shown to localise well in colonic xenografts in nude mice and to significantly inhibit their growth (Pedley et al., 1991). Although siilar tumourto-normal tissues ratios are achieved in mice and humans given the same antibody (Begent & Pedley, 1990), the Fge 2 Antibody distribution data (derived from serial gamma-camera imaging) in tumour (0) and normal tissues (0, liver, 0, lung; x, blood) in patient no. 3 (in Table I) who reeived intact A5B7. Days ore and pos: start of RIT ,ieraov Figwe 3 a, Partal response m size of lung meastases in patient no. 4 receiving intact ASB7 antibody. Sum of products in two dimensions of three measurable hmg metastases in plain chest radiograph. b and c, Cr scan of the liver showing a liver metastasis (b) (arrowed) in patient no. 9, which resolved completely after the first course of RIT (c) with F(ab%. therapeutic effect is not as great in man. There may be many reasons for this difference. Tumour volumes are larger in patients than in mice, and it has been shown that antibody localises less efficiently in larger tumours (Pedley et al., 1987).
Human tumour xenografts used in mice are selected for their good loalisation of antibody, whereas there is great variation in localisation between individual patients (Boxer et al., 1992). Neverthekss, evidence of tumour responses were seen in the patients in this and other studies . One of the limitations of RIT is the low dose rate achieved in the tumour. It has been estimated that, below a certain threshold dose rate, tumour growth rate will exceed cell kill rate (Fowler, 1990). Hence, an increase in the dose rate delivered by RIT may be critical for successful therapy. It is possible that a relatively modest improvement in antibody locaistion in tumour relative to normal tissue could produce a marked improvement in clnical response. F(abj2, by potentially doubling the maximum dose rate for the same administered radioactivity and toxicity, may give a substantal improvement in therapeutic sucess. Pedley et al. (1993) have demonstrated that twice the activity of F(ab% radioantibody must be admini in order to produce similar therapeutic effects with the F(ab%2 fragment as the intact antibody. This is due to more rapid circulatory clearance of the fragment during the initial few hours in mice, resulting in a lower absolute amount delivered to the tumour. In addition Pedley et al. (1993) showed that tumour-to-normal tissue ratios were higher and toxicity was reduced with F(ab%. Reduced toxicity is important, as it may allow for an increase in the amount of radioactivity delivered to the tumour in clinical RIT. Human bone marrow has a lower tokrance for radiation than that of mic (Badger et al., 1985;Bigler et al., 1986;Buchegger et al., 1990) and is the main dose-limiting toxcity in RIT. Severe immediate-type hypersensitivity reactions occasionally occur after administration of murine antibodis, and this was the reason for intradermal testing with antibodes before intravenous administration. While this carries a risk of inducing sensitisation to antibody, ensuring a negative intadermal test is the safest course in terms of reducing the risk of severe immediate-type hypersensitivity. IgG human anti-mouse antibody develops very commonly after murine antibody administration whether intradermal testing is done or not. The distribution data presented here do not show the same pattern of rapid circulatory clarance of F(ab% antibody seen in the xenograft model. This is consistent with the similar bone marrow toxicty in the two groups of patients considering that circulating radioactivity is believed to be the source of bone marrow suppression.
The more rapid clearance of F(ab%2 than intact antibody in mice is usually associated with high renal uptake of radioactivity simila to that seen when Fab' is given. Fab' (50kDa) is smIall enough to be filtered by glomeruli and reabsorbed in the renal tubules. Renal uptake of F(ab32 would not be expected if it remains intact at 100 kDa MW, but could be exlained in mice by its being broken down in mouse serum to Fab' while the preparation used here remains stable as F(ab% in the crculation in man. The similar distibution of intact and F(ab)2 antibody in the circulation and liver refutes the hypothesis that clearance of the intact antibody is substanially mediated by Fc reptor binding-This is consistent with the low affinity of monomenc IgG for the Fc receptor, in contrast to the greatly increased affinity of aggregates (Arend & Mannikr 1972).
The data presented are the first in patients to suggest a more rapid penetration of the F(ab% into tumours as compared with intact antibody. It is proposed that the faster penetration of the fragment is the result of its smaler mokcular weight, and this is consient with the finding that small molcular size in antibodies results in improved penetration in tumour animals models (Yokota et al., 1992) and in tumour spheroids (Sutherland et al., 1987;Sunters et al., 1992). Sunters et al. (1992) showed that further reductions in molecular size of antibody may produce even more rapid tumour penetration. In animal model studies F(ab% cleared more rapidly from the circulation than intact antibody, making it difficult to assess the contribution of molcular size to tumour uptake since uptake is in part dependent on the availability of antibody in the circulation. The observations in our study make it possible to see the effect of reducing moleular size on tumour penetration in man because the clarance from pLam of the F(ab% and intact antibody was very similar. Mocules smaler than F(ab% can be expected to clear more rapidly than F(ab% in man, with the result that the absolute amount of antibody delvered to the tumour will probably be less. F(ab)2 may therefore be a good compromise as a therapeutic molecule, giving high absolute amounts of radioantibody in the tumour with rapid penetration.
The overall improvement in initial uptake and therapeutic ratio associated with the administration of the F(ab')2 antibody has important implications for the future design of antibody-targeted therapy.