A Mouse Model to Assess STAT3 and STAT5A/B Combined Inhibition in Health and Disease Conditions

Genetically-engineered mouse models (GEMMs) lacking diseased-associated gene(s) globally or in a tissue-specific manner represent an attractive tool with which to assess the efficacy and toxicity of targeted pharmacological inhibitors. Stat3 and Stat5a/b transcription factors have been implicated in several pathophysiological conditions, and pharmacological inhibition of both transcription factors has been proposed to treat certain diseases, such as malignancies. To model combined inhibition of Stat3 and Stat5a/b we have developed a GEMM harboring a flox Stat3-Stat5a/b allele (Stat5/3loxP/loxP mice) and generated mice lacking Stat3 and Stat5a/b in hepatocytes (Stat5/3Δhep/Δhep). Stat5/3Δhep/Δhep mice exhibited a marked reduction of STAT3, STAT5A and STAT5B proteins in the liver and developed steatosis, a phenotype that resembles mice lacking Stat5a/b in hepatocytes. In addition, embryonic deletion of Stat3 and Stat5a/b (Stat5/3Δ/Δ mice) resulted in lethality, similar to Stat3Δ/Δ mice. This data illustrates that Stat5/3loxP/loxP mice are functional and can be used as a valuable tool to model the combined inhibition of Stat3 and Stat5a/b in tumorigenesis and other diseases.


Introduction
Genetically-engineered mouse models (GEMMs) lacking gene(s) of interest in the whole animal or in a tissue-specific manner offer a unique opportunity to model the effects of pharmacological inhibitors, as well as to assess toxicity in models of health and disease [1]. Upon identification of a cancer target gene(s), the toxicity and efficacy of future inhibitors can be evaluated by crossing mice harboring alleles of interest flanked by loxP sites (flox mice) with tissue-specific and inducible Cre recombinase expressing mouse strains and deleting the gene(s) in appropriate cancer-mouse models [2]. This strategy offers a straight forward and rigorous method for target validation before undertaking costly and time-consuming development of antitumorigenic targeted inhibitors.
Signal transducer and activator of transcription family members (STAT1, 2, 3, 4, 5A, 5B and 6) are transcription factors which are activated in response to multiple cytokines and growth factors. The binding of ligands to their cognate receptors results in the activation of Janus kinase (JAK) family members (JAK1, 2, 3 and Tyrosine kinase 2, TYK2). Tyrosine phosphorylation of STATs by JAK family members induces homo-and hetero-dimerization, as well as subsequent translocation into the nucleus where STAT proteins bind to their specific DNA sequences and control the transcription of multiple genes [3,4]. STAT3, STAT5A and STAT5B have been implicated in several physiological and disease processes such as cell division, migration, survival, cellular responses to pathogens, breast development, inflammation and cancer [5]. Indeed, the over-activation of STAT3 and STAT5A/B has been reported in human hematopoietic and solid malignancies, suggesting that these transcription factors behave as oncogenic drivers in those tumors [6,7]. These observations have been complemented with functional studies using GEMMs harboring flox alleles for Stat3 [8][9][10][11][12][13] and Stat5a/b [14] genes. Consequently, several research groups have developed inhibitors targeting STAT3 or STAT5A/B to treat malignancies [6]. Unfortunately, monotherapy treatment of tumors almost invariably results in tumor resistance. Tumor cells acquire resistance to drugs by (re)activating alternative or redundant signaling pathways [15]. STAT3 and STAT5A/B have overlapping functions in several (patho)physiological contexts [16][17][18][19]; thus it is expected that tumor cells will acquire resistance to single inhibitors by compensatory mechanisms relaying in the activation of the non-inhibited STAT family member. Therefore, dual inhibitors targeting both STAT3 and STAT5A/B or a combination of single STAT3 and STAT5A/B inhibitors may be more effective [20], albeit perhaps also resulting in increased toxicity. Stat3, Stat5a and Stat5b are genetically-linked genes, clustering within a 160 kb genomic region of mouse chromosome 11. This should preclude modeling simultaneous inhibition of STAT3 and STAT5A/B simply by intercrossing Stat3 and Stat5a/b single flox mice. To overcome this issue, we have generated a new GEMM which contains an allele harboring the Stat3, Stat5a and Stat5b genes flanked by loxP sites (Stat5/3 loxP/loxP mice). Stat5/3 loxP/loxP mice crossed to a transgenic line expressing the Cre recombinase in hepatocytes showed deletion of STAT3 and STAT5A/B in the liver and developed steatosis at the age of eight weeks. Furthermore, embryonic deletion of STAT3 and STAT5A/B resulted in a lethal phenotype. This demonstrates that the Stat5/3 loxP/loxP mouse is functional and can be readily used to model STAT3 and STAT5A/B double inhibition in cancer and other disease-experimental mouse models.

Generation of Stat3 and Stat5a/b Flox (Stat5/3 loxP/loxP ) Mice
A Bacterial Artificial Chromosome (BAC)-based targeting construct [21,22] spanning the Stat3, Stat5a and Stat5b locus was generated as follows: a 200 kb BAC (BAC, RPCI-23-362J7) ( Figure 1A and Supplementary Figure S1) was modified by introducing a loxP-FRT3-Neomycin-FRT3 cassette using BAC homologous recombination [23] upstream of the Stat5b gene as previously described [14]. In a second recombination step, a FRT-Hygromycin-FRT-loxP cassette was introduced within the first intron of the Stat3 gene between exons 1 and 2. Stat3 exon 2 contains the ATG translation initiation codon; thus, the flox strategy predicts the absence of a STAT3 truncated protein (Supplementary Figure S1). The BAC-based targeting construct was linearized using the restriction enzyme Not I and electroporated into HM-1 mouse embryonic stem cells (ES, 129/ola origin [24]). Out of around thirty ES cell clones which survived the neomycin/hygromycin double selection, eighteen appeared to have an undifferentiated morphology and were chosen for further analysis. We performed Southern blot analysis of genomic DNA derived from the selected clones using a DNA probe upstream of the hygromycin cassette to visualize the endogenous (wt) and the transgenic alleles ( Figure 1B). We presumed that equal intensities of wt and transgene bands either indicated correctly targeted ES cells which had incorporated the transgene in the endogenous Stat5/3 locus or random integration of two copies of the transgene. Unequal intensities between the wt and transgene bands would suggest random integration of the transgene into the genome. Clones 9, 10, 11, 13 and 15 ( Figure 1B) showed wt and transgene bands of approximately equal intensities and were further evaluated. We performed Fluorescence In Situ Hybridization (FISH) using the RPCI-23-362J7 BAC as probe for the selected clones ( Figure 2A and Table 1). wt ES cells showed mainly two signals, but we also observed nuclei containing three signals, suggesting chromosomal abnormalities. Notably, mouse ES cells tend to harbor a trisomy of chromosome 11 [25], which harbors the Stat5/3 locus. Clone 8 (which was included in the analysis as a control because the transgene band showed a higher intensity than the wt in the Southern blot, indicating a random integration event) displayed three signals. Clones 10, 11, 13 and 15 also showed mainly three or four signals ( Table 1), suggesting that the transgene had been randomly incorporated into the genome of these clones. In contrast, clone 9 exhibited two signals, indicating correct recombination of the transgene into the endogenous locus. Furthermore, PCR analysis of genomic DNA showed that the 5 and 3 ends of the BAC construct were not present in clone 9, demonstrating that they were lost during homologous recombination ( Figure 2B). Additionally, genomic qPCR using three pairs of primers flanking either the neomycin cassette (P3 and P4), a central region in the Stat5a locus (P5 and P6) or the hygromycin cassette (P7 and P8) confirmed that clone 9 contained one copy of the wt endogenous Stat5/3 locus and one copy of the transgene ( Figure 2C). Of note, primers P3/P4 and P7/8 do not amplify the transgenic construct. Finally, the functionality of the loxP sites was assessed by transfecting a plasmid expressing the Cre recombinase [26] into clone 9. Indeed, deletion of the Stat5/3 locus was achieved as demonstrated by PCR using genomic DNA and primers P9 and P10 ( Figure 2D). Therefore, we used ES cell clone 9 for the generation of chimeric animals via injection of the ES cell clone into blastocysts. Subsequently, these mice were mated to C57BL/6N mice to establish a mouse line harboring the Stat5/3 loxp allele.  Table 1. Quantification of the FISH analysis performed in the ES cell clones. FISH was performed using the RPCI-23-362J7 BAC as a probe in wildtype (wt) ES cells and clones 8, 9, 10, 11, 13 and 15. The result is expressed as the percentage of interphases (two hundred analyzed) containing one, two, three or four signals.

ES Clone
One Signal Two Signals Three Signals Four Signals  Table 2) and do not exhibit any obvious phenotype. Stat5/3 loxP/loxP mice (backcrossed five generations into C57BL/6N) were crossed to a transgenic line expressing the Cre recombinase under control of the Actb promoter [27] to obtain a null allele (Stat5/3 ∆/+ mice). Intercrossing of Stat5/3 ∆/+ mice did result in viable Stat5/3 ∆/∆ blastocysts, but not in living offspring ( Table 2), suggesting that Stat5/3 ∆/∆ mice were embryonic lethal, as previously described for Stat3 ∆/∆ mice [28]. Table 2. Genotype distribution of Stat5/3 wt, heterozygote and homozygote animals. Number of embryos and pups from Stat5/3 ∆/+ × Stat5/3 ∆/+ breeding distributed accordingly to the genotype (expected number in brackets). Statistical differences from the expected numbers were analysed with the chi-square test using GraphPad QuickCalcs. Previous reports have shown that the deletion of Stat5a/b in hepatocytes results in liver steatosis [29]. Therefore, we investigated the effect of concomitant Stat5a/b and Stat3 deletion in the liver. Stat5/3 loxP/loxP mice were intercrossed with transgenic mice expressing the Cre recombinase under the control of the tetO promoter (tetO-Cre mice [30]) and with mice expressing the tetracycline-controlled transactivator (tTA) in hepatocytes from the liver-enriched activator protein promoter (LAP-tTA mice [31]) to obtain triple transgenic animals: Stat5/3 loxP/loxP ; LAP-tTA; tetO-Cre (hereby Stat5/3 ∆hep/∆hep mice, ∆hep: deleted in hepatocytes). Liver-specific deletion of the Stat5/3 locus was confirmed by PCR using genomic DNA from different organs of Stat5/3 ∆hep/∆hep mice ( Figure 3B). Furthermore, Western blot analysis revealed that STAT5A/B and STAT3 proteins were absent in livers of Stat5/3 ∆hep/∆hep mice ( Figure 3C). At the macroscopic level, Stat5/3 ∆hep/∆hep mouse livers were enlarged and pale, suggesting lipid accumulation ( Figure 3D). HE-stained sections from livers of Stat5/3 ∆hep/∆hep mice confirmed a steatotic phenotype. Taken together, these results prove that Stat5/3 loxP/loxP mice are functional and recapitulate the liver steatotic phenotype found in mice lacking Stat5a/b in hepatocytes [29]. Modeling STAT3 and STAT5A/B double inhibition using GEMM in cancer and other diseases is an attractive strategy to assess inhibitor efficacy and side effects. Stat5a/b and Stat3 are genetically linked and located within an approximately 160 kb region of the mouse genome. Consequently, combined deletion of Stat5a/b and Stat3 by breeding currently-available mice harboring conditional alleles for Stat5a/b and Stat3 would be a challenging task. On the other hand, Singireddy and colleagues have reported the generation of double Stat5a/b and Stat3 flox mice by breeding individual flox mice; they found that double Stat5a/b and Stat3 flox mice were born at expected Mendelian ratios [32]. This may indicate that the Stat5a/b and Stat3 genomic region harbors a meiotic recombination hotspot whose efficiency could be genetic background-dependent [33]. We have used a different approach and generated a new transgenic mouse line carrying the Stat5a, Stat5b and Stat3 genes flanked by loxP sites. We have used a BAC-based targeting construct to modify ES cells [21,22]. This strategy has the advantage to flox relatively large genomic regions with one single targeting event in ES cells. Consequently, in contrast to the Southern blot strategy used for the detection of conventional targeting constructs [34], correctly targeted ES cells need to be identified by FISH and/or quantification of the endogenous and the transgenic alleles by qPCR [22]. Since Stat5/3 loxP/loxP mice were generated using ES derived from the 129/ola strain, we backcrossed these mice for five generations into C57BL/6N, a widely-used experimental mouse genetic background. Nevertheless, it is expected that the genomic region around the Stat5a/b and Stat3 genes is flanked by 129/ola ES cell derived DNA. This 129/ola genomic region may contain mutations that confound the interpretation of the phenotypes found upon the loss of Stat5a/b and Stat3 genes [35]. This issue could be overcome by generating the mice directly in C57BL/6N-derived ES cells [36]. It is known that increased distances between the loxP sites have been correlated with lower recombination efficiency [37]. Therefore, we aimed to demonstrate the functionality of Stat5/3 loxP/loxP mice by deleting the flox region in hepatocytes. Indeed, despite the relatively large genomic region (135 kb) flanked by loxP sites, we were able to achieve an efficient deletion of the Stat5a/b and Stat3 genes using a mouse line expressing the Cre recombinase in the liver under the control of the tetracycline-transactivator [30,31]. However, other Cre transgenic lines may not be that efficient. In this sense, crossing Stat5/3 loxP/loxP mice with Alfp Cre mice (a transgenic line which expresses the Cre recombinase in hepatocytes and cholangiocytes starting at embryonic day 9.5) [38] resulted in incomplete recombination.

Genotype
Although loss-of-function approaches using knockout mice may serve to validate targets on oncology, it has also some disadvantages. Due to the redundancy in biology, upon gene deletion, it is not usual that other proteins with similar functions become activated, resulting in compensatory mechanisms which may confound the interpretation phenotype [39]. In this sense, it has been reported that deletion of Stat5a/b (but not Stat3) in hepatocytes results in liver steatosis [29], possibly due to an unphysiological activation of STAT3 as cellular response to the lack of STAT5/B. Interestingly, in our model, the deletion of Stat5a/b and Stat3 in hepatocytes results in liver steatosis. This finding not only validates the functionality of our mouse model, but also suggests that the steatotic phenotype observed upon deletion of Stat5a/b in hepatocytes is not caused by aberrant-activation/regulation of Stat3.

BAC Recombineering
E. coli DH10B bacteria harboring the RPCI-23-362J7 BAC (purchased from CHORI) were electroporated with the pSC101-BAD-abgA plasmid [40] and grown at 30 • C until the OD reached 0.2. L-Arabinose was then added to the culture (0.2% final concentration) and cells were shifted to 37 • C until the OD reached 0.5. Cells were harvested and electrocompetent cells were prepared. A neomycin/kanamycin selection cassette flanked by FRT3 sites and harboring a 5 loxP site was recombined in the BAC upstream of Stat5b by electroporating a DpnI-digested PCR product amplified from a PGK/Tn5-neomycin cassette [41] using the following primers: CCAGAGAACCTGTGGGCAA TGGAATGGGCAGAAGCCTCAACCTACACTGCATAACTTCGTATAATGTATGCTATACGAAGTTAT GAAGTTCCTATACTATTTGAAGAATAGGAACTTCGTGGAGTCGAGGAATTCTACC and TGAGA GGAAAGCATGAAAGGGTTGGAGCCAGGGCATTATGTGTGAGGCAGGAAGTTCCTATTCTTCAAA TAGTATAGGAACTTCCAAAAACCAACACACAGATCATG (Italics: homology region to the BAC, bold: loxP site, underlined: FRT3 site). Putative recombinant clones were selected using kanamycin (25 µg/mL). Correctly recombined BACs were validated by PCR and direct BAC sequencing. In a second round of recombineering, a hygromycin selection cassette flanked by FRT sites harboring a 3 loxP site was recombined into the BAC containing the neomycin/kanamycin resistance cassette by electroporating a Dpn I-digested PCR product amplified from a pUBC/EM7-hygromicin plasmid [42] using the following primers: GTCTGTAACTCAAGAGGCTTGTTTCAGTGTTGATGTT TGGTGTATTTCGCGAAGTTCCTATTCTCTAGAAAGTATAGGAACTTCGGCCTCCGCGCCGGGTTT TGGC and GTTCTTGATGGTCCAGTTAATTGGAAAAAGGCACAAGCTGGTCTCTGGTTATAACTTCG TATAGCATACATTATACGAAGTTATGAAGTTCCTATACTTTCTAGAGAATAGGAACTTCCCTGC AACATGAATATTAGAA (Italics: homology region to the BAC, bold: loxP site, underlined: FRT site). Recombined BACs were selected in kanamycin (25 µg/mL) and hygromycin (50 µg/mL).

ES Cell Targeting
First, 30 µg of the BAC-based targeting construct were digested with NotI, phenol-chloroform extracted, ethanol precipitated, resuspended in 30 µl of PBS and electroporated into HM-1 mouse embryonic stem cells [24] with a BioRad Gene Pulser (230 V, 500 µF). Two days after electroporation, ES cells were selected in neomycin (350 µg/mL) and hygromycin (160 µg/mL) until colonies appeared (7-10 days). ES cell colonies with normal morphologies were picked in 96-well plates, expanded and transgene copy number was estimated by Southern blot using a probe located 3 downstream of the hygromycin cassette, which detects both the wt and transgene alleles

FISH Analysis of ES Clones
FISH was essentially performed as previously described [43]. In brief, BAC RPCI-23-362J7 DNA was labeled with digoxigenin-11-dUTP (Roche, Basel, Switzerland) by nick translation and for the detection of the probe sheep anti-digoxigenin-fluorescein isothiocyanate (FITC; Roche, Basel, Switzerland) and donkey anti-sheep-FITC (Dianova #713-095-147; Jackson Immuno Research, Cambridgeshire, UK) antibodies were used. Two hundred interphase nuclei per ES cell clone were scored for the number of FISH signals.
Blastocyst isolation was performed by superovulation and mating of heterozygous Stat5/3 ∆/+ females and males. Blastocysts were flushed from oviducts and uterus horns on E3.5 and cultured in M16 in an incubator for 24 h. DNA was extracted in 10 µl 1× homogenization buffer (containing 500 µg/mL Proteinase K), and genotyping was performed with the appropriated primers.

Conclusions
In conclusion, we have generated and validated a new mouse strain harboring a conditional allele for the Stat5a, Stat5b and Stat3 loci. This new mouse strain will serve as a valuable tool to address the combined functions of STAT3 and STAT5A/B in tumorigenesis and other (patho)physiological processes, expanding the research tools of the JAK/STAT scientific community.