Direct Hematological Toxicity and Illegitimate Chromosomal Recombination Caused by the Systemic Activation of CreER^T21

Conditional gene inactivation using the Cre/loxP system has become increasingly used in reverse mouse genetics (1, 2, 3, 4, 5, 6). This system takes advantage of the bacteriophage P1 Cre-recombinase ability to catalyze the excision of a DNA sequence flanked by loxP sequences. Inactivation of the target gene in conditional knockout mice is regulated depending on the expression pattern of Cre recombinase under the control of tissue-specific promoters (7, 8). However, to analyze gene functions in adult mice, additional temporal control of gene inactivation is indispensable to circumvent problems such as embryonic lethality or developmental abnormalities arising from the early onset of Cre recombinase activity.

Recently, temporal regulation of Cre recombinase activity has been accomplished using tetracycline-controlled gene expression and IFN-inducible expression (9, 10). Another approach uses engineered recombinase fused to the mutated ligand-binding domain of the estrogen receptor (ER^T2),³ which does not bind endogenous estradiol but is highly sensitive to the synthetic ligand tamoxifen (TM) or its metabolite 4-hydroxytamoxifen (11). The fusion protein is inactivated by binding to heat shock proteins, until the administration of TM, when it is released from the complex, becomes active and excises loxP-flanked DNA regions. Several transgenic mouse lines have been generated that express CreER^T2 fusion genes under the control of tissue-specific promoters, which show ligand-dependent recombination in certain cell types (12, 13, 14, 15, 16, 17).

It has long been assumed that the expression of Cre recombinase does not adversely affect the physiology of the host cell, despite several reports alarming the toxicity of Cre recombinase. High levels of Cre expression have been reported to be toxic in some mammalian cells. Mouse embryonic fibroblasts, NIH3T3 cells, and some human cell lines can be sensitive to the continuous presence of Cre (18, 19, 20). Regarding the adverse effect of Cre in vivo, aberrant chromosomal rearrangement in spermatids and male infertility has been reported in transgenic mice expressing Cre in postmeiotic spermatids (21). Another group reported dilated cardiomyopathy in their transgenic mice expressing Cre under the control of α-myosin H chain (22). Glucose impairment was also reported in commonly used transgenic mice expressing Cre under the control of the insulin 2 promoter (23).

CreER^T2 was first expected to avoid adverse effects of Cre recombinase by keeping Cre recombinase inactive until the administration of TM. However, very recently, Naiche et al. (24) reported that systemic activation of CreER^T2 results in lethal anemia and widespread apoptosis in embryos. In their report however, it is unclear whether the toxicity is the direct effect of CreER^T2 or not.

In this study, we demonstrated widespread hematological toxicity of ubiquitously expressed CreER^T2 in adult mice as well as in embryos. We analyzed two independent mouse lines in which CreER^T2 is inserted into the R26 locus (R26CreER^T2 mice), and both lines showed reduced proliferation, increased apoptosis, and illegitimate chromosomal rearrangement in hematopoietic cells after the administration of TM possibly due to the toxicity of CreER^T2. Flow cytometric analysis revealed that CreER^T2 toxicity affected several hematopoietic lineages, and that immature cells in these lineages tend to be more sensitive to the toxicity. In vitro culturing of hematopoietic cells from these mice further demonstrated the direct toxicity of CreER^T2 on growth and differentiation of hematopoietic cells. We further demonstrated the in vivo cleavage of the putative cryptic/pseudo loxP site in the genome after the activation of CreER^T2, which results in the illegitimate chromosomal rearrangement. These results emphasize the critical importance of including control mice carrying the Cre gene to avoid the misinterpretation of the results.

We further demonstrated that these hematological abnormalities in adult R26CreER^T2 mice recover spontaneously after 1 mo, which ensures the availability of this mice in reverse mouse genetics provided appropriate control mice are included.

One line of R26CreER^T2 mice was purchased from ARTEMIS Pharmaceuticals (17). Another line of R26CreER^T2 mice was generated by Regeneron Pharmaceuticals using Velocigene technology (25), essentially as described. The R26 locus was heterozygous for the CreER^T2 knock-in in all experiments. Both lines were derived on a mixed 129/Svj and C57BL/6J background, and the contribution of 129/Svj was the same in every experiment. R26R Cre reporter mice (R26R mice) (26) were purchased from The Jackson Laboratory. All animal experiments were performed in accordance with the Institutional Guidelines, and were in accordance with National Institutes of Health guidelines.

TM (Sigma-Aldrich) was dissolved in a sunflower oil/ethanol (9/1) mixture at 3.0 mg/ml. In adult R26CreER^T2 mice at 8 to 12 wk, 35 mg/kg or 175 mg/kg TM was administered orally for indicated days depending on the experiments. One to four days later depending on experiments, mice were sacrificed and subjected to experiments. Wild-type (WT) littermates administered with the same amount of TM were used as controls. For R26CreER^T2 embryos, 200 mg/kg TM was administered i.p. into pregnant mothers at E14.5, or 150 mg/kg TM was administered at both E13.5 and E14.5 depending on the experiments.

The following Abs were used: anti-Ter119, anti-Mac1 (M1/70), anti-Gr-1 (RB6-8C5), anti-B220 (RA3-6B2), anti-CD19 (1D3), anti-IgM (R6-60.2), anti-CD3 (145-2C11), anti-CD4 (L3T4), anti-CD8 (Ly2), anti-CD25 (PC61), and anti-c kit (2B8) were obtained from BD Pharmingen. Anti-Ter119, anti-Mac1, anti-Gr-1, anti-B220, anti-NK1.1, anti-CD3, anti-CD4, and anti-CD8, were used as Lin markers. Anti-Ter119 (BD Pharmingen) and anti-Ki67 (Novocastra) Abs were used for immunostaining.

The organs were fixed in Tris-buffered 10% formalin solution and embedded in paraffin. Sections (2 μm) were stained with H&E. β-gal staining was performed as previously described (27). For immunostaining, the specimens were fixed in 4% paraformaldehyde at 4°C overnight, serially soaked in 10, 20, and 30% sucrose/PBS and embedded in OCT compound (Tissue Tek, Sakura Finetech) and 6-μm sections were prepared. The sections were immunostained as previously described (28, 29, 30).

Sections were subjected to the TUNEL staining using the in situ apoptosis detection TUNEL kit (MK500, Takara) and visualized by reaction with 3,3′-diaminobenzidine (SK-4100; Vector) for 1 min.

To assess the effect of CreER^T2 activation on the differentiation of hematopoetic stem cells, Lin⁻c-kit⁺ cells were collected from the bone marrow of R26CreER^T2 mice and WT littermates, cultured on a monolayer of TSt-4 cells (31), and administered with 1 μM of 4-OHT (Sigma-Aldrich) at various time points. We analyzed the differentiation of the Lin⁻c-kit⁺ cells by examining the expression of Ter119 for erythroid potential, CD19 for B cell potential, and Mac-1 for myeloid potential. To assess the effect of CreER^T2 activation on the proliferation of differentiated hematopoetic cells, Mac-1⁺ cells, and CD19⁺ cells were collected, cocultured with TSt-4 cells, and administered with 4-OHT.

R26CreER^T2 mice and WT littermates were treated with vehicle or TM for 5 consecutive days, and were sacrificed 3 days after the last administration. Bone marrow cells of these mice were cultured and chosen randomly for chromosomal number analysis (47 cells for R26CreER^T2 mice treated with TM, 20 cells for WT mice treated with TM, and 50 cells for R26CreER^T2 mice treated with vehicle) and for karyotype analysis (14 cells for R26CreER^T2 mice treated with TM, 4 cells for WT mice treated with TM, and 7 cells for R26CreER^T2 mice treated with vehicle) as previously described (32).

Primers were designed around the candidate locus for cryptic/pseudo loxP site to detect the amount of intact genome. Sequences for primers were described in corresponding figure legend. Real-time PCR was performed with a 7700 Sequence Detection System (Applied Biosystems) using SYBR Green PCR amplification reagent (Applied Biosystems), and the results were normalized with the amount of GAPDH genome.

Data are presented as means ± SD. Statistical significance was assessed by nonpaired, nonparametric Student’s t test.

First, we tested the recombination efficiency of R26CreER^T2 mice using R26R Cre reporter mice (26) (R26R mice), which express lacZ after Cre-mediated excision of a neo cassette. Adult R26R/R26CreER^T2 mice were treated with 175 mg/kg body weight TM for 5 consecutive days, and the recombination of the lacZ reporter was analyzed 4 days after the last administration (Fig. 1,A). Southern analysis of genomic DNA from different organs showed up to 50% recombination (50% in the liver, 30% in the kidney), without detectable background activity in untreated animals, as reported previously (17) (data not shown). Whole-mount tissues from R26R/R26CreER^T2 mice demonstrated strong β-gal expression in almost all tissues (Fig. 1,A) except for brain (data not shown). No background recombination was observed in R26R/R26CreER^T2 mice treated with vehicle (data not shown). The recombination efficiency was also tested during embryogenesis. Female R26R mice mated with male R26CreER^T2 mice bearing E14.5 embryos were injected i.p. with 200 mg/kg TM. Two days later, tissues from E16.5 embryos were stained with X-gal, demonstrating that only the tissues from double transgenic mice exposed to TM showed β-gal expression (Fig. 1 B).

We first noticed the toxicity of R26CreER^T2 mice when we tried to knockdown the expression of BMP-4 in embryogenesis using R26CreER^T2 mice, and administered TM to pregnant BMP-4^flox/flox mice (33) bearing BMP-4^flox/flox; R26CreER^T2 embryos and BMP-4^flox/flox; WT embryos. In this experiment, 150 mg/kg TM was administered for 2 consecutive days (Fig. 2^,A) to achieve complete recombination in both alleles in BMP-4^flox/flox mice. Four days after the last injection, we analyzed the embryos, and observed severe anemia in BMP-4^flox/flox; R26CreER^T2 embryos, but not in BMP-4^flox/flox; Cre ⁻ embryos (data not shown). To test whether the phenotype in BMP-4^flox/flox; R26CreER^T2 embryos was due to the deletion of BMP-4 gene or due to the systemic activation of CreER^T2, we administered the same amount of TM to pregnant WT mice bearing R26CreER^T2 embryos and WT embryos without a floxed allele. Four days later, R26CreER^T2 embryos without a floxed allele showed severe anemia as well (Fig. 2^,B), indicating that the anemia was not due to the deletion of floxed alleles, but is due to the toxicity of CreER^T2. The livers of R26CreER^T2 embryos looked pale (Fig. 2^,C), and body weight as well as liver weight of R26CreER^T2 embryos was lower compared with those of WT embryos (Fig. 2^,D). R26CreER^T2 embryos treated with vehicle did not show anemia, or the reduction in body weight or liver weight. Histological analysis demonstrated the colonization of erythroblasts in the liver in WT embryos in late embryogenesis, while the number of erythroblast was significantly reduced in R26CreER^T2 embryos (Fig. 2, E and F). These hematological changes in R26CreER^T2 embryos were already evident at E16.5 (supplementary Fig. 1),⁴ while the body weight reduction was not observed yet.

Next, we administered TM to adult R26CreER^T2 mice and WT littermates according to the protocol shown in Fig. 3,A. R26CreER^T2 mice administered TM developed severe thymus atrophy, but not R26CreER^T2 mice treated with vehicle, nor WT mice treated with TM (Fig. 3,A). Thymus weight normalized to body weight was significantly reduced in R26CreER^T2 mice treated with TM (Fig. 3,A), which was consistent with the reduced cell density in the cortical region of the thymus (Fig. 3,B). R26CreER^T2 mice treated with TM also exhibited hypocellular bone marrow, and a decrease of erythroblasts in the red pulp of the spleen (Fig. 3,B, ∗), but the cell density in the white pulp of the spleen was not changed. We also analyzed whether the strains recover from the hematological abnormality, and demonstrated that the extent of recovery from the hematological toxicity greatly differed among individual mice 2 wk after the administration of TM (supplementary Fig. 2), while all R26CreER^T2 mice recovered completely after 1 mo (Fig. 3 B).

We further analyzed the hematopoietic lineages affected by the toxicity. Adult R26CreER^T2 mice and WT littermates were treated according to the protocol used in Fig. 3, which exerts severe hematological toxicity in R26CreER^T2 mice. Numbers of cells in the thymus, bone marrow, and spleen decreased in R26CreER^T2 mice after TM treatment (Fig. 4 A).

FACS analysis in the thymus demonstrated that CD4⁺CD8⁺ double positive cells were significantly reduced in R26CreER^T2 mice (Fig. 4 B, DP). In addition, the numbers of the cells in double negative subsets in c-Kit/CD25 profiles of Lin⁻ fraction were reduced in R26CreER^T2 mice.

We also analyzed Ter119/Mac-1, Gr-1 profile and B220/IgM profile of bone marrow cells (Fig. 4,B). The numbers of myeloid cells (Mac-1, Gr-1 positive cells), erythroblasts (Ter119⁺ cells) and immature B lymphocytes (B220⁺/IgM⁻ cells) were significantly reduced in the bone marrow of R26CreER^T2 mice, while the number of mature B lymphocytes did not change (Fig. 4 B). Together with that the number of CD4⁺CD8⁺ double positive cells was significantly reduced in the thymus of R26CreER^T2 mice, immature cells might be more sensitive to the toxicity of CreER^T2.

To analyze the toxicity in the peripheral tissues, we further examined Ter119/Mac-1, Gr-1 profile and B220/IgM profile in the spleen (Fig. 4 B). Similar to the results in the bone marrow cells, the numbers of myeloid cells and erythroblasts decreased in the spleens of R26CreER^T2 mice, but not the number of mature B lymphocytes.

To define the nature of the toxicity of CreER^T2, we analyzed apoptosis and cell proliferation in the hematopoietic tissues in adult R26CreER^T2 mice treated with TM according to the protocol in Fig. 3. These mice were sacrificed at the last day of administration, when viable cells still remain in the hematopoietic tissues (Fig. 5). The numbers of Ki67-positive cells were reduced both in thymus and spleen of R26CreER^T2 mice, while the numbers of TUNEL-positive cells were increased in spleen, but not in thymus of R26CreER^T2 mice.

Considering high rate of apoptosis during thymocyte maturation, we postulate that the loss of immature thymocytes in R26CreER^T2 mice (Fig. 4) might reduce the number of “native” apoptosis, and mask the increased apoptosis due to the toxicity.

Therefore, we conclude that the toxicity of CreER^T2 is due to attenuated proliferation and increased apoptosis.

To exclude the possibility that the hematological abnormality observed in R26CreER^T2 mice is caused secondarily to unknown systemic disorders, we analyzed the direct effect of TM on hematopoietic cells obtained from R26CreER^T2 mice. First, we isolated lineage marker negative (Lin⁻) c-kit⁺ cells from bone marrow and cultured these cells with erythropoietin to induce differentiation into erythroid cells in the presence or absence of 4-hydroxytamoxifen (4-OHT). Ter119⁺ cells were not generated when 4-OHT was administered to the cells from R26CreER^T2 mice (Fig. 6,A). Next, we cultured Lin⁻c-kit⁺ cells on a monolayer of stromal cell line TSt-4, which efficiently supports the generation of B and myeloid cells, for 14 days, and 4-OHT was administered to the culture at various time points (Fig. 6,B). The generation of B cells, as examined by the expression of CD19, was significantly reduced by the administration of 4-OHT to the cells from R26CreER^T2 mice, but not the generation of myeloid cells determined by the expression of Mac-1 (Fig. 6, B and C). Finally, we analyzed the toxicity of CreER^T2 in already differentiated hematopoietic cells. We isolated Mac-1⁺ cells and CD19⁺ cells from bone marrow and cultured them on a monolayer of TSt-4cells in the presence or absence of 4-OHT. The number of CD19⁺ cells was significantly reduced when 4-OHT was administered to the cells from R26Cre-ER^T2 mice, while the number of Mac-1⁺ cells was not affected (Fig. 6 D).

As the endonuclease activity of Cre is reported to cause chromosomal aberrations and growth arrest in MEF in vitro (18), we analyzed whether the chromosomal aberrations are caused in vivo in hematopoietic cells in R26CreER^T2 mice. R26CreER^T2 mice and WT littermates treated with vehicle or TM for 5 consecutive days were sacrificed 3 days after the last administration (Fig. 7,A), and bone marrow cells were analyzed for chromosomal numbers and karyotype. In R26CreER^T2 mice treated with TM only 53% of the cells showed a normal diploid chromosome number of 40 (Fig. 7,B), while 90% of the cells had 40 chromosomes in WT mouse treated with TM as well as in R26CreER^T2 mouse treated with vehicle. In karyotype analysis, 78% of bone marrow cells from R26CreER^T2 mouse treated with TM displayed chromosomal aberrations including chromosome exchanges (Fig. 7,C, 1), chromatic exchanges (Fig. 7,C, 2), and chromatid breaks (Fig. 7C, 3), while no chromosomal aberrations were observed in bone marrow cells from WT mice treated with TM (Fig. 7 C) or in bone marrow cells from R26CreER^T2 mice treated with vehicle (data not shown).

Thyagarajan et al. (35) reported that mammalian genome contains several candidates for cryptic/pseudo loxP sites, and that one locus in mouse genome AF033025 (GenBank) serves as an active site for the Cre recombinase. To clarify whether inappropriate cleavage at cryptic loxP sites occurs after the activation of CreER^T2, we designed real-time PCR primer sets around the cryptic/pseudo loxP site in AF033025 locus to detect the amount of intact AF033025 locus (Fig. 7 D). Intact AF033025 locus in the thymus of three of four R26CreER^T2 mice was almost undetectable after the administration of TM, indicating illegitimate cleavage at the cryptic/pseudo loxP site due to the activation of CreER^T2. The amount of intact AF033025 locus did not change until the administration of TM, excluding the possibility that the gene targeting procedure to generate R26CreER^T2 allele altered the locus.

In this study, we demonstrated that the administration of TM to R26CreER^T2 mice causes severe growth arrest, apoptosis, and illegitimate chromosomal rearrangement in hematopoietic cells, even in the absence of genes targeted by loxP sites. We tested two independent lines of R26CreER^T2 mice from different facilities, and the results were essentially the same. Furthermore, both strains recovered from the toxicity within a month. We also performed in vitro culturing of hematopoietic cells from these mice and demonstrated direct toxicity of CreER^T2 on growth and differentiation of certain cell types.

Previous reports regarding the adverse effects of Cre in vivo could not exclude the possibility that the unexpected phenotypes were due to the disruption of the genome loci where transgenes were integrated. On the contrary, the lines in this report are alleles introduced into the well-characterized R26 locus, and the disruption of the locus was proved not to cause adverse effects. In addition, no hematological abnormalities were detected until the administration of TM, indicating that an effect of the R26 locus is not likely to be the cause. Importantly, the hematological abnormalities was not due to the toxicity of TM, because the administration of TM to WT mice in vivo as well as to the hematopoietic cells from WT mice in vitro did not exert any effect. Therefore, we concluded that the hematological abnormalities observed in this report were due to the systemic activation of CreER^T2, which arrested cell proliferation and induced apoptosis (Fig. 5), and were the direct effect on hematopoietic cells (Fig. 6).

The cause of these hematological abnormalities after the systemic activation of CreER^T2 is likely to be Cre-mediated genomic rearrangements as observed in Fig. 7, perhaps at cryptic or pseudo-loxP sites within the mouse genome, which have recently been shown to serve as substrates for Cre recombinase (34, 35). Thygarajan et al. (35) reported that the sequences in mouse genomes considerably divergent from the consensus loxP sites serve as functional recognition sites for Cre mediated recombination, and the recombination efficiency of one locus (AF033025) was considerably high in bacterial assays. We further demonstrated that intact AF033025 locus in three of four R26CreER^T2 mice was almost undetectable after the administration of TM (Fig. 7 D). Furthermore, recent bioinformatics analysis estimated the frequency of cryptic loxP sites in the mouse genome is 1.2 per megabase, and are homogeneously distributed throughout the genome.

High sensitivity of hematopoietic cells to the systemic activation of CreER^T2 might be due to their rapid proliferation rate, because the genome in rapidly proliferating cells are more easily accessible by CreER^T2 than the tightly packed genome in quiescent cells. FACS analysis also demonstrated that immature proliferating cells in each hematopoietic lineage tend be more sensitive to CreER^T2 toxicity (Fig. 4 B). Positive correlation between Cre-induced toxicity and proliferation was previously reported in fibroblasts (18) as well as in transgenic flies (36). In addition to hematopoietic cells, intestinal epithelial cells also proliferate rapidly, and R26CreER^T2 mice occasionally demonstrated diarrhea and intestinal edema after the administration of TM, possibly due to the toxicity of CreER^T2 in rapidly proliferating intestinal epithelial cells (data not shown).

Sensitivity to the toxicity of CreER^T2 might also be influenced by the amount of CreER^T2 translocating to nuclei, which is defined by the level of CreER^T2 expression as well as dose and tissue distribution of TM. Seibler et al. (17) previously reported relatively high expression of CreER^T2 in thymus, where we observed severe toxicity.

In previous reports demonstrating adverse effects in Cre transgenic mice, the authors suggested that the inducible form of Cre might be tolerated better because it stays outside the nucleus until induction (23). However, our results in this study indicated large amount of activated CreER^T2 was also able to cause cell toxicity. Although the growth arrest is prominent in CD19⁺ cells from R26CreER^T2 mice after the administration of TM (Fig. 6), no hematological abnormalities have been reported in well-characterized CD19-Cre mice, in which Cre recombinase is highly expressed in B cells. One explanation for the discrepancy is that DNA damage in the cells bearing Cre recombinase induces the cells to develop DNA repairing system to counteract the damage, and such systems might be established in CD19-Cre mice, while R26CreER^T2 mice are not prepared when massive amounts of CreER^T2 would be suddenly activated and cause DNA damage.

Although the hematological abnormality in R26CreER^T2 mice might compromise the phenotypic analysis of the gene of interest, the strain is still of great value because of its efficient inducibility without leakage, and of ubiquitous expression of CreER^T2.

In this study, we suggest three points to take note of to make good use of this strain. One way to solve the problem is taking appropriate control for Cre toxicity: the use of the same mouse without floxed allele. In spite of the fact that Cre toxicity has been occasionally documented in the literatures, it seems still to be widely neglected. A recent study has systemically reviewed the use of RIP-Cre mice, which alone display glucose intolerance, and demonstrated that in more than half of the cases, the appropriate control was not included (23).

Second, it is better to postpone the analysis of the mice for at least 1 mo after the administration of TM. The hematological abnormalities will have diminished after 1 mo (Fig. 3 B), possibly due to the proliferation of the surviving cells.

Third, it is better to minimize the dose of TM. The toxicity in R26CreER^T2 mice was dependent on the dose of TM, which regulates the inducibility of CreER^T2. The minimal dose of TM to induce efficient recombination varies between target alleles, depending on the number of floxed alleles, the distance between loxP sites, the expression level of the target gene, and local chromatin structure. One should adjust the minimal dose of TM to induce efficient recombination in the gene of the interest (supplementary Fig. 3). In cell culture analysis, changing the medium after incubation for 6 h with 4-OHT minimizes the toxicity with efficient recombination (data not shown). The experiments where high recombination efficiency is not necessary or even desirable, such as lineage tracing and mosaic oncogene activation, might be ideal for R26CreER^T2 mice. The self-excising Cre vectors might be another option to reduce the toxicity (37, 38, 39).

The result of the study warns of the potential consequences of Cre-mediated recombination between cryptic loxP sites in the genome in Cre/loxP based technologies in human gene therapy protocols. Paradoxically, however, immature and rapidly proliferating cells are more susceptible to the toxicity caused by the activation of CreER^T2, indicating the possible therapeutic implication of the technology for cancer treatment. Schimidt-Supprian et al. (40) reported that the activation of CreER^T2 transgene in c-Myc-driven primary B cell lymphoma leads to death of lymphoma at lower dose of TM compared with our experiment. Because the dose of TM they used in their experiment does not exert severe toxicity in healthy hematopoietic cells (data not shown), selective eradication of malignant cells might be possible. In addition, R26CreER^T2 mice might be useful as an inducible model for hematological abnormalities caused by aberrant chromosomal rearrangements.

We thank Drs. Y. Kaziro, Y. Nabeshima, S. Takeda, and T. Sakurai for valuable comments and discussion. We also appreciate Drs. D. Sakata, T. Matsuoka, N. Watanabe, and S. Narumiya for providing the experimental equipment.

The authors have no financial conflict of interest.