Single-cell Cytotoxicity with Radiolabeled Antibodies¹ Free

We previously described the cytotoxicity for B-lymphoma cells of an Ab to CD74(MHC³class II invariant chain, Ii) conjugated to Auger-electron-emitting radionuclides, including ¹²⁵I, ¹¹¹In, and ^99mTc (1). This Ab is taken up in large amounts by target cells,approximately 10⁷ Ab molecules/cell/day, because of constitutive rapid internalization of molecules at the cell surface and replacement by newly synthesized molecules. Abs internalized from the cell surface are delivered to lysosomes and promptly degraded,together with Ii. If residualizing radiolabels are used, they accumulate within cells to high levels. Residualizing labels are defined as those that generate catabolic products that are unable to efficiently cross membranes and therefore are trapped within lysosomes. The level of radioactivity reached in these experiments was >50 cpm/cell, and 100% of the target cells were killed (5 ×10⁵ cells total). Dosimetry calculations indicated that the radiation dose delivered was consistent with the cytotoxicity observed.

The current study was intended to extend these results in two ways: to test other cell lines expressing Ii, including adherent target cells;and to test other Abs reacting with B-cell lymphomas. Melanoma cells and many carcinoma cells can be induced to express high levels of both mature MHC class II antigen and the invariant chain by IFN-γ. Anti-CD74 uptake by induced SK-MEL-37 melanoma cells was comparable with the uptake by Raji cells (2). However, the very different cellular morphology is likely to have a significant impact on killing by internalized radionuclides. More specifically, the spreading of the cells on plastic would tend to decrease the proximity of nucleus and cytoplasm. Therefore, we have tested cytotoxicity for these adherent cell lines, as well as for other B-cell lymphomas, to establish the generality of the results.

Whereas the massive intracellular uptake of anti-CD74 is not matched by any other Ab to our knowledge, some other Abs can bind to the cell surface in similar amounts. For example, the number of mature MHC class II molecules on Raji B-lymphoma cells is approximately 3 ×10⁶ (2). (Note that the mature MHC class II molecules lack the invariant chain, which must be removed before binding of peptide antigens.) This level of expression suggests that some cytotoxicity might be obtained with anti-MHC class II,although perhaps not as strong as with anti-CD74. A basic difference between anti-CD74 and antimature MHC class II is in their subcellular localization. Thus, whereas the former is delivered to lysosomes, the latter remains primarily on the cell surface. According to dosimetry calculations, isotopes in the cytoplasm are expected to be more potent than isotopes on the cell surface, but the difference is relatively small for the radionuclides used in these experiments. For example, the advantage of internalization for ¹²⁵I and ¹¹¹In is 53 and 67%, respectively (3), as calculated for a cell with the dimensions of Raji(cell diameter, 16 μm; nuclear diameter, 12 μm; Ref. 1), which suggests that isotopes localized to the cell surface would still be quite active. It should also be noted that considerable evidence suggests a recycling pathway for mature MHC class II, in which the molecules enter an intracellular compartment where peptide antigens are exchanged (4, 5). This factor, of course, would further enhance the cytotoxic capacity of the Ab, but the effect would be small because most of the antigen is on the cell surface at any one time (4).

Another Ab was also tested, namely anti-CD20. With approximately 3 × 10⁵ sites/cells (6, 7), this Ab was not expected to provide an effective target for killing in this experimental system, but it seemed possible that some cytotoxicity would be observed. Unexpectedly, radiolabeled anti-CD20 killed cells very effectively, and the explanation for such killing is partially elucidated below.

The cell lines used were the B-cell lymphomas Raji, RL, Ramos, and Daudi and the melanoma SK-MEL-37. The origin of these lines and the culture conditions were described previously (2, 8). Cell lines were tested routinely for Mycoplasma contamination using the Mycotect Assay (Life Technologies, Inc., Grand Island, NY)and were negative.

Anti-CD74 (LL1) was supplied by the Antibody Production facility at Immunomedics, Inc. (Morris Plains, NJ). The source of L243 (anti-MHC class II antigen) was described previously (5, 8). The hybridoma producing 1F5 was obtained from the American Type Culture Collection (Rockville, MD). Ab-producing cells were found to constitute<10% of the cell population, so the Ab-producing cells were cloned by limiting dilution before expansion of the cells. Ab purification was on an affinity column of Sepharose linked to protein A or protein G(Amersham Pharmacia, Piscataway, NJ), as appropriate for the Ab isotype. Abs were labeled with ¹²⁵I and ¹³¹I by the chloramine-T method and with ¹¹¹In with the chelator isothiocyanate-benzyl-DTPA, as described previously in detail (1). Labeled preparations were analyzed by instant TLC,gel filtration high-performance liquid chromatography, or both by methods that have been described (9) to determine the level of radioactivity not bound to the Ab, which was always <10% and usually <5%. Representative preparations of radiolabeled Abs, with each radiolabel, were tested for immunoreactivity (percentage bindable)by incubating with a large excess of cells. Control tubes had excess unlabeled Ab added to block specific binding and therefore to indicate the level of nonspecific binding; specific binding was calculated by subtraction. The range of specific binding was as follows: LL1,53.0–71.7%; LL2, 42.1–60.1%; 1F5, 50.6–74.8%; and L243,56.3–58.6%. Such variation in immunoreactivity could have some effect on the data presented, but this would be at most a small effect, and it would not significantly alter any of the conclusions.

All of the methods were described in detail previously (1). Briefly, cells were incubated for 2 days with varying concentrations of radiolabeled Abs and then diluted 14.3-fold. Cell counts were obtained every 3–5 days until the cells had multiplied 16-fold or until day 21. The fraction of cells killed was calculated from the growth curves, without considering possible division delay induced by irradiation.

SK-MEL-37 melanoma cells were preincubated for 2 days with human IFN-γ (IFNγ-1b; Actimmune; Genentech, South San Francisco, CA) at 250 units/ml, and IFN-γ at this concentration was included in the medium used throughout the assay. After trypsinization, 0.15 ml of a 0.1% suspension (packed cell v/v) was plated in wells of 96-well plates and incubated overnight. After aspiration of the medium, 0.2 ml of medium containing radiolabeled Ab was added. Serial dilutions of the Abs were used, each tested in triplicate, and the incubation was for 48 h. The control wells contained excess unlabeled Ab at 62.5μg/ml to inhibit specific Ab binding and thereby to indicate the level of nonspecific cytotoxicity. Some wells were used to measure uptake of radioactivity, some were used for cell counts after trypsinization, and some were used for a clonogenic assay. To measure uptake, cells were washed three times with tissue culture medium and then harvested with 50 μl 2.0 m NaOH, which was collected with a cotton swab. For the clonogenic assay, cells from a single well were trypsinized and suspended in 6.7 ml medium. Serial 4-fold dilutions were prepared, and 5 ml of each dilution were plated in T30 flasks. After 12–15 days of growth, colonies of >50 cells were counted in those flasks that contained 50–200 colonies. Before counting, colonies were fixed for 10 min with 5.0% glutaraldehyde in 0.1 m sodium cacodylate HCl (pH 7.2), stained for 10 min with 0.5% methylene blue in 25% ethanol, and washed thoroughly with water. Irradiated cells grew more slowly than control cells, which can be attributed to radiation-induced growth delay (10)or to other toxic effects of the radiation. Therefore, compared with the control cells, the treated cells were allowed 2–3 more days to grow before counting colonies. In practice, this allowed all of the healthy colonies to reach the countable size of ≥50 cells. The cloning efficiency of control cells was approximately 60%.

Uptake of cpm was determined in experiments identical to the cytotoxicity experiments. At various times, cells were collected,pelleted, and washed three times with 5 ml tissue culture medium, and the radioactivity was determined. Routinely, 1/20 of the total sample was counted because otherwise the cpm were too high to be counted accurately (>10⁶ cpm). Cell counts were also determined at each time point, although, for reasons discussed below,the data are generally expressed in terms of the initial cell count. The number of Ab molecules bound/cell was calculated from the cpm bound and the specific activity and was corrected for the fraction of cpm in the stock Ab preparation that was not associated with intact IgG(always <10%). For radiation dosimetry, the area under the curve of bound cpm versus time was calculated, which provided the cumulative counts. Cumulative disintegrations were calculated from the gamma counter efficiencies, which were 70.9% for ¹¹¹In and 76.5% for ¹²⁵I. Cellular S-values for Raji-sized cells (diameter, 16 μm; nuclear diameter, 12 μm) were from Goddu et al. (3).

Experiments were set up like the cytotoxicity experiments, with cells in 24-well plates. ¹²⁵I-labeled L243 was included at 5 μCi/ml, which kills approximately 95% of the cells after a standard 2-day incubation. After 2, 22, or 44 h, cells from three wells were collected, pooled, pelleted, washed three times, and suspended in 2.5 ml. Aliquots of 0.5 ml were replated in four wells of a 24-well plate, and 0.5 ml of L243 hybridoma ascites diluted 1:25 was added to two of the wells, whereas the other two wells received control medium. The principle of this assay is that excess unlabeled Ab replaces the labeled Ab on the cell surface, because of the fact that the bound labeled Ab, although predominantly bivalently bound,frequently wobbles on the cell surface (11). After 24-h incubation under tissue culture conditions, the cells and supernatant were analyzed for remaining cpm, as described previously in detail (8). The supernatant was also tested by precipitation with cold 10% TCA to distinguish between intact and degraded Ab. Large fragments of Ab would also be TCA precipitable, but such fragments are not generated from L243 in significant amounts in these experiments, as determined previously (5), and, in any case, would not affect the interpretation of the results in this study.

Previous cytotoxicity experiments were performed only with the Raji B-lymphoma cell line (1). To establish the generality of the results, similar experiments were performed with three other B-cell lymphomas, Ramos, Daudi, and RL. Ramos and Daudi, like Raji, are Burkitt lymphoma cell lines, whereas RL is derived from a diffuse non-Hodgkin lymphoma. Daudi is the only one of these four lymphoma cell lines that has a normal p53 gene (12, 13),which might be expected to increase its sensitivity to radiation (14). Cytotoxicity experiments were performed similarly to those described previously, with serial dilutions of the radiolabeled Ab starting at 90 μCi/ml. The results, shown in Fig. 1, were generally similar to the results described with Raji cells, with high levels of antigen-specific cytotoxicity with all of the cell lines. With three of the four cell lines, 100% killing was obtained(> approximately 6 logs) at the highest Ab concentration tested,whereas the killing of RL cells was slightly less. Daudi was markedly more susceptible to killing than the other three cell lines, with both antigen-specific and nonspecific killing. In the assay of nonspecific killing, a nonreactive Ab at 90 μCi/ml killed 85–96% of Daudi cells, whereas killing of the other three cell lines was only 30–50%. By comparison with previous experiments in which higher μCi concentrations were tested in the killing of Raji (15),Daudi was approximately 3-fold more sensitive to nonspecific killing. In specific killing with ¹¹¹In-labeled LL1, Daudi was also 2–3-fold more sensitive, as shown in Fig. 1. Also, unlike the other target cells, Daudi cells did not gradually become very large before lysis, as described previously for Raji (1), but instead lysed more rapidly. Both the lack of cell enlargement and the susceptibility to killing of Daudi cells are likely to be related to the presence of wild-type p53, which plays a role in the induction of apoptosis (16), but additional studies are required to confirm this possibility.

An adherent cell line, SK-MEL-37, which has a level of Ab uptake similar to that of the B-cell lymphomas (2), was also tested as a target cell. These cells were adherent to plastic throughout the assay, which was considered to be a potentially important difference from the B-cell lymphomas. As shown in Fig. 2, strong, antigen-specific killing was observed. With the highest Ab concentration tested, 10 μg/ml, which had 159 μCi/ml, the colony-forming units were reduced by 97.4%. A low level of nonspecific cytotoxicity was observed; this was attributable to true nonspecific cytotoxicity and not to residual specific binding that was not inhibited by excess unlabeled Ab, because essentially the same results were obtained with a nonreactive Ab, labeled similarly (data not shown). The uptake of radioactivity, determined at day 2 (the time of cloning) in the same experiment, was 3.1 ×10⁶ cpm/8.0 × 10⁴cells/well or 38.8 cpm/cell. At least 88% of this uptake was antigen specific because it was inhibited by excess unlabeled Ab. However,100% killing, as observed with lymphoma target cells, was not obtained with the melanoma targets, despite comparable levels of cpm uptake/cell.

Two other Abs, reacting with relatively high-density antigens, were labeled with ¹¹¹In and tested for their ability to kill Raji cells. Antimature MHC class II (Ab L243) and anti-CD20 both killed cells very effectively (Fig. 3). In comparison with the previous CD74 experiments, L243 was comparable in potency, whereas CD20 was somewhat less potent on the basis of the dose-response curves. At the highest concentration tested, CD20 killed slightly <100% of the target cells. For example, in the experiment shown in Fig. 3, the fraction surviving with 1F5 at 90μCi/ml was 2.99 × 10⁻⁶, which corresponds to approximately 1–2 cells surviving of the initial 5 × 10⁵ cells. Nonreactive Abs labeled in the same way had no significant cytotoxic effect at the concentrations used, as shown in Fig. 3 C, demonstrating that the cytotoxicity was antigen-specific.

For both Abs L243 and 1F5, unlabeled Ab used at the same protein concentration was tested to demonstrate that the cytotoxicity was attributable to the radiation delivered. The Ab concentration was 4–5μg/ml, which is a near-saturating concentration. Results with unlabeled L243 are shown in Fig. 3,A. This Ab appeared to have a transient effect at day 2, which was probably because of the fact that the Ab causes marked clumping of the cells, but by day 5 the cells had multiplied essentially as much as untreated cells. In a similar experiment with unlabeled 1F5, the Ab had no detectable effect on cell growth (Fig. 3 B).

Uptake of cpm was determined in similar experiments. As shown in Fig. 4, the uptake of both Abs was very high. The peak uptake with L243 was 47 cpm/cell, and the peak uptake with 1F5 was 29 cpm/cell. Although we determined the cell count at each time point in all of the experiments,the data are presented in terms of the initial number of cells plated/well. The reason for this is that the cells start to die as early as day 2, and therefore use of the actual viable cell number results in artifactual high values of cpm/cell at later time points. For example, in the experiment shown in Fig. 4, L243 killed approximately 70% of the initial cells at day 2, which resulted in a value of >150 cpm/viable cell, which is clearly misleading. This issue was discussed previously in greater detail (1). Fig. 5 shows typical growth curves in experiments of this type. Note that cells were killed much more rapidly with L243 than with 1F5, but even with 1F5, cell division was blocked rapidly, and there was very little increase in cell number after time 0. This was a consistent finding in experiments in which a large fraction of cells were killed and supports the validity of using the initial cell count to calculate the cpm/cell. The ability of radiation to rapidly block cell division is well known (10).

Fig. 4 also shows the number of Ab molecule-equivalents bound/cell at various times. These are referred to in this way because some of the isotope retained by the cells is no longer on intact Ab but rather in the form of catabolic products. The uptake of L243 by the cells peaked at 7.2 × 10⁶ Ab molecules/cell at 48 h but was already at a high level at 1 h. This pattern of binding kinetics is markedly different from that described previously with LL1,where binding was very low at 1 h. This difference can be attributed to the fact that the number of antigenic sites/cell on the cell surface is much higher with L243 than with LL1. The substantial increase in L243 binding with time that does occur can be attributed to two factors. First, although L243 does not display rapid internalization, it will be gradually internalized and degraded because of normal turnover of cell surface constituents. Although not a major factor, this will contribute to an increase of cpm/cell over time when residualizing labels are used. More specifically, we determined previously that, of the total L243 counts bound initially to Raji cells, 21% were catabolized within 2 days (5). Second,the irradiated cells grow substantially in size as an effect of lethal radiation. As described previously (1), by day 2 the mean cell diameter increased from 15.4 μm to 20.9 μm, corresponding to an increase in surface area of 84% (assuming that the abundance of microvilli and other surface conformations do not change). These two factors can readily explain the uptake of 7.2 ×10⁶ Ab molecule-equivalents/cell at 48 h,given 3 × 10⁶ sites/cell normally.

With 1F5, however, the high level of Ab uptake/cell, peaking at approximately 5 × 10⁶ Ab molecules/cell at day 2, is not so readily explained. The number of CD20 sites/cell on Raji cells was reported to be 3.6 × 10⁵(7), which is similar to the value reported for another B-lymphoma cell line, Daudi (6). We redetermined the sites/cell at saturation and found that there were 2.4 ×10⁵sites/cell⁴, which is consistent with the previous reports. Thus, the number of ¹¹¹In-labeled Ab molecules bound/cell at 48 h, as shown in Fig. 4, was approximately 20-fold higher than the number of sites/cell. The increase in cell size because of lethal irradiation,as noted above, can account for only a 2–3-fold increase, so a 10-fold increase remains to be accounted for. Another factor, as with L243, is the turnover of the antigen (and any Ab bound to it). However, CD20 is considered to turn over slowly, as indicated by low rates of catabolism of bound Ab. Press et al. (6) reported that there was no significant catabolism of a CD20 Ab by Daudi cells. We described significant levels of catabolism using the Ramos cell line,but this was still at a relatively low level (17). The impact of this factor can be readily investigated by comparing residualizing with nonresidualizing radiolabels. A convenient nonresidualizing label is a conventional chloramine-T iodine label,because iodotyrosine rapidly leaves the cell after it is generated in lysosomes (discussed in Ref. 18). Results of such a comparison with anti-CD20 demonstrated that the advantage of a residualizing label is modest,<2-fold.⁵Therefore, the high level of binding of 1F5, much higher than the number of sites/cell, remains to be explained and is discussed further below.

Because both of these Abs are internalized at relatively low rates, the majority of bound Ab is expected to be at the cell surface, at least for the first 2–3 days of incubation with Ab. The implication of this statement is that the cells are killed primarily by radiation emitted from the cell surface. This represents a major difference from the situation with LL1, because this Ab is localized primarily to lysosomes. Because cytoplasmic radionuclides are expected to be somewhat more cytotoxic than radionuclides on the cell surface, as noted in the “Introduction,” it seemed important to confirm further that radionuclides on the cell surface were responsible for cell killing. Accordingly, experiments were performed with conventional iodine labels, both ¹³¹I and ¹²⁵I. These labels, as noted above, rapidly leave the cell after catabolism, so are not trapped to a significant extent in lysosomes. Therefore, they remain localized in target cells primarily at the cell surface.

Fig. 6 demonstrates effective cytotoxicity obtained with both ¹³¹I and ¹²⁵I labels. Both radiolabels were tested at the highest concentration that produces little if any nonspecific killing, as determined previously (1, 15). The ¹³¹I-labeled conjugate was tested at a starting concentration of 16 μCi/ml, whereas the ¹²⁵I-labeled conjugate was tested at concentrations up to 138 μCi/ml. With both isotopes, L243 was considerably more potent than 1F5, but 1F5 still had high levels of antigen-specific cytotoxicity. For example, in Fig. 6,B, the cytotoxicity from ¹³¹I-labeled 1F5 may seem weak in comparison with L243, yet it produced a cell kill of >3 logs. Fig. 6 also includes data with LL1 for comparison, as well as data with a nonreactive control Ab. With LL1, the labeled conjugate is rapidly catabolized and released from the cells, resulting in little accumulation and consequently little cytotoxicity, as shown.

Using an ¹²⁵I label, the uptake of radioactivity was also determined with both Abs, and results are shown in Fig. 7. The uptake of ¹³¹I-labeled Abs is expected to be very similar, because the specific activities and the immunoreactivities with both isotopes were similar. These experiments were set up exactly like the cytotoxicity experiments, using the minimum Ab concentration that kills nearly 100% of the cells. This concentration was 16 μCi/ml for L243 and 138 μCi/ml for 1F5, so 8.6-fold higher for 1F5. As shown, high levels of uptake were observed with both Abs, with a peak uptake of approximately 20 cpm/cell at 2 days.

From the area under the curve on the plot of cpm versustime, the cumulative counts were determined for the uptake experiments described above. The cumulative dpm were then calculated from the gamma counter efficiency. Using the published S-values for ¹²⁵I present on the cell surface, 1.15 ×10⁻⁴ Gy/Bq·sec, for a cell the size of Raji,the total radiation dose delivered was 1386 cGy with ¹²⁵I-labeled L243 and 1530 cGy with ¹²⁵I-labeled 1F5. These values are presented in Table 1, which also lists the calculated cGy dose delivered in 2 days. We have not determined the time of functional cell death, but it is likely that much of the dose delivered at later time points was redundant, because the cells were already dead. Similar values are also shown for ¹¹¹In-labeled conjugates. With the ¹¹¹In label, the situation is complicated by the fact that some of the bound Ab is internalized and catabolized, and the catabolites are retained within lysosomes. Thus, the label is not entirely on the cell surface, and it may be appropriate to use the cytoplasmic S-value for some fraction of the radiolabel. However, we know from previous studies that the great majority of the bound Ab remains on the cell surface with these two Abs (see below, also), so use of the cell surface S-value is appropriate to estimate the radiation dose delivered. We have not attempted to correct these values of radiation dose for the gradual increase in the size of the cells resulting from lethal irradiation, which was described above. This Table is intended only to provide an estimate of the radiation dose delivered in these experiments and does not allow a precise comparison between Abs and radionuclides, because the uptake naturally depends on the Ab concentration used. For ¹³¹I, assuming the same uptake as with ¹²⁵I, the dose delivered would be only slightly lower than with ¹²⁵I,because the S-value is 1.06 × 10⁻⁴Gy/Bq·s. Radiobiological parameters for Raji, using external X-irradiation, was determined previously (1). The D₀ was 90 cGy, and the extrapolation number ñ (10) was 1.31. Therefore, the expected killing is consistent with the cytotoxicity observed. The expected dose for a kill of 6 logs is 1268 cGy.

Although it seems clear that the great majority of the bound iodine must be present on the cell surface at all times because catabolites are released rapidly from the cell, we demonstrated this directly by competitive elution with excess unlabeled Ab, using L243. This method,which was described and characterized previously (11), is based on the fact that Abs bound to the cell surface frequently wobble(releasing one of the antigen molecules), although they are predominantly bivalently bound. For this reason, such Abs can be competitively replaced by unlabeled Ab that is present at a high concentration. As shown in Fig. 8, the great majority of the bound L243 Ab (80–90%) was eluted with excess unlabeled Ab, and this fraction was the same whether the initial Ab incubation was for 2, 22, or 47 h. These data demonstrate that the Ab remains on the cell surface and does not accumulate in any intracellular compartment to a substantial level. Because elution in these experiments requires 24 h for completion, the possibility remains that the bound Ab recycles between the cell surface and an intracellular compartment, such as early endosomes, but we can conclude that the bound Ab spends at least some of the time at the cell surface.

There are four major conclusions of this study: A)target cells in addition to Raji (which were tested previously) were killed effectively by anti-CD74; Susceptible targets included three other B-cell lymphomas and the adherent melanoma SK-MEL-37; B) Abs to antigens other than CD74, linked to either Auger electron emitters or β-particle emitters, effectively killed target cells; C) Abs that remain primarily on the cell surface were effective cytotoxic agents; Internalization of bound radionuclides into lysosomes is not required; and d) with Abs that remain on the cell surface, it was predictably not necessary to use residualizing radiolabels to obtain strong cell killing. Although it remains likely that internalization provides some advantage, as predicted by the calculated S-values, the expected advantage is only slightly <2-fold for the radionuclides used here. On the other hand, there is likely to be an advantage of cell surface-bound Abs, in that the Abs can saturate the antigenic sites rapidly in a few hours. With anti-CD74, in contrast, the level of binding at 2–4 h is very low, and 24 h are required for the maximum level of uptake to occur. Therefore, cell surface-bound Abs such as L243 deliver higher radiation doses during the first 24 h, which probably explains why cytotoxicity with L243 was observed earlier than with LL1. We investigated previously (1, 15) the possibility that radiation crossfire (cell to cell) is a significant factor in experiments of this type by plating cells at different concentrations. With both ¹¹¹In- and ¹³¹I-labeled LL1, crossfire was not a significant factor. Because L243 and 1F5 bound at similar or lower amounts to the cells, it is very unlikely that crossfire would be significant with these Abs either.

Earlier work suggested that radiolabeled Abs would not effectively kill single cells, but this conclusion was based on a limited body of experimental data, much of which we discussed previously (1). Warters et al. (19) first attempted to kill cells with ¹²⁵I bound to the cell surface. They used ¹²⁵I-labeled con A, which is expected to bind to a very large number of sites/cell, but only weak killing was observed. The maximum killing obtained was approximately 90%. But there were several problems associated with the use of con A. First, the bound con A was very rapidly catabolized by the cells, which made it necessary to perform the binding incubation for 36 h at 4°C to maintain the con A on the cell surface. Second, the unlabeled con A was cytotoxic for the cells, necessitating the use of trypsinized, monovalent con A. The avidity of this monovalent ligand is unlikely to be as high as that of bivalent Abs, which bind essentially irreversibly (11, 20). In any case, it should be noted that the D₀ (dose required for 63%kill in the linear part of the semilog dose-response curve) for ¹²⁵I-labeled con A was 12,600 disintegrations/cell with CHO cells, which is in the same range as the values we found with Ab LL1. The D₀for ¹²⁵I-labeled IMP-R2-LL1 was approximately 24,900 with Raji cells (15). The CHO cells appear to be approximately 2-fold more resistant to radiation than Raji cells, as judged by results of external beam irradiation (1, 19). One important variable in the method of calculating D₀ must be recognized. The total cumulative disintegrations used by Warters et al.(19) was over a 2-day period, whereas we used a period of 4–6 days because the isotopes were strongly retained by the cells,primarily within lysosomes. It is quite possible that, in our experiments, the cells were already killed by day 2 and that radiation delivered after that time was redundant. If this were the case, the true D₀ would be much less than our estimate. In general, we consider the results of Warters et al. (19) to be consistent with our data.

Another key study was by Lindmo et al. (21),who attempted to kill melanoma cells with an ¹³¹I-labeled Ab that reacts with a high density antigen on the cell surface. These authors showed that strong killing was obtained only if Ab-coated cells were stored frozen for many weeks to allow radiation damage to accumulate. This result, then, emphasizes the low level of cytotoxicity obtained. In general, the relatively weak cytotoxicity that has rarely been reported with radiolabeled Abs has been attributed to unusual properties of the particular Abs, such as transport to the nucleus (22, 23) or accumulation in macropinosomes (24). Cytotoxicity with ¹³¹I-labeled anti-CD20 was recently reported by Johnson and Press (25). These investigators intentionally pelleted the target cells to increase crossfire from adjacent cells, so their results may or may not demonstrate single-cell kill. We note thatα-particle emitters are exceptions to many of the above statements. This type of radiation can efficiently kill single cells (26); however, the available isotopes have short half-lives of ≤7 h, which seem unsuitable for treatment of solid tumors where substantial time is required for tumor penetration.α-Particle emitters also appear to be very toxic, considering that only 20 μCi of ²¹²Pb-labeled Ab can be administered to a mouse (27).

From the cellular S-values published by Goddu et al.(3), we can calculate the number of disintegrations on the cell surface required to produce a defined level of cell kill. For Raji cells, the radiation parameters, determined from Cs¹³⁷ irradiation experiments, were D₀ = 90 and ñ = 1.31 (1), so the dose required for a kill of 6 logs is approximately 1270 cGy. For a cell the size of Raji(R_C = 8 μm; R_N = 6 μm), this would require approximately 110,000 disintegrations of ¹²⁵I,120,000 disintegrations of ¹³¹I, and 164,000 disintegrations of ¹¹¹In. From our data, the amount of radioactivity delivered/cell reached these levels, and the cytotoxicity observed was generally consistent with the calculated radiation dose. However, as noted above, the time course of irradiation is an important factor that has not been incorporated into the calculations.

Although the ability to kill single cells is clearly critical for therapy of micrometastases, its importance for therapy of solid tumors is uncertain. Indeed, radiolabeled Abs used for radioimmunotherapy, both experimentally and clinically, are very rarely even tested for in vitro cytotoxicity. This is partly because crossfire is a major factor in the killing of cells within solid tumors, whereas it does not apply in the killing of single cells (28). In vivo experiments, both in animal models and in patients, are required to ascertain the relevance of single-cell kill to therapy. The results described herein do demonstrate, however, that two of the Abs that have been most effective in clinical studies of radioimmunotherapy, anti-CD20 (29) and anti-MHC class II (30), are both in fact able to effectively kill single cells. This suggests that the importance of single-cell kill may be underestimated. At the doses used in patients, initial Ab concentrations are high enough to produce substantial single-cell kill, according to our in vitrodata. For example, with a dose of approximately 60 mCi of ¹³¹I-labeled Lym-1, as used by DeNardo et al. (30), the peak concentration in interstitial fluid will be approximately 6 μCi/ml, which from our in vitro data would produce a high level of cell kill. This is also true for the myeloablative dose of up to 800 mCi of ¹³¹I-labeled anti-CD20 used by Press et al. (31). We note that the anti-MHC class II Ab used herein, L243, is not allele specific, unlike Lym-1 (32),so would be expected to have more consistent reactivity with diverse B-cell lymphomas.

An enigma regarding anti-CD20 binding is that the uptake of Ab is much higher than the number of sites/cell by a factor of approximately 20. There are two known factors that contribute to increasing the Ab uptake. First, there is some turnover of the antigen, which provides some increase in the Ab molecule-equivalents binding/cell. Second, once radiation damage occurs, the cells increase markedly in size, which naturally results in an increase in the Ab molecules bound/cell. However, these factors can explain only a 3–4-fold increase in Ab sites/cell, so a further explanation is required. Two possibilities are apparent. First, the level of antigen may be up-regulated as a result of Ab binding. Secondly, there may be a pool of intracellular antigen that equilibrates with the cell surface and is gradually bound by Ab. Which of these possibilities is correct, if either, is currently under investigation.

Although the 98% kill of the adherent target cells, SK-MEL-37, was substantial, it was not as complete as the killing of B-lymphoma cells,which was 100%. This is likely to be attributable at least partially to the shape of the cell. Although we have not attempted to perform dosimetry calculations for cells adherent to plastic, which would be quite complex, it is clear that the flattened shape of the cell will tend to separate radionuclides in the cytoplasm from the nucleus. Moreover, the rim of cytoplasm will be quite thin above and below the nucleus. Thus, the radiation dose delivered to the nucleus will be considerably less than with round lymphoma cells. In addition, the SK-MEL-37 cells may be less sensitive to radiation than B-lymphoma cells. We would also like to suggest another possible contributory factor, though speculative. When the SK-MEL-37 cells form a monolayer,the cells are always most dense along the edges of the wells, and it seems possible that a few cells may be buried beneath other cells at the edges and therefore inaccessible to Ab.

The major advantage of Auger electron emitters over β-particle emitters appears to be their much lower level of nonspecific cytotoxicity (15). This was demonstrated in vitro (15) and has also recently been shown in vivo in a SCID mouse xenograft model.⁶In regard to single-cell kill alone, ¹³¹I was more potent than either ¹¹¹In or ¹²⁵I, on the basis of the μCi/ml required for a particular level of kill, and this was true for all of the three Abs tested. The potency of ¹³¹I can be at least partially attributed to its significant nonspecific toxicity, which will enhance the effect of specific Ab binding. The higher energyβ-particles of ⁹⁰Y (relative to ¹³¹I) are expected to be somewhat less potent at single-cell kill, and this was demonstrated in vitro with Ab LL1 (15).

The three antigens targeted in this study are clinically important. However, the significance of these results will be amplified if the approach can be applied to other antigens that are not present at such a high density. This can potentially be achieved by using more potent radionuclides or higher specific activity. It should be emphasized that the radionuclides we have tested were selected only because of availability and that many more potent radionuclides (based on theoretical S-values) could be selected. There are at least four radionuclides that are expected to be at least 3-fold more potent than ¹¹¹In when delivered to the cell surface, having suitable half-lives of 1.7–15.4 days. These are ¹⁵³Sm, ¹⁹¹Os, ^195mPt, and ^195mHg. ^191mOs would also be potent, because it decays rapidly to ¹⁹¹Os. On the basis of the analysis of Sastry et al. (33), ¹¹⁹Sb and ^119mTe are also expected to be potent(although cellular S-values have not been published for these isotopes). ^195mPt, the most potent of the group,is expected to be 9.9-fold stronger than ¹¹¹In. An isotope like ¹⁵³Sm is attractive because it emits both abundant conversion electrons and moderate-energyβ-particles. None of these radionuclides are currently available carrier-free, and only ¹⁵³Sm has been conjugated to Abs with high stability (34). Large increases in specific activity, using currently available isotopes, are also possible. Conjugation of a single ¹¹¹In atom/Ab would yield a specific activity of approximately 240 mCi/mg, which is 6-fold higher than the highest specific activity we have used in these experiments. Furthermore, it has been demonstrated that some Abs can be conjugated with five chelates without impairing immunoreactivity (35). In summary, there are realistic approaches by which much lower-density antigens might provide effective targets.

We are grateful to Dr. Habibe Karacay of Immunomedics, Inc.,Morris Plains, NJ, for synthesis of isothiocyanate-benzyl-DTPA,to Mark Przybylowski, Tom Jackson, and Phillip Andrews for assistance with radiolabeling, and to Rosana B. Michel for technical assistance.