Introduction

Type 1 Diabetes (T1D) is an autoimmune disease in which immune cells specifically target and kill pancreatic beta cells. A disease that typically manifests in young children, T1D necessitates a lifelong dependence on insulin, and is accompanied by significantly increased health risks and complications, even under the best managed care. Importantly, T1D affects an estimated 11–22 million people globally and its prevalence is increasing [1], [2]. This underscores the urgent need to find and develop treatments that can better regulate glycemia, restore beta cell function and improve patient outcomes.

Imatinib is a tyrosine kinase inhibitor (TKI) that was initially developed as an inhibitor of the Bcr-Abl oncogene for the treatment of chronic myeloid leukemia (CML) [3]. In addition, imatinib potently inhibits c-Kit and platelet derived growth factor receptors (PDGFRs), underlying its clinical use in the treatment of c-Kit-positive gastrointestinal stromal tumors (GIST) and PDGFR-associated myeloproliferative diseases. Interestingly, recent clinical reports have described improved glycemic control in patients with type 1 or type 2 diabetes taking imatinib for CML or chronic myeloproliferative disease [4]. Furthermore, preclinical studies have shown that imatinib treatment has efficacy in the non-obese diabetes (NOD) mouse model of T1D [5]. While several mechanistic studies suggest that inhibition of c-Abl and PDGFR are important for efficacy of imatinib in NOD diabetes [4], [5], the contribution of c-Kit inhibition to this activity has not been clearly addressed.

In this study, we engineer the T670I mutation into the mouse c-kit locus, an allele originally identified in GIST patients with refractory responses to imatinib [6], and bred the strain onto the NOD background to generate imatinib-resistant c-Kit mice that develop T1D. We characterize the imatinib response of diabetic NOD mice expressing wild-type or T670I mutant c-kit alleles and ask whether inhibition of c-Kit is required for efficacy of imatinib in this model. These results allow further definition of the target profile for tyrosine kinase inhibitor drug discovery programs focused on T1D treatment.

Methods

Results

The T670I allele of c-Kit is a mutation originally identified in GIST patients with refractory responses to imatinib [6]. Like other threonine “gatekeeper” mutants identified in TKI resistance [6], [8]–[11], the exchange of threonine for a bulkier isoleucine sterically blocks imatinib from entering the ATP binding pocket, reduces inhibitor binding stability and affinity, and confers drug resistance [8]–[11]. To develop imatinib-resistant c-Kit NOD mice, we generated mice that carried the c-kit T670I mutation (Figure 1a–c), and then bred this allele onto the NOD background for 10 generations and until known genomic insulin-dependent diabetes (Idd) loci were fixed. Henceforth, mice homozygous for the T670I allele will be referred to as NOD.c-Kit^T670I mice, while control littermates homozygous for the wild-type allele will be referred to as NOD.c-Kit^wt mice.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. NOD.c-Kit^T670I mice are imatinib resistant and develop diabetes.

A. Targeting strategy for generation of NOD.c-Kit^T670I mice. PCR genotyping (B) and sequence traces (C) of wild-type (wt) and mutant (mut) mice using primers F and R to generate fragments of 795 bp and 967 bp for wt and mut alleles, respectively. The T670I codon change and accompanying introduction of Bgl-II restriction site are highlighted. D. Expansion of c-Kit⁺/Sca-1⁺ murine HSCs from either NOD.c-Kit^T670I (black bar) or NOD.c-Kit^wt (white bar) littermates in the presence or absence of 5 µM imatinib. (*** p = 0.0002, as determined by two-way ANOVA). E. NOD.c-Kit^T670I (black square) mice develop diabetes comparably to NOD.c-Kit^wt (white circle) littermates (n = 41–45 mice/group; average age of diabetes onset = 14.5 weeks in NOD.c-Kit^wt mice and 15.5 weeks in NOD.c-Kit^T670I mice).

https://doi.org/10.1371/journal.pone.0084900.g001

To determine whether c-Kit activity in NOD.c-Kit^T670I mice was resistant to imatinib, we characterized the effect of imatinib on expansion of hematopoietic stem cells (HSCs), a process dependent on c-Kit [12]. We isolated HSCs from either NOD.c-Kit^T670I or NOD.c-Kit^wt littermate mice and expanded them in vitro in the presence or absence of 5 µM imatinib. After six days, we observed that expansion of HSCs from NOD.c-Kit^wt mice was substantially inhibited with addition of imatinib, as expected. In contrast, expansion of HSCs from NOD.c-Kit^T670I mice was not inhibited by imatinib, confirming that the c-kit allele of NOD.c-Kit^T670I mice is indeed resistant to imatinib (Figure 1d).

Subsequently, NOD.c-Kit^T670I and NOD.c-Kit^wt mice mice were monitored for diabetes onset. We found that both groups of mice developed diabetes with equivalent incidence and age of onset (Figure 1e). Recently diagnosed diabetic NOD.c-Kit^T670I and NOD.c-Kit^wt mice were dosed daily with imatinib over a period of three weeks, and blood glucose values (BGVs) were monitored at least every three days. We found that imatinib had comparable efficacy in reducing BGVs toward normal levels in both groups of mice (Figure 2 and Figure S1). Therefore, these data demonstrate that inhibition of c-Kit is not required for efficacy of imatinib in T1D.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Diabetic NOD.c-Kit^T670I mice are sensitive to imatinib treatment.

Blood glucose values for individual diabetic NOD.c-Kit^wt mice treated with either PBS (A) or imatinib (B), in comparison to diabetic NOD.c-Kit^T670I mice treated with either PBS (C) or imatinib (D) for 21 consecutive days. E. Average blood glucose values in diabetic NOD.c-Kit^wt or NOD.c-Kit^T670I mice treated with either PBS ((thick, solid line) NOD.c-Kit^wt; (thin, solid line) NOD.c-Kit^T670I) or imatinib ((thick, dashed line) NOD.c-Kit^wt; (thin, dashed line) NOD.c-Kit^T670I) for 21 consecutive days (n = 5–10 mice/group; *** p<0.0001 between PBS vs. imatinib within each mouse group for entire data set, as determined by one-way ANOVA). Note there was no statistically significant difference between NOD.c-Kit^wt and NOD.c-Kit^T670I within imatinib treatment groups. F. Area under curve for diabetes progression over 21 days comparing PBS ((open, white square) NOD.c-Kitwt; (closed, black square) NOD.c-Kit^T670I) or imatinib ((grey, slashed square) NOD.c-Kit^wt; (grey, closed square) NOD.c-Kit^T670I) treated mice. (* p≤0.01 between PBS vs. imatinib within each mouse group, as determined by one-way ANOVA).

https://doi.org/10.1371/journal.pone.0084900.g002

Discussion

Using a novel approach, we generated a genetically engineered mouse model of type 1diabetes that carries an imatinib-resistant c-Kit^T670I mutation and develops diabetes to define a target profile. Importantly, both groups of mice comparably develop diabetes and respond to imatinib treatment, ruling out any major contribution of c-kit inhibition in the mechanism of imatinib in this setting. Earlier studies by Louvet et. al. showed that neither an inhibitory anti-c-Kit antibody nor PLX647, a c-Kit/c-Fms inhibitor, had efficacy in diabetic NOD mice [5]. However, these studies were either complicated by anaphylaxis and death, or yielded weak, non-durable responses to inhibitor treatment leading the authors to conclude that c-Kit inhibition was not sufficient for efficacy. Our data provide genetic evidence consistent with Louvet et al, and further show that c-Kit inhibition is not required for efficacy by imatinib. Together, these studies indicate that the primary mechanism of action for imatinib efficacy in T1D is not through inhibition of c-Kit but probably through inhibition of other tyrosine kinase receptor targets such as c-Abl or PDGFR. Indeed, previous studies have implicated c-Abl and/or PDGFR inhibition by imatinib as likely targets in diabetes hyperglycemia reversal [5], [13].

Preclinical efficacy in NOD T1D has had poor positive predictive power in translation to clinical trials of human T1D. However, anecdotal reports from CML patients suggest that tyrosine kinase inhibitors may provide long-term benefit in reversing hyperglycemia in diabetic patients. As a targeted agent, imatinib is better tolerated than non-specific chemotherapeutics, and has been used in a pediatric population [3]. These findings have spurred efforts to initiate a controlled clinical trial evaluating the efficacy of imatinib in T1D patients [14]. However, the mechanism of action for imatinib and other tyrosine kinase inhibitors in treating mouse models of diabetes remains unclear. Studies that address this are crucial to future drug development in which improved, targeted therapies can be generated with better safety profiles. Our data support the development of inhibitors that do not target c-Kit, as inhibition of c-Kit is neither sufficient [5], nor is it required for efficacy of imatinib in NOD T1D. Importantly, loss of function alleles of c-kit have been adversely tied to impaired spermatogenesis, mast cell depletion, melanocyte dysfunction, and mild cytopenia [13]. Concerning treatments that may affect a pediatric population, tyrosine kinase therapies have also demonstrated an increased risk associated with impaired bone metabolism and growth retardation in prepubescent children [3], [15]. Notably, effects on bone have been attributed to inhibition of c-Kit signaling in osteoclasts [15]. In conclusion, this study provides important data in refining tyrosine kinase inhibitors to retain desired targets for the treatment of T1D while excluding irrelevant targets, minimizing potentially adverse events and improving the therapeutic index for new drug candidates.

Supporting Information

Acknowledgments

The authors would like to acknowledge C. Cienfuegos, E. Hesek, M. Corpuz, M. Lindstrom and the GNF Vivarium staff for technical assistance, and B. Wen and C. Filippi for thoughtful discussion.