Tetracycline Treatment Retards the Onset and Slows the Progression of Diabetes in Human Amylin/Islet Amyloid Polypeptide Transgenic Mice

Experimental protocols were approved by the University of Auckland Animal Ethics Committee and performed in accordance with the New Zealand Animal Welfare Act (1999).

Chemicals and kits were from Roche Applied Biosciences, Invitrogen, Life Technologies/BRL, or Sigma and were of analytical grade or better, unless stated otherwise.

The hA/hIAPP transgene (supplemental Fig. S1A [available at http://diabetes.diabetesjournals.org/cgi/content/full/db09-0548/DC1]) was constructed from PCR-derived fragments using the following primer pairs: RIP5 (GAAAGACTCGAGGATCCCCCAACCAC) and RIP3 (CAGGGCCATGGTGGAACAATGACC), hAMY5 (GAAGCCATGGGCATCCTGAAGCTGCAAGTA) and hAMY3 (GTCATGTGCACCT AAAGGGGCAAGTAATTCA), hALB5 (GTGTGTTTCATCGATATGCACGTAAGAAA) and hALB3 (CAACCTCAAGCTTGTCTGGGCAAGGG), hGAPDH5 (GGAGTAAGCTTCCTGGACCACCAGCCCC), and hGAPDH3 (GTGCGA GTCT-AGACTTCCTCCACCTGTCA). It was introduced into the genome by pronuclear injection (24).

RNA was extracted from liquid nitrogen snap-frozen tissues using QIAGEN RNeasy Midi Kits according with the rotor-stator tissue disruption (Ultra-Turrax T8; IKA-Werke, Germany). Total RNA (20 μg) was denatured and transferred to nylon membranes with 20× SSC (25) and hybridized to ³²P-labeled probes (26) at 65°C overnight. Intensities of mRNA bands corresponding to hA/hIAPP and mA/mIAPP were not directly compared as experimental conditions differed (supplemental Fig. S1A and B).

Mice were fed ad libitum with Diet 86 (Tegel NRM, Auckland, Zealand), a dry-pelleted natural-ingredient diet with 3% fat. Blood glucose levels were determined in tail vein blood (Advantage II; Roche) and diabetes defined when concentrations were >11 mmol/l on two consecutive weekly readings. For survival studies, we applied an agreed euthanasia surrogate end point for death, consistent with current ethical practice. Euthanasia was performed in the following circumstances: after 20% loss of maximum body weight, if significant lethargy developed, following loss of exploratory behavior with increasing relative immobility, failure to groom, or any other signs of overt distress. Infection or malignancy were also accepted as indications for euthanasia.

Mice undergoing intraperitoneal insulin tolerance tests (ITTs) were typically fasted for 6 h. Thereafter, actrapid (0.5 mIU/g body wt; Novo Nordisk) was injected intraperitoneally in conscious animals (29 G needle). Glucose levels were measured before injection and at indicated times postinjection. For intraperitoneal glucose tolerance tests (GTTs), mice were overnight fasted (16–18 h) with free access to water. Glucose was administered (1 mg glucose/g body wt) followed by tail blood sampling.

Cardiac puncture blood (EDTA) was separated (3,000g, 4°C, 15 min). Pancreata were excised and snap frozen (liquid nitrogen) and peptides extracted (homogenization, acid/ethanol) (27).

Murine insulin was determined in plasma, serum, or pancreatic extracts by Ultrasensitive Mouse Insulin enzyme-linked immunosorbent assay (ELISA) (Mercodia, Uppsala, Sweden) or rat/mouse insulin ELISA (Linco). Pancreatic extracts were diluted (1:500 and 1:5,000) in PBS (pH 7.4). Plasma hA/hIAPP was determined by ELISA (Linco) in plasma and pancreatic extracts; mA/mIAPP was determined using our in-house radioimmunoassay.

Pancreatic tissue was paraffin embedded, serially sectioned (5 μm), and stained with hematoxylin and eosin. On adjacent sections, Congo red (1%; 15–20 min, then saturated lithium carbonate, 30 s) and hematoxylin staining were used to measure frequency and extent of islet amyloid (polarization microscopy). Islet morphology and amyloid content were analyzed by a single-blinded histologist who scored nine or more islets/animal. Pancreata were sectioned at sufficient levels (≥200 μm apart) to ensure analysis of ≥9 (but generally ∼20–30) distinct islets/animal.

Representative sections were serially incubated with guinea pig anti-insulin serum, donkey anti-guinea pig IgG–fluorescein isothiocyanate, rabbit anti-glucagon, and donkey anti-rabbit IgG–Texas Red then counterstained with Congo red and reimaged to quantitate islet amyloid, morphology, and insulin and glucagon cells. Alternatively, sections were incubated with combined rabbit anti-glucagon and anti-somatostatin and imaged by indirect avidin-biotin-peroxidase.

Tetracycline was administered orally via the drinking water in light-proof bottles either from the time of weaning (21 days of age) or from diabetes onset in different studies, at final concentrations (0.03 mg/ml or 0.5 mg/ml in water, 18 mol · lΩ⁻¹ · cm⁻¹, milliQ; Millipore) and fresh solutions constituted weekly. Polydipsia was defined as fluid intake exceeding twice the average daily intake of the average nondiabetic adult mouse. Controls comprising hemizygous male mice and nontransgenic littermates received milliQ water alone.

Data were analyzed using GraphPad Prism 4 (GraphPad Software, San Diego, CA) and diabetes onset and survival using the Mantel-Haenszel log-rank test and the Gehan-Breslow-Wilcoxon test. Correlation analyses were performed using Pearson correlation with two-tailed hypotheses. Descriptive variables were contrasted by one-way ANOVA with Tukey-Kramer post hoc tests. Blood glucose and fluid intake values were contrasted using mixed models fitted by restricted maximum likelihood (JMP 5.1; SAS Institute). P values of <0.05 were considered significant.

Transgenic animals were generated within a FVB/N background (supplmentary Fig. S1). One transgenic line, hereinafter designated as Line 13, had integrated transgene copy numbers of 36 ± 7 and 76 ± 2 for hemizygous and homozygous animals, respectively (data not shown).

Diabetes developed spontaneously and reproducibly in both homozygous and hemizygous Line 13 male and female mice, while nontransgenic littermates did not develop hyperglycemia (Fig. 1). All subsequent studies were performed in male animals (28,29). In homozygous mice, diabetes developed in 100% of animals (n = 6) by 41 days and all were dead by 91 days (Fig. 1,A). By comparison, median values for time to diabetes onset and lifespan within the hemizygous group were longer (175 and 272 days, respectively) compared with homozygous animals (35 and 63 days, respectively). In addition, a second, independent line (designated Line 20) generated with the same construct and methods also developed spontaneous diabetes (Fig. 1 A). Thus, the diabetic phenotype was not an artifact arising from insertional mutagenesis.

Intraperitoneal ITTs performed at 85 days showed that hemizygous mice developed spontaneous diabetes within a background of normal insulin sensitivity (Fig. 1,B). Time to diabetes onset was not correlated with the period between diabetes onset and death, showing that age did not influence the rate of progression of diabetes once it had begun (Fig. 1 C).

In homozygous mice, at an age where they were glucose intolerant (30 days; intraperitoneal GTTs not shown), plasma hA/hIAPP concentrations were 1.8-fold higher than in hemizygous animals and ∼3.7-fold above nontransgenic control mA/mIAPP concentrations (Table 1). Normal plasma insulin was maintained in homozygous mice at 30 days despite a 72% reduction in pancreatic insulin content. Thus, at a time point similar to the median time to diabetes (35 days), substantial pancreatic insulin depletion had occurred in homozygous transgenic mice before a reduction in circulating insulin concentrations was detectable. Interestingly, and by contrast, pancreatic insulin concentrations in hemizygous animals were relatively unchanged at a time point (183 days) similar to the respective median onset time to diabetes (175 days). These data indicated that an ∼2- to 2.5-fold increase in expression of hA/hIAPP above endogenous and nonamyloidogenic mA/mIAPP in hemizygous animals was sufficient to evoke diabetes.

To investigate the relationship between amyloid deposition and diabetic phenotype, islet amyloid area was quantitated microscopically. Of the diabetic hemizygous animal cohort examined, there was a positive correlation between amyloid content and lifespan in animals that had survived >220 days (44% frequency, P < 0.001) (Fig. 1,D and Fig. 2, top two panels). However, amyloid was rarely observed in terminally diabetic animals killed at lifespans <220 days (49% frequency). As expected, amyloid deposition was never observed in the islets of nontransgenic animals (Fig. 2, middle panel). Interestingly, some hemizygous animals remained nondiabetic (7% frequency) but showed the presence of islet amyloid (Fig. 1,D and Fig. 2, second to bottom panel) with a similar relationship between amyloid content and lifespan as observed for diabetic hemizygous animals (P < 0.05). Amyloid was also absent from all terminally diabetic homozygous Line 13 mice analyzed (Fig. 2, bottom panel).

Immunohistochemistry of pancreatic islet sections using antisera to stain both α- and δ-cells showed that almost all the islet cells in terminally diabetic homozygous and hemizygous mice were non–β-cells (Fig. 2). Glucagon staining also revealed a change in α-cell distribution from the physiological location at the islet periphery to a more dispersed distribution throughout diabetic islets, a substantive change in histomorphology (Fig. 2). Taken together, these findings show that end-stage diabetes in both homozygous and hemizygous animals was associated with islet β-cell loss and altered islet architecture. Consistent with these observations, plasma insulin and amylin concentrations in terminally killed homozygous and hemizygous animals were below the limits of detection (data not shown). We also performed studies for terminal deoxynucleotidyl transferase-mediated dUTP nick-end labeling and caspase-3 activation in islets from animals at different stages of diabetes, but rates were not significantly different between those from diabetic and control (nondiabetic) animals (results not shown). These findings are consistent with the observation that although apoptosis is widespread in biology, dying cells are seldom seen in situ because of their rapid clearance by phagocytosis (30). They do not, however, exclude the possibility that the sensitivity of the assays used may not have been sufficient to detect any differences in numbers of rare apoptotic cells.

We previously reported that tetracycline interacts with amylin fibrils based on evidence from thioflavin-T fluorescence and radioprecipitation assays (31). Here, we provide additional in vitro evidence for specific interactions between tetracycline and hA/hIAPP by circular dichroism spectroscopy (supplementary Fig. S2). In addition, to investigate whether tetracycline might modulate indexes of aggregation-evoked diabetes in vivo, hemizygous mice were randomly assigned to groups that were administered either water alone (control) or water containing 0.03 mg/ml tetracycline. Transgenic animals treated with tetracycline (0.03 mg/ml) developed diabetes with an apparently earlier median time to onset (85 days, n = 17) compared with those administered water only (121 days, n = 20), which is a difference that was not significant by the Mantel-Haenszel log-rank test. As proportional hazards might not apply in this case, however, the data were reanalyzed using the Gehan-Breslow-Wilcoxon test, which does not rely on this assumption. The latter analysis did reveal a significant difference in median time to diabetes onset (P < 0.05). Following diabetes onset, progressive hyperglycemia resulted in the development of polydipsia at blood glucose concentrations of ∼25 mmol/l (Fig. 3 A and B). Since the upper limit of blood glucose measurement here was 33 mmol/l, the maximum reported values are lower-limit estimates of actual glucose concentrations in advanced diabetes.

Most importantly, in tetracycline-treated hemizygous animals, there was a significant delay in disease progression following onset of hyperglycemia, as measured by retardation in the rates of both blood glucose elevation (Fig. 3,A, P < 0.01) and fluid intake (Fig. 3,B, P < 0.01) compared with water-treated transgenic animals. Tetracycline had no effects on either blood glucose concentrations or fluid intake in nontransgenic littermates over the same time period. We did not quantitate blood tetracycline levels; however, assuming a similar bioavailability to oral doxycycline or minocycline administered to mice (32), we estimate that plasma tetracycline concentrations during the polydipsic phase could reach low micromolar levels. Survival analysis (Fig. 3 C) showed that tetracycline-treated transgenic animals had a significantly increased median survival (34%, P < 0.05) for the period from diabetes onset to death (155 days, n = 17) compared with the water-treated control group (116 days, n = 19).

Examination of pancreata from terminally killed tetracycline-treated animals also showed a positive correlation between islet amyloid content and lifespan (Fig. 3 D, P < 0.001, n = 12, R² = 0.91). This finding was similar to that observed for water-treated terminally diabetic hemizygous mice wherein the corresponding regression slope was not significantly different (Fig. 1D). Interestingly, of all the pancreata examined in this study the highest amyloid content occurred in a tetracycline-treated mouse, a finding consistent with the correspondingly increased lifespan observed in this group.

To investigate whether the observed antidiabetes effects of tetracycline might be attributable to potential indirect insulin-sensitizing effects, parallel GTTs and ITTs were performed in independent cohorts of animals with or without equivalent tetracycline treatment (Fig. 4). Hemizygous mice were administered tetracycline in their drinking water from 30 days postweaning at 0.5 mg/ml to approximate the increased drug intake observed during polydipsia in the previous study. GTTs performed on a hemizygous cohort revealed no baseline differences (0 days treatment; Fig. 4,A) but significantly improved glucose tolerance in the tetracycline-treated mice compared with water-treated controls at 30 days (Fig. 4,B) and 60 days treatment (Fig. 4 C). Survival analysis indicated a trend toward delayed diabetes onset in the tetracycline-treated group (P = 0.06, data not shown).

In a parallel study, a second independent hemizygous cohort was treated with or without tetracycline immediately postweaning. ITTs performed after 60 and 90 days treatment showed that improved glucose tolerance was not due to any insulin-sensitizing effect of tetracycline (Fig. 4,D and E). Tetracycline treatment also significantly delayed the median time to diabetes onset in this cohort of mice as compared with the corresponding water-treated control group (Fig. 4 F, 98 vs. 66 days, P < 0.05). These studies showed that improved glucose tolerance and delayed onset of diabetes in tetracycline-treated hemizygous mice were not due to extrapancreatic insulin-sensitizing actions or other putative systemic effects of tetracycline on glucose homeostasis.

To replicate and extend our previous findings, in particular with regard to dosage-related effects, further groups of hemizygous animals that had been administered water from the time of weaning were randomly assigned at the time of disease onset to one of the following groups: water-only control (n = 12), water containing 0.03 mg/ml tetracycline (n = 12) (study 3a, Fig. 5), or water containing 0.5 mg/ml tetracycline (n = 10) (study 3b, Fig. 5). The extent of suppression of disease progression in the tetracycline-treated groups was dosage dependent with significant alleviation of hyperglycemia (P < 0.01; Fig. 5,A) and corresponding fluid intake (P < 0.05; Fig. 5,B) in the 0.5 mg/ml tetracycline–treated group compared with the water-treated controls. The 0.5 mg/ml tetracycline–treated group had a 254% increase in median survival from diabetes onset to death compared with the water-treated control group (208 vs. 82 days, respectively, P < 0.05; Fig. 5 C), and there was a significant trend to dosage dependency among the three groups (P < 0.05).

To examine more precisely the effects of tetracycline treatment on islet β-cell mass, pancreatic islets were examined in a further independent cohort of diabetic mice studied at 6 weeks after disease onset. Hemizygous animals were randomly assigned at disease onset to one of two groups: either water-treated control or treatment with 0.5 mg/ml tetracycline, as was described above for study 3b. Interestingly, visual inspection of islets from hemizygous mice with early- to mid-stage diabetes (Fig. 6,A) showed they were structurally well preserved, with insulin areas similar to those of matched nontransgenic littermates (Fig. 6,B). Also, quantitative analyses of islet areas revealed no significant between-group difference at this time-point (Fig. 6,C). By contrast, animals with advanced diabetes had comparatively elevated blood glucose concentrations and reduced insulin areas, consistent with islet β-cell loss (Fig. 6 D).

These qualitative observations were confirmed in an exhaustive, quantitative blinded comparison of insulin areas of 9–30 islets from individual tetracycline-treated (Fig. 6,E) and water-treated (Fig. 6,F) hemizygous mice. Although there was no significant difference in mean insulin area–to–islet area ratios between the two groups (75 ± 4%, n = 8 vs. 71 ± 7%, n = 12, respectively), there was a significant difference in their variance (F test, P = 0.029), pointing to a between-group difference. This finding was extended in further correlational analyses, which showed that there were significant inverse relationships between blood glucose concentrations and mean islet insulin area in each group (Fig. 6,G). Here, there was a significant difference in the regression slopes between tetracycline- and water-treated mice (Fig. 6,G). In particular, insulin area–to–islet area ratios in diabetic hemizygous mice, wherein blood glucose concentrations were <15 mmol/l, were comparable not only between tetracycline and water-treated mice but also with values in nontransgenic nondiabetic mice, for example as represented by mouse no. 352 (Fig. 6 G, ♦).

We quantified pancreatic and serum insulin concentrations by ELISA in further separate groups of hemizygous mice 6 weeks after the onset of diabetes. Interestingly, although some reduction in total pancreatic insulin content was evident in some animals, there was no apparent relationship between insulin content and blood glucose concentrations (Fig. 6,H). Similarly, there was no evident reduction in serum insulin concentrations, and indeed some hemizygous animals displayed comparatively high serum insulin concentrations during early diabetes (Fig. 6 I). These findings showed that insulin area was maintained within the islets during early- and mid-stage diabetes in hemizygous transgenic mice with no substantial reductions in pancreatic insulin content or serum insulin concentrations. They are consistent with islet β-cell loss as a correlate of later-stage, more severe diabetes (blood glucose >15 mmol/l).

Our findings demonstrate that the amyloid deposits found in the pancreatic islets of hA/hIAPP transgenic mice are not intrinsically cytotoxic, an observation consistent with the reported occurrence of islet amyloid in nondiabetic humans (2). Perhaps most notably, amyloid deposition was not observed in homozygous animals with severe, early-onset diabetes but was present in both diabetic and nondiabetic hemizygous animals, where it was positively rather than negatively correlated with lifespan.

In homozygous transgenic animals, the lack of visible amyloid deposition coupled with the higher intrinsic hA/hIAPP expression and significantly depleted pancreatic insulin concentrations at the median time to diabetes onset were all consistent with islet β-cell loss as the major diabetogenic mechanism. Evidence from other hA/hIAPP transgenic murine lines has pointed to a role for soluble oligomers in the increased frequency of β-cell apoptosis in late-stage diabetes (15,18,19). The corresponding lack of amyloid in our homozygous mice is consistent with this pathogenic mechanism and supports a growing body of experimental data consistent with the hypothesis that cytotoxic effects of prefibrillar aggregates of other amyloidogenic proteins, such as β-amyloid and α-synuclein, can elicit cell death (33,,,–37).

Our findings also show that tetracycline can partially suppress the progression of diabetes in hemizygous animals. In study 1, it slowed the rate of deterioration in blood glucose and polydipsia after diabetes onset, compared with matched control animals, which translated into a 34% increase in median survival. By contrast, tetracycline did not delay diabetes onset in this study, although the lack of an effect may simply have reflected the relatively low dosage administered during the pre-diabetic phase. The likelihood that this explanation is correct was confirmed in study 2, wherein tetracycline administration at the higher dosage of 0.5 mg/ml from 30 days after weaning, significantly delayed disease onset and progression. The observed improvement in glucose tolerance in tetracycline-treated hemizygous mice after 30 and 60 days of treatment is also consistent with observations from study 3, where we again used the higher drug dosage (0.5 mg/ml) but with administration from the time of disease onset. In this study, tetracycline exerted a clear, dosage-dependent effect to delay the deterioration in blood glucose regulation and polydipsia, with the higher dosage causing an increase in median survival of 254% in the period following diabetes onset.

We found no evidence that the antidiabetes effects evoked by tetracycline in the hemizygous mice were due to any putative extrapancreatic effects, since ITTs revealed that it exerted no detectable systemic insulin-sensitizing effects in either hemizygous mice or their nontransgenic littermates. Furthermore, tetracycline had no measurable effects on glucose regulation in nontransgenic littermates. When taken together with the histological analysis of islets 6 weeks after diabetes onset, these data point toward preservation of islet β-cell function as the mechanism underlying tetracycline's actions to delay onset and progression of diabetes, possibly through interactions with soluble nascent prefibrillar aggregates.

Finally, our findings show that β-cell dysfunction and not β-cell loss was responsible for the initial development of diabetes in these hemizygous mice. Compared with the rapid onset and development of diabetes in homozygous animals, diabetes onset and progression were significantly more gradual in the hemizygous group. Strikingly, nontransgenic controls and hemizygous animals had similar total pancreatic insulin content at the median time of diabetes onset. Significantly, their insulin areas were also comparable in the early- and mid-stages of diabetes, indicating that no significant loss of β-cells had ensued to this point in disease progression. Direct measurements of pancreatic and serum insulin concentrations 6 weeks following diabetes onset, although variable, were also consistent with the above finding, in that there were no substantive reductions in insulin content and no clear associations with blood glucose concentrations. This effect is consistent with the development of glucose blindness in the islets of at least some hemizygous mice (38).

Our experimental analyses were clearly able to discriminate a 20% loss in insulin staining area or in islet β-cell mass, but such absolute reductions only occurred in association with advanced diabetes. That no significant reduction in insulin staining area occurred in early- to mid-stage diabetes clearly demonstrates that hA/hIAPP expression does not initially evoke diabetes in hemizygous mice by eliciting overt islet β-cell loss. In support of this conclusion, extensive immunohistochemical analyses revealed no evidence for β-cell apoptosis in pancreatic sections from hemizygous mice 6 weeks following diabetes onset, as measured by caspase-3 or terminal deoxynucleotidyl transferase-mediated dUTP nick-end labeling staining (data not shown), although the assays we used may have been insufficiently sensitive to detect rare apoptotic cells in affected islets.

Our findings indicate that below a certain threshold, hA/hIAPP expression causes islet β-cell dysfunction leading to impaired insulin production, processing, and/or secretion at the outset of diabetes (14). Here, we could not investigate insulin processing due to the unavailability of an appropriate mouse-specific proinsulin ELISA. As amylin expression in our mice is under control of the rat insulin 2 promoter, we expect that rising blood glucose would establish a positive feedback cycle that progressively facilitates islet β-cell loss. Presumably, this latter mechanism occurs by default in homozygous mice which possess higher intrinsic hA/hIAPP expression.

In summary, our findings show that islet β-cell dysfunction and not mature extracellular amyloid is the underpinning cause for diabetes pathogenesis in hemizygous hA/hIAPP transgenic mice. Moreover, deposition of microscopically visible amyloid was positively correlated with lifespan, showing that tissue hA/hIAPP deposits are not intrinsically cytotoxic. Treatment with an effective dosage of tetracycline delayed the onset and impeded the progression of diabetes in hemizygous mice. Consequently, any intervention that allows progressive deposition of (apparently benign) islet amyloid through mechanisms that reduce the cytotoxicity of prefibrillar aggregates might be expected to prevent islet β-cell degeneration.

These studies were supported by the Endocore Research Trust, the University of Auckland Research Committee, the Maurice & Phyllis Paykel Trust, the Auckland Medical Research Foundation, and Lottery Health (New Zealand). G.C. acknowledges support by program grants from the Foundation for Research, Science, and Technology and by the Health Research Council of New Zealand.

No potential conflicts of interest relevant to this article were reported.

We thank John Todd, Cynthia Tse, and John Scott for helpful discussions and gratefully acknowledge Xiaoling Li, Vita Chien, Rosemary Smith, Nataliya Olerskeya, Beryl Davy, and Vernon Tintinger for technical assistance and Vivian Ward for excellent graphics support.