Human CTLA4 knock-in mice unravel the quantitative link between tumor immunity and autoimmunity induced by anti–CTLA-4 antibodies

Antibodies have emerged as one of the most valuable immunotherapeutics for cancer.¹  Therapeutic antibodies can be divided into 2 categories. The first category of antibodies binds directly to cancer cells.^1-3  This binding results in the death of cancer cells by immune-dependent and/or -independent mechanisms.⁴  The second category of antibodies causes tumor rejection by binding to and activating cells of the immune system, such as the T lymphocytes.^5,6  Because this category of antibodies targets lymphocytes regardless of antigen specificity, a major concern of immunotherapy based on this category of antibodies is the risk of severe autoimmune side effects.⁷ 

A prominent example of a category II therapeutic antibody is the anti–cytotoxic T-lymphocyte antigen 4 (CTLA-4) antibody.⁶  CTLA-4 is the high-affinity receptor for B7-1 and B7-2.^8,9  Anti–CTLA-4 monoclonal antibodies (mAbs) have been shown to promote antitumor immunity against a variety of tumors including colon carcinoma,¹⁰  fibrosarcoma,¹⁰  prostate cancer,^11-13  melanoma,^14-16  ovarian carcinoma,¹⁷  mammary carcinoma,¹⁸  and myeloma.¹⁹  These observations have led to enthusiasm for the translation of CTLA-4 antibody therapy to human cancer. More recently, an anti–human CTLA-4 mAb has been generated and tested in clinical trials of patients with advanced ovarian cancer and melanoma.^20,21  In one of these trials, anti–CTLA-4 mAb induced grades 3 and 4 autoimmune toxicities.²⁰ 

To facilitate translation of this concept, it would be helpful to establish preclinical models to identify anti–human CTLA-4 antibodies that can induce anticancer immunity with acceptable autoimmune side effects. Unfortunately, in vitro cultures of human T cells have proven to be an unsuitable model, as the same antibody can have opposite effects on different clones of T cells in the same culture.²²  We have recently reported the use of the human peripheral blood lymphocyte–severe combined immunodeficient (PBL-SCID) model to screen for therapeutic anti–CTLA-4 antibodies in vivo.²³  While this model allows us to demonstrate the protective effect of the antibody against human Epstein-Barr virus (EBV) lymphoma, it does not permit us to evaluate autoimmune side effects. Taking advantage of the fact that human CTLA-4 is capable of interacting with mouse B7-1 and B7-2,^9,24  we created a mouse with a knock-in of the human CTLA4 gene. Using this model we compared the autoimmune side effects and cancer immunity of 3 anti–human CTLA-4 antibodies. Surprisingly, the antibody that induced the most potent cancer immunity provoked the least autoimmune side effects. These results demonstrate that autoimmunity does not quantitatively correlate with cancer immunity and that selective tuning of cancer immunity over autoimmunity is possible with careful choice of antibodies.

Anti–human CTLA-4 monoclonal antibodies L3D10, K4G4, and L1B11 have been described previously.²³  Antibody was purified from hybridoma culture supernatant using a Protein G column. Mouse immunoglobulin G (IgG) was purchased from Sigma (St Louis, MO).

The P1 clone containing a 100-kb murine CTLA4 gene was purchased from Genomic Systems (St Louis, MO). A 3.8-kb DNA fragment containing the 5′ promoter region, exon 1, and part of intron 1 of the murine Ctla4 gene was amplified using 2 primers: CTGAAGCTTCAGTTTCAAGTTGAG, which corresponded to a sequence starting at base 734 of the 5′ promoter region, and TTGGATGGTGAGGTTCACTC, which corresponded to base 4524 of the exon 2 region. The polymerase chain reaction (PCR) product was digested with HindIII and the 3.0-kb fragment was cloned into a HindIII-digested pFlox vector (Dr Raj Muthusamy, Children's Hospital, Columbus, OH). The vector was a 6.5-kb plasmid containing a neomycin resistance gene/HSV (herpes simplex virus) thymidine kinase gene cassette flanked by loxP sites.

DNA containing a 14-kb fragment of the human CTLA4 gene was prepared from a lambda phage clone^25,26  and digested with the restriction enzyme HindIII. A 3.2-kb HindIII fragment containing part of intron 1, exon 2, intron 2, and exon 3 of the human CTLA4 gene was purified and inserted into a HindIII-digested pBluescript plasmid. Plasmid DNA with the insert in the correct orientation was linearized by XhoI digestion and partially digested with BamHI to obtain a 3.2-kb BamHI fragment for use in further cloning. The pFlox plasmid containing a 3-kb exon 1 of mouse CTLA-4 was linearized by XhoI digestion and partially digested with BamHI. The 9.5-kb fragment was purified and ligated with a 3.2-kb fragment of human CTLA4 exons 2 and 3.

A 2.9-kb DNA fragment containing part of intron 3, exon 4, and part of the 3′ sequence of the murine Ctla4 gene was cloned from the P1 clone using primers ATCCTCTAGAAGCTTCAAAGCAGGTTATCA, corresponding to base 6160 through base 6181 of intron 3, and TCTAGTCGACCACAGAGAGTCAAGGCCCTG, corresponding to base 8617 through base 8588 of the 3′ region. The PCR product was digested by XbaI and SalI and inserted into the pFlox clone containing mouse Ctla4 exon 1 and human Ctla4 exons 2 and 3. The final construct is illustrated in Figure 1A.

Embryonic stem (ES) cell line R1 was transfected with the DNA construct by electroporation (Figure 1A), and drug-resistant ES cell colonies were obtained as described.²⁷  To verify that homologous recombination had occurred, DNA was extracted from ES clones for analysis by PCR. Fragments were amplified using forward primer CCAAGACTCCACGTCTCCAG, corresponding to a region upstream of exon 1 of the mouse Ctla4 gene that is outside of the region used in the transgene construct, and reverse primer CCTCTGAGCATCCTTAGCAC, corresponding to a region in exon 2 of the human CTLA4 gene. These 2 primers gave rise to a PCR product of 3.3 kb only when the human exon was inserted into the mouse Ctla4 gene by homologous recombination. Eight of 153 DNA samples screened were positive for this product. The positive clones were analyzed by Southern blot to further confirm homologous recombination. Briefly, the genomic DNA from PCR-positive and -negative ES clones were isolated, digested with EcoRI, and transferred to Nylon membrane (Osmonics, Westborough, MA). A 0.9-kb probe was generated by PCR, targeting the region upstream from exon 1 between the EcoRI and HindIII sites using primers CTGCAGTGAACACCCCTCTC and ACGTCTCCAGGTCCTCAGAG. The probe was labeled with ³²P using the DECAprime DNA labeling kit (Ambion, Austin, TX), and hybridized to the membrane. The blot was exposed to BIOMAX MS film (Kodak, Rochester, NY) with a Kodak HE intensifying screen for 2 days at –70°C. The endogenous murine Ctla4 gene yielded a band of 4.7 kb, whereas homologous recombination yielded a band of 7.0 kb by the replacement of the 0.9-kb murine exon 2 with the 3.2-kb human exons 2 and 3.

To remove the Neo-TK selection cassette, we transfected ES cells of clone no. 63 with the pCre-Pac plasmid described by Taniguchi et al,²⁸  by electroporation. Two sets of PCR reactions were carried out to detect the floxed and deleted alleles of the CTLA4 locus. The first PCR reaction used 5′-TCCCTCTCAGACACCTCTGC-3′ as the forward primer and 5′-GTCATAAACATCTCTCAGGTAA-3′ as the reverse primer. This reaction amplified the alleles in which Neo/TK had been deleted with a product of 1.1 kb. While this reaction should theoretically also amplify the endogenous murine Ctla4 alleles, the PCR conditions used did not allow amplification of a large product of 4 kb. The second PCR reaction used 5′-TCCCTCTCAGACACCTCTGC-3′ as the forward primer and 5′-CGACCTGTCCGGTGC-3′ as the reverse primer.

Chimeric mice were prepared by an aggregation method essentially as described.²⁷  The chimera mice were bred to C57BL/6 mice to obtain founders with germ-line transmission of CTLA4 knock-in allele. The founders were then backcrossed to either C57BL/6 or BALB/c background for at least 6 generations. Homozygous mice were used for screening anti–human CTLA-4 antibodies.

P1CTL transgenic mice expressing a T-cell receptor specific for the P1A35-43:L^d complex have been previously described.²⁹  BALB/c and C57BL/6 mice were purchased from Charles River Laboratories under contract from the National Cancer Institute. All mice were maintained in the University Laboratory Animal Research Facility at the Ohio State University under specific pathogen-free conditions. MC38 colon carcinoma cells were purchased from American Type Culture Collection (Manassas, VA).

Spleen cells were obtained from human CTLA4^+/– and human CTLA4^+/+ C57BL/6 mice and stimulated with anti-CD3 (2C11; 0.1 μg/mL) for 30 hours in culture. Following culture, total RNA was isolated and a reverse transcriptase (RT)–PCR was performed with cDNA from stimulated splenocytes to amplify the full-length CTLA4 sequence from exons 1 to 4. The forward primer began at base pair 5 on mouse exon 1 (5′-CTTGTCTTGGACTCCGGAGGTAC-3′) and the reverse primer at base pair 652 on mouse exon 4 (5′-AAGGCTGAAATTGCTTTTCACATTC-3′) for a total amplified fragment size of 648 base pairs. To determine the coordinate expression of both mouse and human CTLA4 genes, cDNA was amplified with forward primers specific for mouse (5′-TGTGCCACGACATTCACAGA-3′) or human exon 2 (5′-GAGGCATCGCCAGCTTTGTG-3′) and a common reverse primer for mouse exon 4 (5′-CACATAGACCCCTGTTGTAAGA-3′). The amplified fragment using forward and reverse primers for mouse Ctla4 was 354 base pairs (bp), whereas the fragment using a forward primer for human CTLA4 was 455 base pairs. Forward and reverse primers for hypoxanthine phosphoribosyltransferase (HPRT) were used as an internal control and gave rise to a 100-bp product.

To determine if human (hu) CTLA-4 protein was expressed properly, we bred P1CTL transgenic mice with huCTLA4^+/– mice to create P1CTL⁺huCTLA4^+/– mice. Freshly harvested spleens from these mice were stimulated with 0.1 μg/mL P1A peptide and harvested after 66 hours in culture. Spleen cells were stained with fluorescein isothiocyanate (FITC)–conjugated anti–mouse CD3, phycoerythrin (PE)–conjugated anti–mouse CTLA-4 (intracellular), and CyChrome-conjugated anti–human CTLA-4 (intracellular). To further confirm that CTLA-4 protein was appropriately regulated we also stained naive spleens from wild type (WT), human CTLA4^+/–, and human CTLA4^+/+ C57BL/6 mice. Spleen cells were stained with PerCP-conjugated anti–mouse CD4, FITC-conjugated anti–mouse CD25, PE-conjugated anti–mouse CTLA-4 (intracellular), and allophycocyanin (APC)–conjugated anti–human CTLA-4 (intracellular). Conjugated antibodies and CytoFix/CytoPerm intracellular staining kit were purchased from BD Pharmingen (San Diego, CA).

Mice used for tumorigenicity studies have been backcrossed to C57BL/6 for at least 6 generations. MC38 cells (5 × 10⁵) suspended in serum-free RPMI (100 μL) were injected subcutaneously in the lower abdomen of mice. For the minimal disease model, mice were treated once a week beginning on day 2. In the established disease model, mice were treated every 4 days with treatments beginning 10 to 14 days after challenge. In both models, the tumor-bearing mice received identical doses of either anti–human CTLA-4 mAb or control mouse IgG (200 μg/mouse per injection). Tumor size and incidence were determined every 2 to 5 days by physical examination. The tumor volume was calculated using the following established formula: volume = ¹/₂ (long × short² ). All mice were killed when the tumor volume reached 4000 mm³. The number of days required for tumors to reach this end point was used for survival analysis.

Anti-DNA antibodies were measured by enzyme-linked immunosorbent assay (ELISA) according to published procedure.³⁰ 

Frozen sections of kidney were prepared from euthanized mice and fixed in acetone. After blocking with 10% normal goat serum, the sections were stained with Rhodamine-conjugated goat anti–mouse IgG and FITC-conjugated goat anti–mouse C3 antibodies (ICN Biomedicals, Santa Ana, CA).

The gene encoding CTLA-4 is composed of 4 exons in both mice and humans, with 76% overall homology between murine and human CTLA-4 proteins and 100% homology between their cytoplasmic domains.^25,26  Since human CTLA-4 is able to bind to murine B7-1 and B7-2,^9,24  it is likely that the interaction of human CTLA-4 and murine B7 would maintain normal signal transduction by CTLA-4. We have created a chimeric DNA construct in which the exons coding for the extracellular (exon 2) and transmembrane (exon 3) domains of murine CTLA-4 have been replaced with those of human CTLA-4 (Figure 1A). As the gene product of exon 1 is a signal peptide not expressed in the mature protein, and the cytoplasmic domain (exon 4) is completely conserved between human and mouse, replacement of only exons 2 and 3 was required to create a humanized CTLA-4 knock-in mouse.

We transfected an embryonic stem (ES) cell line R1 with the human CTLA-4 DNAconstruct in Figure 1A by electroporation.After selection with G418, DNAwas isolated from the drug-resistant ES cell clones and screened by PCR for homologous recombination, which was then confirmed by Southern blot (Figure 1B). Probing for a sequence at the 5′ end of the CTLA4 gene, homologous recombination of the human CTLA4 knock-in gene yielded a band of 7.0 kb, whereas the endogenous mouse CTLA-4 gene yielded a band of 4.7 kb. ES cell clone 63, which had undergone homologous recombination, was transfected with the plasmid pCre-Pac that expresses both the Cre-recombinase and puromycin resistance gene. After selection with puromycin and potential excision of Neo/TK by Cre-recombinase, we further selected with gancyclovir, which eliminated all cells in which the Neo/TK gene was not excised. PCR analysis of DNA from several surviving colonies indicated that the Neo/TK cassette was excised from the knock-in locus (Figure 1C). Based on the analysis of DNA and RNA, ES cell clone 20 was chosen for the production of chimera mice, which were bred with C57BL/6 and BALB/c mice to obtain germ-line transmission of the human CTLA-4 gene. Those mice that have been backcrossed to C57BL/6 for at least 6 generations and homozygous for huCTLA-4 were used for tumorigenicity studies.

To determine whether human CTLA4 was properly expressed and spliced at the RNA level, spleen cells were obtained from human CTLA4^+/– and human CTLA4^+/+ C57BL/6 mice and stimulated with anti-CD3 (2C11) for 30 hours in culture. RNA was extracted from these cells and RT-PCR was performed with cDNA to amplify the full-length CTLA-4 sequence from exons 1 to 4. Primers spanning mouse exons 1 to 4 amplified a band of 648 base pairs. As shown in Figure 1D, the overwhelming majority of CTLA4 mRNA contained exons 1 to 4. In addition, by using primers specific for mouse and human CTLA-4, we were able to observe expression of both mouse and human CTLA4 alleles. As shown in Figure 1D, primers designed to amplify mouse Ctla4 exons 2 to 4 fail to produce a product in the homozygous knock-in mice, further confirming homologous recombination.

To determine whether human CTLA-4 protein was properly expressed, spleen cells from P1CTL⁺huCTLA4^+/– and P1CTL⁺huCTLA4^–/– littermate mice were stimulated in culture with P1A peptide for 66 hours. Cells were harvested and stained for both mouse and human intracellular CTLA-4 expression. As shown in Figure 2A (top panels), both mouse and human CTLA-4 proteins are detected in the human/mouse CTLA4 heterozygous mice. In addition, diagonal distribution of the human and mouse CTLA-4 molecules reveals that the 2 alleles are regulated by the same mechanism (top left). The specificity of the staining was confirmed by both isotype control staining as well as by the lack of binding of anti–human CTLA-4 antibodies in WT mo/mo littermates. To test whether the mouse and human CTLA4 alleles were similarly regulated in vivo, we analyzed freshly isolated spleen cells from WT mice or those that were either heterozygous or homozygous for the human CTLA4 alleles. It has been reported that the only subset that constitutively expresses CTLA4 in the peripheral lymphoid organs are the CD4⁺CD25⁺ regulatory T cells (Tregs).³¹  As shown in Figure 2B (bottom middle panel), among the Tregs diagonal distribution of human and mouse CTLA-4 proteins was observed in the heterozygous mice. As expected, homozygous knock-in mice did not express murine CTLA-4 protein. A minute non-Treg population in all 3 strains of mice expressed appreciable levels of CTLA-4 protein (top row). As with the Treg population observed in the heterozygous mice, these cells also expressed mouse and human CTLA-4 at similar levels as revealed by the diagonal distribution (Figure 2B, top middle panel).

CTLA4 knock-out mice are known to develop profound lymphoproliferative disorder and die within 4 weeks of birth.³²  Our extensive observations have indicated that the homozygous human CTLA4 knock-in mice have a normal life span with no sign of autoimmune disease development over a period of observation longer than one year. As shown in Figure 3A, no enlargement of the lymphoid organs was observed in the human CTLA4 knock-in mice. Moreover, the extent of T-cell activation in vivo was essentially the same as that observed with WT T cells (Figure 3B). Therefore, the human CTLA-4 allele has functionally replaced the mouse CTLA-4 gene. As such, the knock-in mice may be used to study antibodies targeting the human CTLA-4 molecule in vivo.

We have recently described a panel of anti–human CTLA-4 antibodies that promote expansion of human T cells in the human PBL-SCID mouse model. Moreover, the antibody-treated mice survived longer than the control Ig–treated mice.²³  To test whether human CTLA4 knock-in mice are useful in discriminating the therapeutic effect of the anti–CTLA-4 antibodies, we injected colon cancer cell line MC38 subcutaneously into the CTLA4 knock-in mice. Two days later, the tumor cell–bearing mice received either control IgG or 1 of 3 isotype-matched anti–CTLA-4 antibodies. Among them, L3D10 and K4G4 have the same affinity and binding kinetics, whereas L1B11 has approximately 3-fold lower affinity. As shown in Figure 4A, all 3 antibodies demonstrated a statistically significant delay in tumor growth compared with mouse IgG control antibody. In addition, L3D10 proved to be the most potent antibody when compared with the other 2 treatment antibodies (Figure 4A-B). As seen in Figure 4C, all 3 antibodies led to enhanced survival compared with control Ig–treated mice. A survival advantage of L3D10-treated mice was also observed over those treated with L1B11 and K4G4 (Figure 4C).

To explore the therapeutic potential of the L3D10 antibody for large established tumors, we delayed treatment until approximately 2 weeks after tumor cell challenge. As shown in Figure 5, in comparison to the control Ig–treated group, the L3D10 antibody delayed tumor growth, and prolonged survival of tumor-bearing mice. Nevertheless, it should be noted that L3D10 alone did not cause complete tumor rejection. Therefore, it is likely that even the most efficient anti–CTLA-4 antibody will need to be used in combination with other reagents in order to achieve complete rejection of established tumors.

Given the tendency of anti–CTLA-4 antibodies to exacerbate autoimmune diseases in experimental autoimmune models, it is of interest to determine whether the autoimmune side effects quantitatively correlate with antitumor immunity. Our analysis revealed that in wild-type mice, anti–mouse CTLA-4 antibody 4F10 suppressed tumor growth, but enhanced anti–double-stranded DNA antibodies. In contrast, anti–4-1BB antibody 2A induced cancer immunity without triggering anti-DNA antibody response (Figure S1, available on the Blood website; see the Supplemental Figure link at the top of the online article). Thus, the anti-DNA antibodies can serve as a useful marker for autoimmunity associated with anti–CTLA-4 antibody. We first compared mice treated with 3 different anti–CTLA-4 antibodies for their production of anti–double-stranded (ds) DNA antibodies. As shown in Figure 6A-B, although anti-dsDNA antibodies were detected in all tumor-bearing mice treated with anti–CTLA-4 antibodies, the mice that received K4G4 and L1B11 had 3- to 5-fold higher levels of anti-dsDNA antibodies than mice treated with L3D10. The difference was stable over the course of the treatment. Consistent with this variability in anti-dsDNA antibody induction, we observed more IgG deposition in kidney glomeruli of K4G4, L1B11-treated mice than in those treated with L3D10 (Table 1).

A comparison between the amounts of the anti-dsDNA antibody and the sizes of the tumors suggests that for mice that received control Ig, L1B11, or K4G4, tumor size correlated inversely with the amount of anti-dsDNA antibodies. This observation suggests that, among these 3 groups, the intensity of the antitumor immune response correlates with that of the anti-DNA antibody response (Figure 6C). However, the group that received L3D10 treatment had the smallest tumor size with the lowest anti-dsDNA antibody levels. Thus, stronger cancer immunity does not have to be coupled with more severe autoimmune side effects.

As measured by Biacore, L3D10 and K4G4 have essentially identical affinity for human CTLA-4.²³  In addition, these antibodies have identical isotypes (IgG1, κ). To determine whether the antibodies have overlapping binding sites, we tested whether they compete with each other in binding to human CTLA-4. As shown in Figure 7A-C, all 3 antibodies cross-blocked each other's binding to CTLA-4, with efficiency that grossly correlates with their affinity to CTLA-4.²³  Moreover, all antibodies were capable of blocking the binding of CTLA-4 to its natural ligand B7-1 (Figure 7D). The similarity of the immunochemical properties of these antibodies highlights the need for preclinical models to screen for anti–CTLA-4 antibodies with favorable therapeutic activity and acceptable autoimmune side effects.

We have demonstrated that human CTLA-4 gene knock-in mice can serve as a valuable model for the preclinical screening of cancer therapeutic antibodies targeting the human CTLA-4 protein. The utility of this approach is based on 3 factors. First, human CTLA-4 has natural ligands in the mouse, as previously reported.^9,24,33,34  Second, the signaling pathways used by mouse and human CTLA-4 are similar. Although this is difficult to verify because the mechanism of signal transduction for CTLA-4 is still unclear, the fact that the cytoplasmic domain of mouse and human CTLA-4 is 100% identical⁸  suggests that the signaling pathway is likely the same. Third and most importantly, human and mouse CTLA-4 must have the same biologic function. In support of this notion, we have demonstrated that the homozygous knock-in mice do not develop lethal autoimmune diseases, which were observed in the CTLA-4 knock-out mice.³²  At the same time, polymorphisms of both mouse and human CTLA-4 genes affect genetic susceptibility to autoimmune diseases.³⁵ 

Previously, we used the hu-PBL-SCID mouse model to explore the potential efficacy of anti–CTLA-4 antibody treatment.²³  In this model, SCID mice are reconstituted with human peripheral blood, thereby creating a functional human immune system. Although this model is useful in screening antibodies for their potential anticancer effect, it is somewhat limited in evaluating other clinical parameters such as autoimmunity. By comparison, huCTLA-4 gene knock-in mice offer several important advantages, foremost of which is the fact that the immune response takes place in a natural setting. In contrast to knock-in mice, SCID animals require significant intervention to promote and maintain responses. Indeed, in the SCID model human T-cell survival is predicated on repeated injections of anti–natural killer (NK) cell antibodies as well as cytokines such as granulocyte macrophage–colony-stimulating factor (GM-CSF). These factors must be taken into consideration when the therapeutic effects are interpreted. Nevertheless, it is of interest to note that in both models L3D10 treatment led to the most efficacious response. In the SCID model, mice undergoing L3D10 treatment exhibited the most substantial T-cell expansion, as well as the longest survival. Similarly, L3D10-treated huCTLA-4 knock-in mice displayed the most significant reduction in tumor growth among treatment groups, as well as the most enhanced survival benefit.

Perhaps the most important advantage with the knock-in model is our ability to evaluate autoimmune side effects associated with potential human therapeutic antibodies. A previous trial with a humanized anti–CTLA-4 antibody reported considerable side effects, with 43% of patients showing grades 3 to 4 autoimmune toxicity, including dermatitis, colitis/enterocolitis, hypophysitis, and hepatitis.²⁰  In another trial, reactivity to melanocytes in skin and retina was associated with T-cell infiltration and necrosis in tumors.²¹  Autoimmune reactivity in anti–CTLA-4–treated mice has not been systematically analyzed, although depigmentation has been reported.^14,36  Perhaps because of the relative short course of transplanted tumors, the autoimmune side effects in the mouse tumor model are relatively mild. However, a model that recapitulates autoimmune side effects will not only allow us to select antibodies with fewer side effects, but also to develop approaches to abrogate remaining side effects. In this regard, our quantitative comparison of the anti-dsDNA antibody titers and pathologic examination of tumor-bearing mouse kidney have revealed considerable heterogeneity among different antibodies in their autoimmune side effects. Surprisingly, L3D10, which induced the strongest therapeutic effect, provoked the least autoimmune side effects. The discordance between cancer immunity and autoimmunity reveals that autoimmune side effects and cancer therapeutic effects are not quantitatively linked. Such uncoupling provides a theoretical basis for selecting optimal anti–CTLA-4 antibodies or other therapeutic agents with the most desirable balance between cancer immunity and autoimmunity. Nevertheless, it should be pointed out that our extensive search for pathologic changes, including blood chemistry and histologic examination of all major organs, has failed to reveal severe autoimmune diseases in tumor-bearing mice that can be attributed to anti–CTLA-4 antibodies (data not shown), which is different from experience with a clinical trial using a different anti–human CTLA-4 antibody.²⁰ 

Several different mechanisms may be responsible for the differential effects of different anti–CTLA-4 antibodies. For instance, cancer immunity and autoimmunity may involve different effector cells. Alternatively, cancer targets and normal tissues may differ in their resistance to immune attack. In this context, previous studies by others have revealed that even when the antigen is shared between tumor and normal tissue, the antibody doses required for tumor rejection and autoimmune side effects differ.^37-39  Regardless of the immunologic basis, the uncoupling of the quantitative link between autoimmunity and cancer immunity demonstrated here suggests that autoimmunity may not be a necessary price for cancer immunity. These findings provide a theoretical basis for the selective modulation of cancer immunity over autoimmunity.

We thank Lynde Shaw for secretarial assistance, and Jin Wen and Priya Joshi for antibody purification.