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

Kenneth D. Lute, Kenneth F. May Jr, Ping Lu, Huiming Zhang, Ergun Kocak, Bedrick Mosinger, Christopher Wolford, Gary Phillips, Michael A. Caligiuri, Pan Zheng and Yang Liu

Data supplements

Article Figures & Data


  • Figure 1.

    Creation of human CTLA4 knock-in mice. (A) Schematic diagram of the structure of construct. The primer positions for screening the floxed and deleted genotypes are also illustrated. PCR Reaction A used primers outside the loxP sites spanning the Neo/TK gene. A successful excision (deleted) of Neo/TK produced a 1.1-kb fragment, whereas undeleted (floxed) Neo/TK did not produce a fragment due to the PCR conditions used. PCR Reaction B used a forward primer outside of and a reverse primer within the Neo/TK cassette. (B) Southern blot of DNA from ES cells transfected with the human CTLA4 construct. A 7-kb band represents successful homologous recombination with the human CTLA4 construct, whereas a 4.7-kb band represents an unaltered mouse Ctla4 gene. (C) Excision of Neo/TK by Cre-recombinase. As depicted schematically in panel A, Reaction A produced the expected 1.1-kb fragment, whereas Reaction B amplified no fragment, consistent with successful deletion of Neo/TK. (D) Expression of human and mouse CTLA-4 RNA in homozygous (left panel) and heterozygous (right panel) knock-in mice. Spleen cells from human CTLA4+/– and human CTLA4+/+ mice were stimulated for 30 hours in vitro with 0.1 μg/mL anti-CD3 mAb 2C11. RNA was extracted and RT-PCR was performed. Primers spanning the full-length CTLA4 RNA sequence were used to confirm that full-length RNA of the knock-in gene was being expressed (left reaction), whereas those that were specific for either mouse (mE2) or human (hE2) CTLA-4 exon 2 were used to identify mouse and human CTLA4, respectively.

  • Figure 2.

    Codominant expression of human and mouse CTLA-4 protein by T cells from human CTLA-4 knock-in heterozygotes. (A) Codominant expression of human and mouse CTLA-4 in T cells after antigen stimulation. Spleen cells from human CTLA-4+/–×P1CTL F1 mice were stimulated for 66 hours in vitro with 0.1 μg/mL P1A peptide. Cells were harvested and stained for cell surface mouse CD3, followed by intracellular mouse and human CTLA-4. The top left panel shows the codominant expression of human and mouse CTLA-4 protein on the same cells as indicated by the diagonal staining pattern. Non–knock-out littermates demonstrated a complete lack of human CTLA-4 expression (bottom left panel). Middle and right panels show isotype controls for each intracellular antibody. All profiles represent cells within the CD3+ gate. The same staining pattern has been observed with anti-CD3 mAb–stimulated T cells (K.M., unpublished observations). (B) Expression of mouse and human CTLA-4 molecules in unstimulated spleen CD4 T cells. Spleen cells from WT (CTLA-4 mo/mo), homozygous (human CTLA4+/+) and heterozygous (human CTLA4+/–) mice were surfaced-stained with anti-CD4 and anti-CD25 and then stained for intracellular mouse and human CTLA-4 protein. Data shown were gated CD4+CD25 (top panels) and CD4+CD25+ subsets (bottom panels).

  • Figure 3.

    Functional replacement of mouse Ctla4 with the human CTLA4 gene in vivo. (A) Normal appearance of secondary lymphoid organs in 1-year-old homozygous human CTLA4 knock-in mice. (B) Normal expression of activation markers among spleen CD4 and CD8 T cells. Data shown are dot plots of gated CD8 (top panels) and CD4 (bottom panels) T cells.

  • Figure 4.

    Anti–human CTLA-4 antibodies with different potency in delaying tumor growth. (A) Growth kinetics of MC38 tumors in minimal disease model. CTLA-4 (hu/hu) mice were challenged with MC38 (5 × 105/mouse) in the lower abdomen. Two days later, the mice received either control mouse IgG or anti–CTLA-4 antibodies K4G4, L1B11, or L3D10 and the tumors were measured every 3 to 4 days. Data shown represent means and standard error of the mean (SEM) of tumor volumes until day 55, when some mice in antibody-treated groups reached their tumor burden end point (n=4). (B) Log transformation of tumor volume. The tumor growth over time was analyzed using the StataR XTGEE (cross-sectional generalized estimating equations) model. Six tests were done to compare the exponential slopes. All mAbs significantly delayed the growth kinetics of tumors (P < .001). In addition, significant delay of tumor growth was observed in mice that received L3D10 in comparison to those that received either L1B11 or K4G4 (P < .001). (C) Kaplan-Meier survival curves of mice that received either control IgG or one of the anti–CTLA-4 antibodies. Complete rejection of tumors was observed in 2 of 9 mice in the L3D10-treated group. A log-rank test revealed that the 3 mAbs significantly prolonged mouse survival (P < .001–P = .004). Data shown in panels A and B are representative of those from 2 independent experiments. Data in panel C involve 8 to 9 mice per group.

  • Figure 5.

    L3D10 treatment delays growth of established tumors in human CTLA-4 knock-in mice. MC38 tumor cells were injected subcutaneously into the human CTLA-4 knock-in mice. At 10 to 14 days after tumor injection, when the tumors reached a mean diameter of 8 mm, the mice were injected with either L3D10 or control Ig every 4 days for 4 weeks. (A) Growth kinetics of established tumors in mice treated with either control IgG or L3D10 (n = 9). Data shown are means and SEM of tumor volumes. The volumes of large holes caused by necrosis in some mice were subtracted. Student t tests were used to compare the tumor size at each time point; those with P < .05 are indicated with an asterisk (*), whereas those with P < .01 are indicated with 2 asterisks (**). (B) Kaplan-Meier survival curves of mice that received control IgG or L3D10. A log-rank test revealed that L3D10 significantly prolonged mouse survival (P = .011).

  • Figure 6.

    Autoimmune side effects associated with different anti–CTLA-4 antibodies. Serum samples from mice that received anti–CTLA-4 treatment, as detailed in the Figure 2 legend, were collected on day 30 (A) and day 55 (B) and tested for anti-dsDNA antibodies. Data shown are means and standard deviations (SDs) of OD at 490. (C) Correlation between tumor growth suppression and anti-DNA antibodies in control IgG, L1B11, and K4G4, but not in L3D10-treated mice. Data shown are the means and SEM of tumor sizes and OD 490 of ELISA test using a 1:270 dilution of sera from tumor-bearing mice. Tumor size and anti-DNA antibody levels reflect data collected at 30 days after tumor challenge. The relative strength of anticancer immunity and autoimmunity has been repeated in 2 independent experiments involving 8 to 9 mice per group.

  • Figure 7.

    Anti–CTLA-4 antibodies with distinct antitumor and autoimmune effects bound to an overlapping site on CTLA-4 and blocked B7-1/CTLA-4 interaction. (A-C) Cross-competition. Unlabeled anti–CTLA-4 antibody (100 μg/mL) was added to plates coated with CTLA-4 Ig. Given concentration of the biotinylated antibodies were added to the wells after 10 minutes. The amounts of biotinylated antibodies bound were determined by adsorption of horseradish peroxidase (HRP)–labeled streptavidin to the plates. Data shown are means and SEM of OD 490. (D) All anti–CTLA-4 antibodies used in the study block B7-1–CTLA-4 interaction. Chinese hamster ovary (CHO) cells transfected with human B7-1 were incubated with a mixture of CTLA-4 Ig and given anti–CTLA-4 antibodies. After washing away the unbound antibodies, the binding of CTLA-4 Ig was determined by flow cytometry using APC-labeled goat anti–human CTLA-4 antibody. Data shown are histograms depicting CTLA-4 Ig binding to human B7-1–transfected CHO cells. Frozen sections of kidney were analyzed after the mice were euthanized, when they reached early removal criteria (tumors reach 4000 mm3), with the exception of 2 mice in the L3D10-treated group in which tumors never reached the criteria for early removal. The incidences of IgG deposition in mice treated with K4G4 (P = .029) and L1B11 (P = .029), but not L3D10 (P = .47) are significantly higher than the control group.


  • Table 1.

    Incidence of antibody and complement C3 deposition in kidney glomeruli

    IgG 4/8* 4/8* 2/9 0/9
    C3 0/8 0/8 1/9 0/9
    • Frozen sections of kidney were analyzed after the mice were killed (ie, when the mice reached early removal criteria [tumors of 4000 m3]), with the exception of 2 mice in the L3D10-treated group, in whom tumors never met the criteria for early removal.

    • * The incidence of IgG deposition in mice treated with K4G4 (P = .029) and L1B11 (P = .029), but not L3D10 (P = .47), is significantly higher than in the control group.