Quantitative intratumoural microdistribution and kinetics of ¹³¹I-huA33 antibody in patients with colorectal carcinoma

There is increasing interest in the use of humanized monoclonal antibodies to treat epithelial cancers. This technique is attractive owing to limited chemotherapy options for many such cancers and the relatively low toxicity of using tumour-specific radiolabeled monoclonal antibodies. Many unlabeled monoclonal antibodies (mAbs) do not demonstrate clinically significant anti-cancer activity, despite induction of complement-dependent cytotoxicity (CDC) and antibody-dependent cellular cytotoxicity (ADCC). Hence, attempts have been made to utilize the targeting ability of mAbs to deliver radioactive isotopes or other cytotoxic molecules to individual cancer cells [1, 2].

In order for antibody therapy approaches to be effective, it is necessary for the antibody to adequately penetrate within tumours after intravenous administration. Previous studies have found variable uptake in epithelial tumours depending on tumour size, histological type, vascularity, degree of necrosis, antigen expression, antibody size and affinity, antibody internalization, and other factors that are less well understood [3‐10]. Suboptimal response to antibody therapy may be due to poor or non-uniform penetration of antibody into certain tumour areas such as central necrotic regions which lack an adequate blood supply [11]. Effective intratumoural penetration is even more important for antibody-drug conjugates which rely on direct cellular contact to effect cell kill. Despite this, detailed understanding of the impact of antigen expression and vascular density on the distribution of humanized antibodies in tumours, and penetrance into metastatic lesions following systemic infusion, is lacking.

One of the most promising targets in colorectal cancer is the A33 antigen (hereafter referred to as GPA33): a transmembrane glycoprotein of the immunoglobulin superfamily with a molecular weight of 43 kDa [12‐14]. GPA33 consists of two extracellular Ig domains, a single transmembrane domain, and a short intracellular tail containing four acylation sites proximal to the transmembrane domain [12, 14]. Extensive immunohistochemical analysis of malignant and normal tissues has demonstrated that the antigen is homogeneously expressed by more than 95% of colon cancers and in the normal intestinal mucosa, but not in other epithelial tissues [15, 16]. Previous radioimmunotherapy studies using iodinated murine A33 mAb have shown the therapeutic potential of the GPA33 antibody system in patients with metastatic colorectal cancer [17‐19].

HuA33 is a humanized monoclonal antibody (mAb) directed against GPA33 and has been studied in phase I trials in patients with colorectal cancer [20‐25]. Biodistribution studies with ¹³¹I and ¹²⁴I have shown selective and rapid localization of huA33 to colorectal carcinoma on imaging [20, 22, 23, 26]. Tumour localization of huA33 was observed within 24 to 48 h post injection, increased over 7 days, and was retained in tumour for over a month following infusion [20, 22, 23]. As expected, the uptake in the colon and small bowel was also seen, but decreased over 7 to 10 days. Importantly, there was no evidence of non-selective binding to normal tissues. The present study examined tumours from one of these trials [23] in order to explore the intratumoural microdistribution and kinetics of huA33 in vivo, with particular focus on the quantitative cellular uptake and vascular penetrance of huA33 in viable and necrotic tumours.

The demographics and tumour characteristics of the 12 patients in this protocol are shown in Table 1. Gamma camera and SPECT imaging after infusion of ¹³¹I-huA33 was sensitive and specific for tumour (representative images are shown in Figure 1). All lesions in 12 patients that were detected by imaging were confirmed to be of primary or metastatic colorectal cancers at surgery. Patient 7 had a total colectomy yielding three small primary adenocarcinomas. Tumours from seven patients (nos 1, 2, 3, 4, 5, 6, and 12) were noted by the pathologist to have central areas of necrosis, with the tumour from patient 6 also having stromal elements admixed with necrotic tissue. A representative image of a tumour with macroscopic areas of necrosis is shown in Figure 2. Importantly, all specimens showed that viable tumour cells expressed GPA33 as determined by immunohistochemistry (IHC) (Figure 3).

Table 2 shows ¹³¹I-huA33 uptake in tumour measured by radioactivity gamma well counter and autoradiography. The mean whole tumour ¹³¹I-huA33 uptake measured by gamma well counter was 5.13 ± 1.76 × 10⁻³%ID/g, and mean whole tumour ¹³¹I-huA33 uptake measured by autoradiography was virtually identical at 5.12 ± 2.71 × 10⁻³%ID/g. Intratumoural uptake of ¹³¹I-huA33 was compared in the periphery and centre of these tumours. There was reduced uptake in tumour centres compared to tumour periphery (1.61 ± 1.76 × 10⁻³%ID/g vs 4.11 ± 2.53 × 10⁻³%ID/g, p = 0.07). As a subset of patients exhibited macroscopic necrosis in their tumours, these results were stratified by the presence of tumour necrosis (Table 2). This stratification revealed that patients whose tumours were macroscopically viable in both the centre and the periphery had minimal difference in uptake between regions (4.12 ± 0.212 × 10⁻³%ID/g vs 5.39 ± 2.63 × 10⁻³%ID/g, p value not available due to limited patient numbers), whereas patients whose tumours had necrotic centres showed a trend towards less uptake in the tumour centres than the tumour periphery, though this was not significantly different (0.606 ± 0.493 × 10⁻³%ID/g vs 2.98 ± 2.17 × 10⁻³%ID/g, p = 0.06). However, the number of viable tumour cells was also found to be significantly lower in central necrotic areas compared to peripheral viable areas (Table 3; 54 ± 42 vs 313 ± 146, p < 0.001). Given the significantly different numbers of viable cells in necrotic versus viable regions, huA33 uptake was corrected for the differing cell count by calculating the percent injected dose per cell (%ID/cell). Interestingly, cellular uptake of huA33 in tumour central necrotic areas was not significantly different from that in peripheral viable areas (7.10 ± 5.10 × 10⁻⁹ vs 3.82 ± 3.67 × 10⁻⁹%ID/cell respectively, p = 0.4).

We next examined whether there were differences in MVD between these tumours (Table 3). There was a corresponding reduction in MVD in necrotic versus viable tumour (19 ± 11 vs 52 ± 20 vessels/mm², p < 0.001). Interestingly, we observed that MVD in stromal tissue was approximately threefold higher than in adjacent peripheral viable tumour. The mean ICD was also significantly wider in the necrotic tumour than areas of viable tumour, being 207.4 ± 121.7 μm and 69.4 ± 34.5 μm, respectively (p < 0.001).

This study clearly demonstrates the ability of an IgG1 antibody, huA33, to rapidly penetrate at high concentration the central necrotic areas of large metastatic colorectal tumours. At a cellular level, huA33 can target viable tumour clusters at a similar concentration in both necrotic and viable tumour areas. Importantly, huA33 can traverse large distances from intratumoural blood vessels, independent of the interstitial pressure effects present in central necrotic regions of tumours. These novel observations highlight the ability of a high affinity and low internalizing intact antibody to effectively distribute throughout large tumours and target viable tumour cells at relatively uniform and high concentration.

Initial histologic and quantitative autoradiography analysis of huA33 showed that uptake in necrotic areas was moderate (0.606 ± 0.493 × 10⁻³%ID/g), suggesting that intratumoural interstitial pressure may reduce antibody penetration. However, detailed histologic and IHC analysis demonstrated that necrotic tumour centres had significantly lower number of viable tumour cells compared to viable peripheral regions and when corrected for the number of viable tumour cells, uptake of huA33 in necrotic areas was comparable to that in peripheral viable tumour areas (7.10 ± 5.10 × 10⁻⁹ vs 3.82 ± 3.67 × 10⁻⁹%ID/cell respectively, p = 0.4). Whilst most published studies have shown reduced antibody uptake in necrotic areas, Boxer et al. found that, in some cases, patients infused with antibody to CEA had increased concentrations of antibody in necrotic more than viable areas of tumour [4]. However, they did not correct for cell density. In contrast, we found that tumours without central necrosis demonstrated relatively even antibody distribution following intravenous administration of the radiolabelled antibody. Areas of necrosis had lower absolute antibody uptake, but upon examining uptake per viable tumour cell, the uptake was similar to homogeneous areas of viable tumour cells.

Interestingly, the ability of huA33 to penetrate into the necrotic centres of large tumours and localize to viable tumour cells was not reduced despite the demonstrably poorer perfusion of these necrotic central areas (long ICD and low MVD) compared to the periphery (short ICD and high MVD). Other studies have not found consistent relationships between degree of vascularity and antibody uptake in tumours [3‐5, 8, 9, 27]. Given the avascular nature of the necrotic tumour centre, we conclude that huA33 can penetrate long distances from blood vessels before binding to tumour cells. HuA33 exhibited an average maximum penetration distance of 82.5 ± 29.3 μm, with upper limit of this range being 118 μm. This distance might be even larger if the nearest blood vessels in the necrotic centre have impaired function and the antibodies therefore may diffuse in from the vascular capsule of the tumour. Given that huA33 is a high-affinity antibody [21], this long penetration distance is an unexpected finding. This is because very high affinity interactions between antibodies and tumour antigens are predicted to impair efficient tumour penetration of the monoclonal antibodies and thus diminish effective in vivo targeting [28‐32]. Similar findings were recently reported by Rudnick et al. [27], who studied the impact of affinity on the in vivo tumour-targeting properties of anti-HER2 antibodies. They found that the highest affinity mAbs exhibited an average penetration of 20.4 ± 7.5 μm from tumour blood vessels. Conversely, lowest affinity mAbs revealed the greatest average penetration distance (84.8 ± 12.8 μm) and the mAb with a moderate monovalent affinity penetrated to an average distance of 59.7 μm. It is therefore likely that the low internalization rate of huA33 [33] contributes to its penetrance capability in tumour, as internalization plays an important role in antibody capture and degradation [27].

Other intratumoural factors may contribute to the long penetration distance of huA33, such as an enhanced permeability and retention (EPR) effect which is known to lead to a prolonged retention of drugs into tumour interstitium [34, 35]. Whilst the EPR effect might contribute to the huA33 accumulation and retention in the necrotic areas of the tumours, preclinical data suggest that this may be an antibody-specific effect. Ackerman et al. showed the ability of antibodies against GPA33 to progress towards a tumour spheroid centre even at very low concentrations [33]. They hypothesized that the slow rate of GPA33 turnover contributes to the ability to penetrate spheroids, allowing the bound antibody front to penetrate without being slowed by binding to replenish consumed antibody [33]. Aside from antigen turnover rate, it is also known that for an antibody of any given affinity, an increase in binding density is associated with significantly higher monoclonal antibody uptake in tumour [36‐38]. This also appears to be true for huA33, with O’Donoghue et al. observing a linear relationship between ¹²⁴I-huA33 uptake and antigen concentration in tumour, with estimated binding site occupancy for both tumour and normal colon of 20% to 50% [22]. Interestingly, O’Donoghue et al. obtained an antibody uptake of 0.1%ID/mL in focal regions with an average tumour uptake of 0.017%ID/mL which is among the highest reported for radiolabeled antibodies, as the authors concluded. HuA33 uptake observed in our study was similar to that reported by Welt et al. [39]. It is important to also note that unlike some antibodies against other tumour antigens [5], huA33 uptake appears to be universally and strongly present within human colorectal tumours, a result consistent with previous studies of murine A33 antibody in vivo[39]. Previous studies have demonstrated a high and uniform expression of GPA33 in colon cancers and throughout the normal intestinal mucosa [15, 16]. Overall, it seems reasonable to hypothesize that the favourable GPA33 expression level and slow internalization rate are major contributors in determining ¹³¹I-huA33 uptake. This is highly relevant to the design of future studies investigating the therapeutic potential of huA33, particularly for payload delivery strategies. One obvious implication of our results is that patients with large or necrotic tumours should not be excluded from potential treatment with huA33 because of concerns of poor tumour cell targeting. Another is that combination therapy with antivascular agents will also require careful design given their potential effects on antibody penetration, as also shown in other studies [40, 41] where bevacizumab treatment demonstrated to impair the penetration of antibodies into the tumour, resulting in a reduced tumour uptake of these antibodies.

The limit of our study is the small patient numbers, so further studies are needed to confirm the present findings.

In summary, targeting GPA33 with huA33 presents a very attractive option given the universal and uniform expression of this antigen in colorectal cancers, and, within the limit of the study, the data presented here shows excellent uptake in primary and metastatic tumours of varying size and locations. Tumour cellular uptake is very high, even in the necrotic centres of tumours where tumour cell numbers and microvessel densities are very low, an unusual and therapeutically important finding for an intact IgG. These findings provide an important mechanistic support for ongoing clinical trials of huA33 in patients with metastatic colorectal carcinoma and provide data to rationally design them.