A low-toxicity IL-2-based immunocytokine retains antitumor activity despite its high degree of IL-2 receptor selectivity.

PURPOSE
The goal of the study was to engineer a form of interleukin 2 (IL-2) that, when delivered as a tumor-specific antibody fusion protein, retains the ability to stimulate an antitumor immune response via interaction with the high-affinity IL-2 receptor but has lower toxicity because of the reduced activation of the intermediate-affinity IL-2 receptor.


EXPERIMENTAL DESIGN
We investigated changes in the proposed toxin motif of IL-2 by introducing a D20T mutation that has little effect on the activity of free IL-2. We expressed this IL-2 variant as a fusion protein with an antibody (NHS76) that targets the necrotic core of tumors and characterized this molecule (NHS-IL2LT) in vitro and in vivo.


RESULTS
NHS-IL2LT was shown to have near normal biological activity in vitro by using T-cell lines expressing the high-affinity IL-2 receptor, but little or no activity by using cell lines expressing only the intermediate IL-2 receptor. Relative to the control antibody fusion protein containing wild-type IL-2, NHS-IL2LT retained antitumor activity against established neuroblastoma and non-small cell lung cancer metastases in syngeneic mouse tumor models but was much better tolerated in immune-competent mice and in cynomolgus monkeys.


CONCLUSIONS
The qualities of low toxicity and single-agent efficacy shown suggest that NHS-IL2LT is a good candidate for therapeutic approaches combining standard cytotoxic and immune therapies. In fact, this molecule (also known as Selectikine or EMD 521873) is currently in phase I clinical trial.


Introduction
Despite many years of clinical use, the mechanisms of interleukin 2 (IL-2) toxicity and therapeutic efficacy for cancer are not well understood. Clinical dose schedules showing the highest objective response rates in renal carcinoma and melanoma are associated with severe toxicities, especially those of the vascular compartment such as vascular leak syndrome (1). Many mechanisms have been proposed for this vascular toxicity. In one case, toxicity has been attributed to direct binding of IL-2 to endothelial cells via a motif resembling a component of bacterial toxins (2) and centered around aspartic acid residue 20 (D20). Another group has reported a vasopermeability enhancing fragment of IL-2 extending from residues 22 to 58 that increases vascular permeability independent of IL-2 bioactivity (3). A third group of investigators has proposed that activation of cells bearing the intermediate-affinity IL-2 receptor in the vascular compartment leads to inflammatory cytokine release by natural killer (NK) and other cells (4). In the latter case, it was proposed that a receptorselective form of IL-2, that is, one that effectively binds the high-affinity and not the intermediate-affinity IL-2 receptor, would avoid this activation in the vascular compartment, but provide IL-2 to activated T cells expressing the high-affinity receptor.
This approach of selective receptor activation is not unlike the use of low-dose IL-2 in the clinic that has been shown to result in immune activation with far less side effects (5). Surprisingly, this treatment regimen seems to result in the expansion of a subset of CD56 bright NK cells rather than T cells, at least in the circulation, so it is not clear whether this approach can effectively stimulate an antitumor T-cell response. It is generally accepted that this treatment approach results in fewer long-term clinical responses (1).
One potential reason for this lack of efficacy with lowdose protocols may be because of the inability to deliver sufficient levels of IL-2 to the tumor microenvironment, where it is needed to activate antitumor T cells. We and others have been using antibody targeting of cytokines (immunocytokines) such as IL-2 to deliver these immune stimulators to the tumor microenvironment with the hope of increasing efficacy without associated toxicity (reviewed in ref. 6). Numerous studies have documented that it is possible to generate (or enhance) potent T-cell responses to tumors in mice and that these responses are mostly MHC class I-restricted CD8 T-cell responses and are dependent on specific tumor targeting. Unlike many studies with IL-2, our treatment is based on short-term dosing (e.g., 3-5 successive days followed by at least 2 weeks without dosing) rather than repeated dosing over extended periods. The consequence is that we do not provide IL-2 to cells after their initial activation that includes upregulation of highaffinity IL-2 receptor and enhanced responsiveness. In this way toxicity is minimized, but, apparently, the targeting aspect results in efficacy that is not possible with short-term dosing with nontargeted IL-2. In fact, many of our studies have included the use of IL-2 (at equal or higher doses) that was found to be completely ineffective in situations where our targeted molecules were completely effective in eradicating tumor cells (7). In this study, we used an antibody NHS76 that was selected from a phage display library by using Raji Burkitt's lymphoma nuclear extracts as the binding ligand (8) to target the DNA-histone complex in the necrotic core of tumors. Such tumor necrosis treatment antibodies have been shown to be effective in targeting tumors in animal models and cancer patients (9,10).
Despite the ability to target IL-2 to tumors, systemic administration of immunocytokines still exposes the cells in the vascular compartment to IL-2 prior to their delivery to the tumor. Therefore, we have investigated ways of reducing their potential toxicity to increase the doses that can be administered. We first investigated changes in the proposed toxin motif (discussed above) by testing mutations of D20 of IL-2 that were reported to maintain normal IL-2 bioactivity. One such mutation to threonine (D20T) has little or no effect on activity of free IL-2 but would be predicted to eliminate the toxin motif responsible for endothelial cell binding (2). Surprisingly, when this form of IL-2 was expressed as a whole antibody immunocytokine, we found it to be highly specific for activating the high-affinity IL-2 receptor. The resulting molecule would be expected to have 2 beneficial properties, including selectivity for activated T cells and reduced binding to endothelial cells. In this study, we describe the characterization of this molecule in vitro and in vivo and show for the first time that a form of IL-2, with receptor specificity for the mouse high-affinity IL-2 receptor, can show antitumor responses without the normal levels of IL-2 toxicity. An immune stimulator with such a low side effect profile is a good candidate for combination with traditional chemoand radiotherapies that by themselves have immunepotentiating activity (11).

Cell lines and animals
The mouse myeloma NS/0 cell line was obtained from the European Collection of Animal Cell Cultures. The murine T-cell line CTLL-2 and the human astrocytoma U-87 MG were obtained from American Type Culture Collection (ATCC). The murine neuroblastoma NXS2 was a gift from Dr. Ralph Reisfeld at Scripps Research Institute. The murine Lewis lung carcinoma (LLC) was a gift from the late Dr. Judah Folkman at Children's Hospital, Boston. The human T-cell line Kit-225 (K6) was provided by Dr. Angus Grant, EMD Pharmaceuticals, Durham, NC. The human TF-1b cell line was provided by Dr. Paul Sondel, University of Wisconsin at Madison, WI. All cell culture media were purchased from Invitrogen and all cytokines from R&D Systems. C57BL/6 (female), A/J (female), severe combined immunodeficient mice (SCID) CB17 mice (male), and BALB/c (female) mice (8-9 weeks old) were purchased from Taconic Farms and Jackson Lab.

Expression and purification of NHS-IL2 immunocytokines
The pdHL vector used for the expression of NHS-IL2 immunocytokines was described previously (12). The genes encoding the V regions of the NHS antibody were based on the protein sequences of NHS76, a single-chain Fv derived from screening a human scFv phage library against Raji Burkitt's lymphoma cell nuclear extracts (8). The VH and the VL were chemically synthesized, using optimized codons for mammalian expression. The VH was joined to the human IgG2 (immunoglobulin G 2) constant regions, which was in turn fused in frame to the sequence encoding the mature IL-2 in the vector. At the fusion junction, the Cterminal lysine residue of the CH3 was changed to alanine to increase serum half-life (13). The sequence of the VL

Translational Relevance
The approach of selective interleukin 2 (IL-2) receptor activation such as the use of low-dose IL-2 in the clinic and the screening for IL-2 mutants that selectively activate the high-affinity receptor is aimed at immune activation with less side effects. We have been using antibody targeting of cytokines to deliver immune stimulators to the tumor microenvironment to minimize systemic toxicity. In this study, we describe the characterization of such an immunocytokine called Selectikine, which contains an IL-2 (D20T) variant that is selective for both the human and mouse high-affinity IL-2 receptors, and, hence, translational research can be done in mouse models to evaluate antitumor efficacy and toxicity. Selectikine is currently in phase I clinical trial, and this potent tumor-targeting immune stimulator with a low side effect profile is a good candidate for combination with traditional chemo-and radiotherapies that by themselves have immune-potentiating activity.
revealed that it was derived from a human lambda chain and contained the J3 region, hence it was joined to the constant region of human lambda 3, which was used to replace the human kappa chain in the pdHL vector. The D20T mutation in the IL-2 moiety and the N297 mutation in the CH2 domain were introduced by overlapping PCR. The antibody-IL2 fusion proteins were expressed in NS/0 cells by transfection and selection of producer cell clones as described (13). Proteins were purified from conditioned cell culture media by binding to and elution from protein A Sepharose, followed by diafiltration into PBS.

IL-2 bioactivity assays
The IL-2 activities of IL-2-containing immunocytokines were tested in standard T-cell proliferation assays by using the mouse CTLL-2 cell line (14), human T-cell line Kit-225 (K6; ref. 15), or human TF-1b cell line (16), which are all dependent on IL-2 for growth. The CTLL-2 cell line expresses the high-affinity mouse IL-2 receptor, the Kit-225 (K6) cell line expresses the high-affinity human IL-2 receptor (17), and the TF-1b cell line expresses only the intermediate-affinity human IL-2 receptor. For the Kit-225 assay, proliferation was measured by the reduction of Alamar Blue instead of 3 H thymidine uptake (18). Briefly, washed Kit-225 cells that had been starved for 4 days in AIM-V serum-free media (12,500 cells/well) were incubated with IL-2 or IL-2-containing immunocytokines for 36 hours. Alamar Blue (Trek Diagnostics Cleveland) was then added (30 mL/well) and the incubation continued for an additional 34 hours. The fluorescence was then read by using a fluorescence plate reader (excitation 530 nm, emission 590 nm). The maximum response used for ED 50 calculation was attained by using an internal reference standard in each assay. The ED 50 concentration was calculated by using least squares analysis (TREND analysis from Excel).

Mouse NK cell isolation and bioassay
Cells were harvested from spleens taken from C57BL/6 mice and NK cells enriched by using a kit from Stem Cell Technologies. NK cells were separated from the rest of the cells through negative selection by using a depletion cocktail tailored to highly enrich NK cells from suspensions of murine spleen cells. NK cell purity was measured by using cell staining for NK1.1 and DX5, known mouse NK cell markers. Cells were incubated with IL-2 or IL-2-containing immunocytokines for a total of 96 hours, with 3 H thymidine added during the last 16 hours. The cells were harvested from the wells with water onto glass microfiber filter plates and radioactivity was measured by liquid scintillation counting.

PBMC proliferation and FACS analyses
Peripheral blood mononuclear cell (PBMC) from normal donors were prepared by Ficoll gradient centrifugation and labeled with carboxyfluorescein diacetate succidimidyl ester (CFSE, Molecular Probes Inc.) for 15 minutes at 37 C. After washing with culture medium (RPMI containing 10% FBS), 2 Â 10 6 cells were cultured per mL, alone or with added IL-2 or immunocytokine. In some cases, cultures included OKT3 anti-CD3 antibody at a final concentration of 0.1 mg/mL. Cell samples were collected at the indicated time points and stained with the following antibodies: anti-CD56PE, anti-CD4PE, anti-CD8PE, and anti-CD25FITC (fluorescein isothiocyanate) according to the instructions of the manufacturer (BD Pharmingen). After incubation on ice for 30 minutes and washing, labeled cells were analyzed by using a Coulter Epics XL-MCL flow cytometer.

Pharmacokinetic analysis
BALB/c mice were injected with 25 mg of an immunocytokine in a volume of 0.2 mL in the tail vein by using a slow push. At various time points, small blood samples were taken by retro-orbital bleeding and collected in tubes coated with heparin to prevent clotting. After centrifugation to remove the cells, the plasma was assayed by capture with anti-human IgG H&L antisera and detection with an anti-human IL2 antibody (12). Results were normalized to the initial concentration in the serum of each mouse taken immediately after injection.

Pilot toxicity study in cynomolgus monkey
The animal experiment was approved by the local authorities and was conducted in compliance with the principles of good laboratory practice (GLP) and the local animal welfare and health regulations. In this pilot monkey toxicity study (conducted by MDS Pharma Services), a group of 2 male and 2 female cynomolgus monkeys (Macaca fascicularis), with an age of 25 to 27 months at the beginning of treatment, were treated in a 3-week cyclic regimen that reflected the intended clinical dosing scheme of NHS-IL2LT. Animals were treated by a 1-hour intravenous infusion on 3 consecutive days (days 0-2), followed by an 18day treatment-free period (21-day cycle) for 3 cycles. The vehicle control was 0.9% (v/v) NaCl solution. NHS-IL2LT was administered at 1, 3, and 10 mg/kg/d in 0.9% (v/v) NaCl solution. Animals were analyzed for standard toxicity parameters, that is, clinical observations, body weight determinations, vital functions (e.g., blood pressure measurements, electrocardiogram), clinical pathology (hematology including lymphocyte subset analysis, serum chemistry, and urinalysis), toxicokinetic and immunogenicity investigations, and histopathology examinations.

Tumor models
Experimental liver or lung metastases were induced by tail vein injection of viable single cells of NXS2 (0.4 Â 10 6 ) or LLC (0.5 Â 10 6 ) in 0.2 mL PBS into A/J mice or C57BL/6 mice (n ¼ 7), respectively, on day 0. Mice received 5 daily intravenous injections of PBS, fusion proteins, or free IL-2 with or without the naked NHS antibody, on days 4 to 8. Metastases were scored and organ weights were measured on day 28 as previously described (12). For depletion studies, mice received intraperitoneal injections of 50 mg of rat IgG2b anti-murine CD4 mAb (monoclonal antibody, clone 2.43, TIB-210; ATCC), 100 mg of rat IgG2b antimurine CD8 mAb (clone GK1.5, TIB-207; ATCC), or 20 mL of NK-cell-specific rabbit anti-asialo GM1 antiserum (WAKO) on day 3 and once weekly thereafter for 3 weeks, which had been shown previously to result in about 95% depletion of the respective T-cell subsets or NK cells by indirect immunofluorescence staining and cytofluorometric analysis of lymph nodes and spleens.

Bioactivity of a mutant IL-2 immunocytokine
We originally tested the D20T mutation of IL-2 based on data suggesting that this change conferred no selectivity for activating either the intermediate-or high-affinity forms of the IL-2 receptor. Such a mutation would allow us to test the effect of removing this key residue in the proposed LDL toxin motif (2) and its possible role in the vascular toxicity of IL-2, independent of its receptor selectivity. Although this mutation has little selectivity as free IL-2 (unpublished data), we found that when it is fused to the carboxyl terminus of the human IgG H chain, the selectivity for the high-affinity receptor-mediated proliferation is increased dramatically. Biological activity, as measured by cell proliferation, was compared by using 2 human cell lines: the Kit-225 T-cell line expressing both intermediate bg and high-affinity abg, and TF-1b, an erythroleukemic line, naturally expressing the common g chain and transfected to express the human b chain. The resulting cell line responds to IL-2 through the heterodimeric bg intermediate-affinity receptor (16). Bioactivity of the final antibody-IL2 fusion molecule, NHS-IL2LT (LT stands for low toxicity), containing the D20T mutation, was compared by dose titrations to both free human rIL-2 and the same immunocytokine bearing wild-type IL2, NHS-IL2. Although the ability of NHS-IL2LT to induce proliferation of Kit-225 cells was reduced 4-to 5-fold relative to free IL-2, the ability to induce proliferation of TF-1b was reduced more than 6,000-fold ( Fig. 1A  selectivity for high-affinity IL-2R of more than 1,000-fold. Because we would be using mouse models for in vivo testing of toxicity and antitumor efficacy, we also examined the receptor selectivity on murine immune cells. In this case, we used the standard IL-2-responsive T-cell line, CTLL-2, for the high-affinity measurement and purified mouse NK cells for measuring response through the intermediate IL-2R. We found that the NHS-IL2LT mutant immunocytokine had an even more profound selectivity for the highaffinity IL-2R ( Fig. 1C and D). We figured that such biological activity assays are more sensitive and meaningful than just measuring binding affinity to the different receptors, because the selectivity for activating the high-affinity receptor may not be completely accounted for by the difference in binding. In fact, we found that NHS-IL2LT retained binding affinities to both the high-and intermediate-affinity IL-2 receptors (Supplementary Fig. S1A and B), and its differential activities on the proliferation of Kit-225 and TF-1b cells seemed to be mainly due to impaired signaling, as measured by phosphorylated Stat5a (Supplementary Fig. S1C and D). NHS-IL2LT attained substantial levels of pStat5a signaling on TF-1b cells only at very high concentrations of about 100 nmol/L to 1 mmol/L (i.e., containing respectively 1,800 and 18,000 ng/mL of IL-2 equivalents; Supplementary Fig. S1D), consistent with the proliferation data of these cells in Figure 1B.
These results are in contrast to earlier studies with the N88R mutant form of IL-2 that reported selectivity for high versus intermediate IL-2R but was later found to show this effect only for human and monkey cells, but not for mouse immune cells (19). Therefore, in the case of NHS-IL2LT, it is justified to use mouse models to determine relative toxicity and efficacy and thereby establish any potential improvements in the therapeutic index. Although not the original intention of our studies, we found it possible to test for the first time in relevant tumor models, the hypothesis that a form of IL-2 with selectivity for the high-affinity IL-2R would be less toxic and still maintain antitumor efficacy. We also found in these studies that immunocytokines with wild-type IL-2 also show a moderate degree of selectivity for the high affinity over the intermediate form of IL-2R, relative to free IL-2 ( Fig. 1). We have also confirmed this to be the case with other immunocytokines utilizing different antibody V regions (data not shown).

IL-2 receptor selectivity in cultured PBMC
Because both mouse and human immune cells responded with the same degree of IL-2 receptor selectivity, we used human PBMC to study differential proliferative responses in CD56 þ NK cells and CD4 þ and CD8 þ T-cell populations. We also monitored expression of the CD25 component of the high-affinity receptor in the same experiment. For proliferation analysis, resting PBMC were labeled with CFSE and cultured with rIL-2 or equimolar amounts of either NHS-IL2 or NHS-IL2LT immunocytokines. Significant proliferation of NK cells was observed in cultures containing rIL-2 and NHS-IL2, as visualized by CFSE dilution in the CD56-gated population on days 6 and 9 of culture ( Fig. 2A). In contrast, little or no proliferation was observed in the culture containing NHS-IL2LT as measured by CFSE dilution. This result is consistent with the fact that little or no CD25 expression was observed in this CD56 þ cell population and that only cells expressing intermediate receptor accounted for the observed proliferation. Despite this lack of proliferation, the number of CD56 þ cells in the culture supplemented with NHS-IL2LT was increased on day 6 compared with the culture with no added IL-2, suggesting that NK cell survival was prolonged by this molecule.
Analysis of proliferation and IL-2R expression in T-cell populations was carried out in the same manner, but in the presence and absence of anti-CD3 antibody stimulation. Again, PBMC were labeled with CFSE and cultured with rIL-2, NHS-IL2, or NHS-IL2LT. In this case, CD4 þ and CD8 þ cells were gated and analyzed separately for proliferation and CD25 expression. In the absence of anti-CD3 stimulation, little or no CD4 þ T-cell proliferation was observed with rIL2 or either immunocytokine (Fig. 2B), because the IL-2Rb chain is expressed constitutively in CD8 T cells, but not in unstimulated CD4 T cells (20). In cultures activated by anti-CD3 antibody, strong proliferative responses were seen by day 6 in all cultures, including those with NHS-IL2LT. Response to this molecule correlated with a strong upregulation of CD25 in the CD4-gated cell population as measured on day 6 of culture (Fig. 2B). Because anti-CD3 activation is known to induce CD25 expression (21), we assume that the CD4 þ CD25 þ cells observed at this time have survived and proliferated in response to the selective NHS-IL2LT molecule, whereas the culture containing anti-CD3 antibody, but no IL-2, would not have maintained expression of CD25 until day 6.
When CD8 þ cells were examined in the same cultures, a somewhat different pattern of response was observed. In this case, cultures without anti-CD3 activation showed proliferative responses to the wild-type forms of IL-2, that is, those capable of triggering both intermediate-and highaffinity receptors, but not to the selective NHS-IL2LT molecule (Fig. 2C). Also unlike CD4 þ cells, some degree of CD8 þ cell proliferation was observed in response to anti-CD3 antibody alone (no added IL-2), but that the addition of either the wild-type or mutant IL-2 molecules increased the number of cell divisions dramatically and to roughly the same extent. Again, the ability of NHS-IL2LT to stimulate cell proliferation was associated with strong induction of CD25 expression in the CD8-gated population in the anti-CD3-activated culture.

Effects on pharmacokinetic behavior in mice
Before comparing wild-type and mutant immunocytokines, it was essential to eliminate other parameters that could effect vascular toxicity and antitumor efficacy, especially changes in pharmacokinetics. We reported earlier that the fusion of IL-2 to an antibody molecule has unpredictable consequences to the pharmacokinetic behavior of immunocytokines in mice and man (22,23). Generally, the distribution a phase in the blood is One culture was activated with anti-CD3 antibody while a second was not. C, CD8 þ cells were gated and analyzed for CD25 expression and CFSE dilution on day 6 of culture. One culture was activated with anti-CD3 antibody while a second was not.
far more rapid than that of a whole antibody and the t 1/2 in the vascular compartment is shorter as well. These effects can be partially abrogated by reducing FcR binding or by modifying the linker between the antibody and IL-2, so that intracellular degradation is reduced (13). The NHS-IL2 immunocytokines described in this study have been optimized for both reduced FcR binding (by using the g2 isotype) and intracellular degradation (by mutating the C-terminal residue of the CH3 domain from K to A). Despite these changes, we found that the mutation of D20 to T had a dramatic effect on clearance from the circulation, primarily as a result of prolonging the distribution a phase [t 1/2 a ¼ 0.

Tolerability of NHS-IL2LT in mice
Our initial tolerability studies utilized immune-competent mice that make high-titered antibody responses to this human fusion protein beginning around day 5. For this reason, we limited intravenous dosing to 5 consecutive days and increased dose amounts until significant body weight loss or death were observed. For NHS-IL2 containing wild-type IL-2, a dosing regimen of 50 mg/mouse/d Â 5 days resulted in animal deaths consistently in multiple experiments (Fig. 3A). This is a lower tolerated dose than we had seen in earlier studies with other immunocytokines, but was likely because of longer circulating t 1/2 of this molecule that led to accumulation with daily administration. In contrast, by using NHS-IL2LT, daily intravenous doses of 1 mg/mouse/d Â 5 could be administered with no animal deaths, however at this point, mice were beginning to show signs of toxicity, including transient weight loss (Fig. 3B). These data show that NHS-IL2LT is greater than 20-fold less toxic than the same molecule containing wildtype IL-2 when administered with this dosing schedule.
Because treatment with IL-2 is expected to modulate IL-2R on immune cells, we also examined longer dosing schedules to see whether there was IL-2R upregulation or expansion of cells with high-affinity IL-2R. Because of antiimmunocytokine antibody responses, we conducted these studies in immune-deficient SCID mice recognizing that the lack of functional B and T cells would likely make any results difficult to interpret in the context of immune competence. In the study shown in Figure 3C, consecutive dosing for 5 days repeated weekly resulted in significant toxicity and death with doses as low as 20 mg per day and with a delay of only about 5 days, compared with the NHS-IL2 wild-type control molecule. Additional pilot experiments in SCID mice identified that the shortest time between 3-day repeated courses of treatment was 21 days (not shown), a schedule that is quite similar to those used for standard chemotherapies. These data show that the benefit of receptor selectivity is lost with continuous dosing, but that intermittent dosing protocols may be able to maintain the low toxicity benefit of this immunocytokine and its potential for use in combination therapies. To test this hypothesis in a more relevant model, where biological responses are more similar to humans, and where immunogenicity should be far less pronounced, we next studied repeated intermittent dosing in nonhuman primates.

Toxicity and pharmacologic activity of NHS-IL2LT in cynomolgus monkey
We assessed NHS-IL2LT in a GLP-compliant pilot toxicology study in cynomolgus monkeys, a relevant species for the assessment of IL-2 toxicities (24). The lowest dose administered (1 mg/kg) corresponded to the highest, but poorly tolerated, dose we have administered to cynomolgus monkeys by using other immunocytokines containing unmutated IL-2. The other doses were 3-and 10-fold higher. The maximum tolerated dose (MTD) of immunocytokines in humans have been determined to be 6.4 and 7.5 mg/m 2 for huKS-IL2 and hu14.18-IL2, respectively, when administered as once-daily 4-hour infusions in a 3-day cycle repeated every 4 weeks (23,25). This MTD range for humans corresponds to about 0.52 to 0.61 mg/kg for monkeys based on allometric scaling (26). However, the actual MTD of IL-2 containing immunocytokines in monkeys was determined empirically to be several-fold lower, about 0.1 mg/kg, and monkeys and, importantly, humans dosed at their respective MTDs had very similar C max and AUC (23,25, and unpublished data).
Monkeys were treated by using the intended clinical regimen, that is, a 21-day cycle consisting of treatment (1 hour, intravenous infusion) on 3 consecutive days (days 0-2) followed by an 18-day treatment-free period, which is a more aggressive schedule with the anticipated lower toxicity of Selectikine. Toxicokinetic analysis showed a systemic and dose-proportional exposure of all NHS-IL2LT-treated animals. The peak plasma levels were in the range of 103 to 161 mg/mL in high dose treated male and female animals, which are many times higher than the peak plasma levels of 2 to 5 mg/mL obtained at the MTD of huKS-IL2 and hu14.18-IL2 in humans (23,25), and in monkeys (unpublished data). Although most animals developed low levels of antibodies against NHS-IL2LT, this fact did not affect its kinetic properties and, therefore, drug exposure. NHS-IL2LT elicited no overt clinical signs or changes in body weight. In addition, no treatment-related effects at the injection sites or on cardiovascular parameters were observed. Treatmentrelated alterations with respect to hematology, biochemistry, and pathology were mild to moderate; the major target organs being the lymphatic system, liver, kidney, and intestine with lymphoid cell infiltrations. In summary, NHS-IL2LT induced moderate and typical IL-2-derived toxicity (27) and was clinically well tolerated by using the intermittent dosing scheme, defined in our mouse models, up to the high dose of 10 mg/kg tested.
In addition to the comprehensive toxicity assessment, we also evaluated some pharmacologic IL-2 response in the cynomolgus monkeys to prove the relevance of this species and to show pharmacologic activity of NHS-IL2LT in vivo. We first analyzed the response of total lymphocytes, Tlymphocytes (CD3 þ ), and NK (CD8 þ CD16 þ ) cells to NHS-IL2LT treatment. The total lymphocytes (Fig. 4A) and T cells (Fig. 4B) first showed a very mild lymphopenia (considering these high doses) followed by a strong lymphocytosis around days 5 and 7 that were increased (2-to 3-fold) over baseline, compared with the vehicle control. Interestingly, all dose groups showed essentially the same degree of total lymphocyte and T-cell expansions indicating saturation of receptors at the lowest dose level tested. NK cells showed a somewhat different pattern of response (Fig. 4C). In this case, the degree of lymphopenia was much stronger as a percentage of total NK cells and lymphocytosis in the lowest dose group was clearly reduced compared with the 2 higher-dose groups. In fact, the peak level after recovery in the low-dose group was essentially a return to baseline. This level of response was not different from what was observed in the vehicle control group.
We then looked at more specific IL-2 markers such as sIL2-Ra and CD25-positive (activated) CD4 þ and CD8 þ lymphocytes in serum. The soluble form of the high-affinity IL-2 receptor subunit a increases in serum concomitant with its cellular expression. Although the function of sIL-2Ra is unclear, it correlates with increased T cell and immune system activation (28). It has been shown previously that mutated IL-2 molecules are able to induce activated (CD25 þ ) T-helper (CD4 þ ) and cytotoxic T-cells (CD8 þ ) in nonhuman primate species (4). Cynomolgus monkeys showed a typical and strong IL-2 response with a sharp and dose-dependent increase in sIL-2Ra one day after the last treatment in each cycle (days 3, 24, and 45; Fig. 4D), whereas CD25-positive CD4 þ (Fig. 4E) and CD8 þ (Fig. 4F) cells increased on day 5 in each cycle. The peak expansion of activated CD4 þ and CD8 þ cells and the increase of sIL-2Ra was 20-to 40-fold compared with the vehicle control. Overall, the effects were reversible and returned to baseline values at the end of each cycle (days 20 and 41). These results confirmed our observations in vitro and show that NHS-IL2LT also has a selective pharmacologic activity on CD4/CD8-positive cells in nonhuman primates.

Antitumor activity in mouse models
Next, we tested whether treatment with a low-toxicity form of IL-2 could show efficacy in mouse tumor models. We were also interested in testing whether tumor necrotic targeting could be effective in the setting of minimal residual disease. Previous studies of the uptake of radiolabeled antibodies in mouse tumor models had shown that tumor necrosis can be used to target micrometastases in well-oxygenated organs, when lesions are greater than 20 to 30 cells in diameter (9). At this stage, cells undergoing apoptosis are known to express DNA or DNA-containing complexes on their surface (29,30). Furthermore, for tumor cells with high rates of mutation and daughter cell deaths, in vitro experiments by using tumor spheroids to model micrometastases had shown that 20% to 30% of the cells in the proliferating rim are nonviable and necrotic (31).
To test antitumor activity, we induced experimental lung metastases by intravenous injection of a lethal dose of LLC cells in immune-competent BL/6 mice or experimental liver metastases by intravenous injection of NXS2 neuroblastoma cells in immune-competent A/J mice. Treatment was initiated on day 4 when only small metastases were established. For these studies, we limited dosing to 5 consecutive days to avoid treatment after IL-2R upregulation and thereby maintained receptor selectivity. Animals were sacrificed when control mice showed signs of toxicity from tumor burden (approximately day 28) and lungs or livers were removed for analysis of tumor burden by 2 criteria. Surface metastases were apparent after staining, but most surface lesions tended to fuse and could not be easily counted. Therefore, surface metastases were semiquantitated by estimation of percent surface coverage. For the LLC model, we first correlated this assessment method with tumor burden as measured by histologic examination after hematoxylin and eosin (H&E) staining and found a strong correlation between percent surface coverage and overall tumor burden (Fig. 5A and B). In this experiment, potent antitumor activity was shown with a daily dose of 80 mg and an intermediate level of activity was observed by using 20 mg per dose. The control groups treated with either the same amounts of IL-2 or combinations of IL-2 and the NHS antibody showed no activity at all in this model.
In the NXS2 neuroblastoma model, we measured both percent of surface coverage by metastases and overall organ weight as a measure of tumor burden. Again, we confirmed that diseased livers consisted of tumor mass by H&E staining (not shown). Therefore, tumor-containing liver weight (approximately 5% of body weight) in excess of normal liver weight was taken as a measure of tumor mass. Under these conditions, we found that equal doses of NHS-IL2 and NHS-IL2LT showed nearly the same level of activity as assessed by the percent of surface coverage by tumor (Fig. 6A), and both treatment groups were significantly   different from the PBS control (P < 0.001). Treatment was somewhat better with NHS-IL2, but this was not significant (P ¼ 0.18). When data were assessed by organ weights-a more objective measure of tumor burden-the similarity between the treatment groups (shown as individual animal organ weights) were even more striking. Recombinant IL-2 had no antitumor activity in the same model (not shown).
To assess what effector cells were responsible for antitumor activity, we tested the effect of antibody depletion of CD4, CD8, or NK cells in the LLC experimental metastasis model. Depletion of the respective cell type was confirmed by fluorescence-activated cell sorting (FACS) analysis of peripheral blood for each antibody. The in vivo efficacy results showed that depletion of CD4 þ cells did not abrogate the antitumor effect of the 2 immunocytokines containing either wild-type or mutated IL-2 (Fig. 6B) and may have improved their activities, although the difference was not statistically significant (P ¼ 0.16 for NHS-IL2 and 0.096 for NHS-IL2LT). This improvement in antitumor activity suggests that elimination of CD4 þ CD25 þ Treg (T regulatory) cells might make this treatment approach more effective.
In contrast to the results with CD4 depletion, treatment with anti-CD8 antibody resulted in a complete loss of antitumor activity for both NHS-IL2 and NHS-IL2LT (Fig. 6B), indicating that CD8 þ cells are the primary effec-tors in this model. This is in agreement with most of our earlier results in other syngeneic tumor models (6). The results with NK cell depletion were less dramatic and suggested some role for these cells in antitumor activity (Fig. 6B), possibly providing a supporting function to the more essential CD8 þ effector cells. Interestingly, there was no difference between the groups treated with NHS-IL2 and NHS-IL2LT with respect to NK cell depletion, despite the dramatic selectivity differences in cytokine response in vitro and toxicity in vivo.

Discussion
Several factors can play a role in the ability of a cytokine to effectively activate the immune system against a foreign agent or a tumor cell. One of the first cytokines to be approved for treatment of cancer is IL-2, despite the fact that its biology is one of the most complex and, in some cases, contradictory to its proposed role in immune stimulation. Not only does IL-2 potently stimulate NK and T cells to expand in numbers and increase their cytolytic activity, but in the case of T cells, it sensitizes them to activationinduced cell death (32) and is required for Treg cells to suppress ongoing immune responses (33 through the production of secondary cytokines and overactivation of NK cells-both of which can result in vascular injury (34).
The initial goal of this study was to engineer a form of IL-2 as an immunocytokine with reduced potential for causing side effects while retaining the ability to stimulate an antitumor immune response. An earlier report from another group reported such a study with free IL-2 containing an N88R mutation (4); however, it was later found that the basis for the reduced toxicity (selectivity for highaffinity over intermediate IL-2R) was not maintained in the mouse tumor model used to show efficacy (19). In our case, we tested the biological and antitumor activity of a different IL-2 mutation (D20T) that was found by others to be nonreceptor selective in the context of free IL-2. This aspartic acid residue is the critical part of a toxin-like domain that is believed to be responsible, in part, for direct vascular toxicity of IL-2. Surprisingly, we found that the D20T mutation, in the context of a whole antibody immunocytokine, is highly selective for the high-affinity IL-2R (as measured by IL-2-induced proliferation) for both human and mouse immune cells. Thus, for the first time, it was possible to test the hypothesis of whether such a selective molecule could be both less toxic and retain antitumor activity in a mouse model. Results indicate that the D20T form of IL-2, contained in the NHS-IL2LT molecule, retained the majority of its antitumor activity when tested in syngeneic mouse tumor models of experimental metastases. Even when it was dosed at higher levels than the control, NHS-IL2 molecule containing wild-type IL-2, such doses represent a far greater therapeutic index. For example, we showed that for a single-dosing cycle, the NHS-IL2LT molecule is roughly 20-fold less toxic than NHS-IL2 (containing normal IL2). In the tumor efficacy experiments, we found no more than a 4-fold increase in dose was required to achieve the same level of efficacy. Thus, the therapeutic index was improved a minimum of 5-fold. The tolerability of NHS-IL2LT was also tested in cynomolgus monkeys and shown to be dramatically improved over what we have observed with similar immunocytokines containing normal IL-2. The lowest dose in these studies was 1 mg/kg that resulted in sustained blood levels of NHS-IL2LT of several mg/mL-a level approaching the ED 50 of this molecule for the intermediate IL-2R. Interestingly, this level showed a diminished ability to increase circulating levels of NK cells (following a brief lymphopenia), compared with the 2 higher doses. T cells, on the other hand, appeared to respond just as well at the low dose and proliferated to a much greater extent than NK cells. This expansion correlated to the upregulation of CD25 on both CD4 and CD8-positive T cells, as well as an increase in sIL-2R in the blood. We do not know yet whether the cells responding at the lower dose were already expressing CD25 or whether the initial response was stimulated through intermediate receptor-mediated induction of CD25 on resting cells. In any case, the overall effect was to selectively expand both T-cell subsets, especially those expressing CD25, and to do this at increasingly high-dose levels in the apparent absence of typical IL-2 side effects.
How this selectivity for high-affinity IL-2 receptor in NHS-IL2LT (Selectikine) may translate to antitumor activity and lower toxicity remains to be tested clinically. In fact, Selectikine (EMD 521873) is currently in phase I clinical trial. The inherent problem with immune therapy is that it works less well as tumor burden increases and, in the case of NHS-IL2LT, we used a targeting approach (anti-DNA) that might be expected to require large tumors as a source of antigen. The fact that even small metastases could be treated with this approach, likely because of the appearance of DNA-containing structures on apoptotic cells, allowed us to show antitumor activity in this minimal disease setting after a single-dosing cycle, and this activity was due mostly to CD8 þ effector cells. Only low titers of antihuman antibodies were seen in monkeys after an 8week treatment, and these have not limited the number of cycles of NHS-IL2LT that can be administered. Therefore, it is likely that multiple cycles of treatment in the clinic could be administered safely and with the potential for improved efficacy. One challenge with IL-2 in general, and with forms specific for the high-affinity receptor in particular, is the possibility of generating more Treg cells than antitumor effectors. In this regard, it was interesting to note that our antitumor activity in the Lewis lung model was improved when mice were treated with anti-CD4 antibody, although significant antitumor activity was seen without CD4 depletion. Nonetheless, many approaches to reducing Treg cells (11) could be combined with NHS-IL2LT, including many standard chemotherapies and local radiation. The fact that its administration, even at high doses, is associated with only mild side effects makes such an approach clinically attractive.

Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.