TLR2 knockout protects against diabetes-mediated changes in cerebral perfusion and cognitive deficits

Abstract

The risk of cognitive decline in diabetes (Type 1 and Type 2) is significantly greater compared with normoglycemic patients, and the risk of developing dementia in diabetic patients is doubled. The etiology for this is likely multifactorial, but one mechanism that has gained increasing attention is decreased cerebral perfusion as a result of cerebrovascular dysfunction. The innate immune system has been shown to play a role in diabetic vascular complications, notably through the Toll-like receptor (TLR)-stimulated release of proinflammatory cytokines and chemokines that lead to vascular damage. TLR2 has been implicated in playing a crucial role in the development of diabetic microvascular complications, such as nephropathy, and thus, we hypothesized that TLR2-mediated cerebrovascular dysfunction leads to decreased cerebral blood flow (CBF) and cognitive impairment in diabetes. Knockout of TLR2 conferred protection from impaired CBF in early-stage diabetes and from hyperperfusion in long-term diabetes, prevented the development of endothelium-dependent vascular dysfunction in diabetes, created a hyperactive and anxiolytic phenotype, and protected against diabetes-induced impairment of long-term hippocampal and prefrontal cortex-mediated fear learning. In conclusion, these findings support the involvement of TLR2 in the pathogenesis of diabetic vascular disease and cognitive impairment.

the risk of cognitive decline in diabetes (Type 1 and Type 2) is significantly greater compared with normoglycemic patients, and the risk of developing dementia in diabetic patients is doubled (26, 36, 47, 48). Type 1 diabetes (T1D) and Type 2 (T2D) have been shown to lead to significant reductions in overall cognition, intelligence, speed of information procession, psychomotor efficiency, and mental flexibility (5, 27). While the cognitive decline in diabetes is certainly multifactorial, decreased cerebral perfusion plays a large role in its pathogenesis. It is known that brain function heavily depends on constant perfusion and that changes in vascular function and cerebral perfusion precede cognitive deficits in diabetes. However, our limited understanding of the role and mechanisms by which cerebrovascular dysfunction leads to cognitive impairment in diabetes remains a fundamental gap in the field.

The innate immune system has been implicated in the pathological changes in diabetes, notably through Toll-like receptor (TLR) signaling. Activation of TLRs by pathogens or endogenous danger signals known as damage-associated molecular patterns (DAMPs) stimulates NF-κB and IFN-β, leading to subsequent release of proinflammatory cytokines and chemokines (23, 54). TLR2 is involved in the development of diabetic microvascular complications, such as nephropathy, as well as the endothelial dysfunction that presents in diabetes (20, 32). However, whether TLR2 activation in the cerebrovasculature is an early event linking vascular dysfunction and inflammation to cognitive impairment is unknown. The developing brain can be extremely susceptible to inflammation, and postnatal activation of TLR2 has been shown to induce neuroinflammation that impairs cognitive spatial and fear learning, leading to the impairment of motor skills in adulthood (34). Additionally, there are conflicting reports as to the effect of TLR2 knockout on exploratory behavior (34, 40), suggesting that TLR2 could play a critical role in the plasticity of cognitive behavior during developmentally crucial periods. In this study, we hypothesized that genetic deletion of TLR2 in a T1D model would protect from endothelial dysfunction and improve cerebral perfusion and indices of cognitive function, such as short-term and long-term hippocampus-dependent learning.

METHODS

Animals and induction of diabetes.

All protocols were approved by the Institutional Animal Care and Use Committee at Augusta University. All experiments were performed using male wild-type (WT) C57BL/6J (stock no. 000664; Jackson Laboratory, Bar Harbor, ME) or TLR2-knockout (KO) B6.129-Tlr2^tm1kir/J (stock no. 004650; Jackson Laboratory) mice. Male mice were used in this study to build on published literature exploring the effects of TLR2-KO mouse model on cognitive function (34, 40). Additionally, our use of streptozotocin (STZ)-induced diabetes in the male TLR2-KO mouse model allows for expansion on the work examining TLR2-KO-mediated renal protection in STZ-induced diabetes (12, 32). The current study serves a bridge between these different areas of research (diabetes/cognitive function) in regard to the literature on TLR2-knockout mice. The animals were housed at Augusta University animal care facility that is approved by the American Association for Accreditation of Laboratory Animal Care. Animals were fed standard rodent chow and tap water ad libitum throughout the duration of the study. Body weights and blood glucose measurements were taken biweekly. Blood glucose measurements were taken from tail vein samples using a commercially available glucometer (Freestyle, Abbott Diabetes Care, Alameda, CA). At 10 wk of age, mice were fasted overnight and injected with 50 mg/kg ip STZ or vehicle for 5 days consecutively. Glycosylated hemoglobin values (HbA1c%) (A1CNow-plus; PTS Diagnostics, Indianapolis, IN) were used as a measurement of long-term blood glucose levels. Diabetes in mice was defined as having a HbA1c% > 8.0% to ensure a significant level of hyperglycemia. All nondiabetic control mice had a HbA1c% < 6.0%. All behavioral experiments were performed in a blinded fashion at Augusta University’s Small Animal Behavioral Core facility. At the time of euthanasia, the heart, kidney, and spleen were harvested to determine effect of diabetes on organ weight in WT and TLR2KO mice.

High-resolution laser speckle contrast imaging of cortical perfusion.

High-resolution Laser Speckle Contrast Imaging (LCSI, Perimed) was used to record cerebral perfusion at 6 wk after induction of diabetes with STZ (mice aged 16 wk old) to determine the effect of early-stage diabetes. A region of interest (ROI) between the bregma and lambda in the two hemispheres was selected. Briefly, the mice were anesthetized using isoflurane, body temperature was maintained at 37 ± 0.5°C, and the skull was shaven and exposed by a midline skin incision. Perfusion images were acquired using PSI system with a 70-mW built-in laser diode for illumination and 1,388 × 1,038 pixels charge-coupled device camera installed 10 cm above the skull (speed: 19 Hz; exposure time: 6 ms). Acquired images were analyzed for changes in cerebral perfusion using a dedicated PIMSoft.

Magnetic resonance imaging FAIR-RARE of whole brain perfusion.

MRI was conducted at the Augusta University Core Imaging Facility for Small Animals at 14 wk postinduction of diabetes with STZ (mice aged 24 wk old) to measure relative cerebral perfusion using the FAIR technique. Anesthesia was maintained and induced for each mouse using a vaporized mixture of isoflurane and compressed oxygen throughout each MR session. Each mouse was secured and stabilized in the prone position to a dedicated holder and then positioned into a 7.0 T 20-cm bore BioSpec MRI spectrometer (Bruker Instruments, Billerica, MA) with standard transmit/receive volume coil (35 mm ID) for MRI acquisition. Throughout the entire scan period, cardiopulmonary status and respiration were monitored (SA Instruments Physiological Monitoring System, Stony Brook, NY). The brain was imaged in five sections. Perfusion-sensitive images were acquired using a FAIR RARE sequence with the following parameters: echo time = 46 ms, repetition time = 12,000 ms, flip angle = 180°, field of view = 3.2 cm × 3.2 cm, slice thickness = 1.5 mm, number of slices = 5 matrix = 128 × 128. Total acquisition time to collect multiple MR images for each subject was ~42 min for the entire session. Perfusion maps were calculated using a macro in ParaVision 5.1 and exported as 32-bit raw image files for post-processing analysis using ImageJ (1). ROI was the whole brain for each section. Values were converted to units of milliliter per 100 g times minutes (i.e., units of ml/100 g × min) using slope conversation factor derived from the Paravision perfusion map.

Vascular reactivity.

After MRI FAIR-RARE and behavioral testing (fear-conditioned learning) were completed, animals were euthanized and aortas were quickly excised from nondiabetic and diabetic WT and TLR2KO mice and used within 45 min of isolation to ensure viability of the vessels. Aortas were cut in to 2-mm segments. Isometric tension exerted by the vessels was recorded via a force transducer using the pin-myograph technique (model no. 610M; Danish Myo Technologies, Aarhus, Denmark). The myograph chambers were filled with Krebs buffer (in mM: 118 NaCl, 25 NaHCO₃, 4.7 KCl, 1.2 MgSO₄, 1.2 KH₂PO₄, 1.5 CaCl₂, and 11.1 dextrose), gassed with 95% O₂ and 5% CO₂, and maintained at 37 ± 0.5°C. Vessel segments were mounted in the chamber using 200-µm pins and adjusted to a baseline tension (5.5 mN). After stabilization, the vessels were challenged with 120 mM KCl, washed, and allowed to equilibrate for 30 min. Cumulative dose-response curves to the dipeptidyl-peptidase-IV (DPP-IV) inhibitor linagliptin (1 nM to 0.1 mM) were performed, and the force generated was expressed as a percent change from maximum phenylephrine-induced preconstriction. Although linagliptin is used in the treatment of Type 2 diabetes, past studies, including ours, suggest that linagliptin exhibits glucose-independent effects in a model of Type 2 diabetes (18, 19, 27, 31). In addition, it has been shown that linagliptin has a protective effect on the microvasculature of diabetic Wistar rats (13) and preserves endothelial function in STZ-induced diabetic rats by reducing vascular superoxide levels and preservation of both NO and EDH-mediated responses (50). Given that TLR2 contributes to endothelial dysfunction via a proinflammatory state that reduces NO (20), linagliptin was used as a vasodilator in the current study. Maximum relaxation response (R_max) was assessed, and to determine the contribution of NO to this relaxation response, inhibition of nitric oxide synthase (NOS) was achieved using preincubation with N^G-nitro-l-arginine methyl ester (l-NAME; 1 mM) for 30 min.

Locomotor activity assessment.

Locomotor activity assessment was recorded 6 wk after induction of diabetes with STZ (mice, 16 wk old). Animal movement was captured and recorded continuously both vertically and horizontally by photobeam sensors by way of an automated and digital monitoring system (Med Associates, St. Albans, VT). Each mouse was placed in an open field chamber (20 cm × 20 cm × 28 cm) to freely explore for 30 min. Photobeam breaks were subsequently analyzed and converted to directionally specific movements using Activity Monitor (Med Associates) and then exported in total and subdivided further to central (middle box) and peripheral (outside edge of box) zones.

Novel object recognition.

Novel object recognition (NOR) was recorded 6 wk after induction of diabetes with STZ (16-wk-old mice). Mice were placed in testing boxes (approximate dimensions = 40 cm × 25 cm × 20 cm) for three consecutive days (15, 10, and 10 min, respectively) to habituate before training and testing. On day 4, the mouse was placed in the box for 10 min of free exploration training session in the presence of two equidistant and identical objects (eye bolts). After a 1-h delay to test short-term memory, one of the objects was exchanged (six-pointed head screw knob), and the mouse was reintroduced for 10 min of free exploration session: the novel object recognition session. Animals were excluded that did not spend >20% total time exploring both objects or exhibited a bias for either object during training session (i.e., >55% of total session time spent with one of the objects) (All sessions were video recorded and analyzed using automated video-tracking software (Ethovision XT 11.0; Noldus, Leesburg, VA). The time a mouse spent with its nose in each object was used to indicate object preference, and distance traveled and mean velocity were used as indicators of locomotor activity. The object recognition index was calculated by dividing the total time spent with the novel object (in seconds) by the total time spent with both objects (in seconds).

Y-maze.

Y-maze testing was recorded 6 wk after induction of diabetes with STZ (16-wk-old mice). The Y-maze was opaque white-cast acrylic with three arms (each = 30 cm L × 8 cm W × 6 cm H) joined at one end to form a “Y” shape with a common central zone. With one arm blocked, each trial was initiated by releasing a mouse from the end of one arm for a 10-min free exploration period. After a 15-min intertrial interval, the mouse was reintroduced to the Y-maze for a second trial and 3 min of free exploration with all arms unblocked. All trials were video-recorded and analyzed using video-tracking software (Ethovision XT, Noldus, Leesburg, VA) to obtain the total number of entries per arm, time spent in each arm (seconds), the distance traveled in each arm (cm), and the mean velocity (cm/s). All measures chosen are indicators of the ability to hold information on the spatial relations from the first trial.

Fear-conditioned learning.

Fear-conditioned learning was recorded 14 wk after induction of diabetes with STZ (24-wk-old mice). Associative learning was assessed using a fear-conditioned learning method, an ideal tool to assess hippocampus-dependent (contextual) and -independent (cued) learning, respectively. Mice were trained (i.e., conditioned) in sound-attenuated chambers (Med Associates) for 7 min to associate a context (sensory and visual information gathered while in the chamber) and cue (30 s 80-dB white noise), as well as the conditioned stimulus (CS). The unconditioned stimulus (US) was a 2-s electric footshock (0.5 mA) applied to the grid floor. Training sessions began with a free exploration followed by delivery of two CS-US “pairings” at 3 and 5 min (i.e., “2-pairings training method”). Twenty-four hours later, contextual fear learning was evaluated for 5 min (no presentation of CS or US), and then cued fear learning was evaluated by moving each mouse to a different apparatus with numerous contextual changes to odor, lighting, exploratory space, and floor texture. The new contextual changes were validated with 3-min free exploration (no CS or US), followed by a 3-min period with the CS held constant. Fear extinction learning was evaluated by returning mice to same chamber for 5 min to evaluate contextual learning (no CS or US) 4 and 14 days following the training session. All sessions were video recorded, and freezing behavior was scored manually by a hand-tallying method (8). Data are expressed as the total percent freezing.

Statistics.

Data were analyzed using the following softwares: GraphPad Prism (La Jolla, CA) and NCSS Statistical Software (Kaysville, UT). Data and graphs are reported and plotted as means ± SE with P < 0.05 considered to be significant. Groups were analyzed by multifactorial [factors = genotype: WT, or TLR2KO; treatment: control (CTL) or STZ; group: WT-CTL, WT-STZ, TLR2KO-CTL, and TLR2KO-STZ] ANOVA followed by Tukey’s post hoc analysis with multiple comparisons when applicable. For paired analyses, two-way Student's t-test was performed and P ≤ 0.05 was considered significant.

RESULTS

STZ-induced hyperglycemia and changes in body and organ weight are not dependent on TLR2KO.

Injection of STZ led to hyperglycemia in both WT and TLR2KO mice, as indicated by both fasting blood glucose levels (Fig. 1A) and HbA1c (Fig. 1B). These indices provide a measure of acute and chronic glycemic control, respectively. Additionally, the baseline blood glucose levels and HbA1c in the control groups were also the same between WT and TLR2KO. These results indicate that there is no effect of genetic deletion of TLR2 on glycemic control. Diabetes induced by STZ led to a decrease in body weight of WT and TLR2KO mice compared with their nondiabetic controls, with no differences observed between body weights in the control groups themselves (Fig. 1C). The effect of diabetes was the same in each genotype on organ weights as well, with a decrease observed in heart weight (Fig. 1D) and spleen weight (Fig. 1E). There was a trend toward an increase in kidney weight in both the WT and TLR2KO diabetic mice compared with their nondiabetic controls; however, this effect did not reach significance in either group (Fig. 1F).

Fig. 1.

Glycemic data and body/organ weights of Toll-like receptor-2 (TLR2)-knockout (TLR2KO) nondiabetic and diabetic mice. A: streptozotocin (STZ) induces fasting hyperglycemia in both the wild-type (WT) and TLR2KO mice to the same degree. B: long-term glycemic levels during the experiment were also increased to the same degree in the diabetic WT and TLR2KO STZ groups, as indicated by hemoglobin A1c (HbA1c)%. C: body weights in the diabetic WT and TLR2KO mice at 24 wk of age were reduced compared with their respective controls, with no differences observed as a result of genotype. Both heart weight (D) and spleen weight (E) were reduced in the diabetic WT and TLR2KO mice compared with their respective controls, with no effect of genotype. F: kidney weight in the diabetic mice exhibited a trend toward being increased in both the WT and the TLR2KO groups similarly, but this effect was not significant (P = 0.088). For WT control (CTL), n = 7; for WT STZ, n = 17; for TLR2KO CTL, n = 13; for TLR2KO STZ, n = 14. *P < 0.05 vs WT CTL. #P < 0.05 vs. TLR2KO CTL.

TLR2KO protects against decreased cortical perfusion in early-stage diabetes and increased whole brain perfusion in long-term diabetes.

Cortical perfusion was assessed in control and diabetic WT and TLR2KO mice at 6 wk postinduction of diabetes using high-resolution laser speckle imaging (Fig. 2A). It was revealed that in early-stage diabetes, there is a reduction in cerebral perfusion in diabetic WT mice in the region between the lambda and the bregma, an area that is supplied by the middle cerebral arteries (MCAs) and is commonly affected in diabetic cerebrovascular disease (P < 0.05). This decreased cerebral perfusion was absent in diabetic TLR2KO mice, and neither control nor diabetic TLR2KO mice had a statistically significant difference in cerebral perfusion compared with control WT mice (Fig. 2B).

Fig. 2.

Cerebral perfusion changes in early- and late-stage diabetic mice. A: cortical perfusion in WT and TLR2KO nondiabetic and early-stage diabetic mice. B: Laser speckle imaging reveals that cortical perfusion in brain areas supplied by the middle cerebral artery is decreased in WT STZ, and TLR2KO protects against this change in diabetes. No differences were observed between nondiabetic TRL2KO control and WT control; n = 4–6, *P < 0.05 vs. WT CTL. C: whole brain perfusion in WT and TLR2KO nondiabetic and long-term diabetic mice was assessed with MRI FAIR-RARE perfusion analysis. D: WT STZ mice exhibit an increase in whole brain perfusion compared with WT CTL, and TLR2KO was able to prevent this diabetes-induced change in CBF. No difference was observed between nondiabetic control groups. n = 4/group, *P < 0.05 vs. WT CTL.

Laser speckle imaging does not reveal subcortical perfusion, and given diabetes's ability to affect the microvasculature, as well as the macrovasculature, we subsequently employed MRI perfusion analysis using FAIR-RARE imaging to determine whole brain perfusion in long-term diabetes (Fig. 2C). The whole brain perfusion was increased in diabetic WT mice compared with WT controls (P < 0.05), and this effect did not occur in the diabetic TLR2KO mice compared with TLR2KO controls (Fig. 2D).

TLR2KO prevents diabetes-induced impairment in vascular relaxation to linagliptin.

To determine whether endothelial dysfunction, potentially resulting from an inflammatory process mediated by TLR2, was occurring in our STZ model, endothelium-dependent vasorelaxation was assessed in aortic rings using the DPP-IV inhibitor linagliptin. Diabetic WT mice exhibited an impairment in maximum relaxation to linagliptin compared with control WT mice (P < 0.05), and diabetic TLR2KO mice were protected from this vascular dysfunction (Fig. 3A). To assess the contribution of NO to the vasorelaxation in each genotype, we used the inhibitor of NOS, l-NAME (1 mM). Inhibition of NOS in the WT mice significantly impaired maximum vasorelaxation to linagliptin. In the TLR2KO mice, inhibition of NOS dramatically reduced the level of vasorelaxation and led to a sustained vasoconstriction. The effects of l-NAME in the TLR2KO mice were significantly different from both the TLR2KO control vessels, as well as the WT vessels treated with l-NAME (P < 0.05), suggesting that the endothelium-dependent response to linagliptin is more dependent on tonal NO in the TLR2KO mice (Fig. 3B).

Fig. 3.

Vascular response to linagliptin in aorta WT and TLR2KO nondiabetic and diabetic mice. A: WT STZ mice exhibit a decreased maximum relaxation response to linagliptin, and TLR2KO protects against this cerebrovascular dysfunction in diabetes. B: inhibition of nitric oxide (NO) synthase (NOS) with N^G-nitro-l-arginine methyl ester (l-NAME) significantly prevents vasorelaxation to linagliptin, indicating the response is dependent on NO. Inhibition in TLR2KO mice leads to a significantly greater impairment in relaxation than in WT, suggesting that in TLR2KO mice vasorelaxation is more dependent on NO. For WT CTL, n = 5; for WT STZ, n = 11; for TLR2KO CTL, n = 6; for TLR2KO STZ, n = 8, for WT l-NAME, n = 4; for TLR2KO l-NAME, n = 6 *P < 0.05 vs. WT CTL. #P < 0.05 vs. TLR2KO CTL.

TLR2KO mice exhibit a hyperactive phenotype compared with WT mice, and diabetes does not affect exploratory behavior in either genotype.

There was a significant effect of TLR2KO deletion on locomotor activity, as evidenced by the increases in ambulatory counts (Fig. 4A) (P < 0.05), total distance traveled during the test (Fig. 4B) (P < 0.05), and the ratio of distance traveled in the center versus distance traveled on the periphery of the enclosure (Fig. 4C) (P < 0.05). There was no effect of diabetes on locomotor activity among either genotype; however, the diabetic TLR2KO mice did not exhibit the increase observed in WT controls. This genotype-specific effect on locomotion prevents an accurate, appropriate comparison of cognitive function when using cognitive tests that rely on locomotion, such as NOR or Y-maze. As there was no significant effect on diabetes compared with control in either genotype, it was more appropriate in subsequent cognitive testing to compare the effects of diabetes on WT and on TLR2KO mice individually instead of in a 2 × 2 (genotype × disease) analysis.

Fig. 4.

Exploratory behavior WT and TLR2KO nondiabetic and diabetic mice. Locomotor activity testing examining exploratory behavior as indicated by ambulatory counts (A), total distance traveled (B), and center-to-periphery distance traveled (C) over 30 min shows that there was no significant difference in WT control or diabetic mice, meaning these animals can be compared against each other in future cognitive tasks. TLR2KO mice, however, exhibit an increased exploratory behavior in all three indices, which suggests that it is not appropriate to compare WT and TLR2KO mice in cognitive tasks relying on locomotion. n = 9–11. *P < 0.05 vs. WT CTL.

Increased exploratory behavior in TLR2KO mice impacts cognitive assessment testing of short-term hippocampus-dependent learning.

The hyperactivity observed in locomotor activity testing is likely to influence cognitive tests that rely on locomotion, as the differences in behavior lead to confounding assessments of actual cognitive function. With this limitation in mind, we still attempted to assess short-term hippocampus-dependent learning using two different behavioral tests, the NOR test, and Y-maze. NOR assessment revealed a similar hyperactivity in the TLR2KO control mice compared with the WT control mice, as indicated by total distance traveled (Fig. 5A) (P < 0.05) and mean velocity over the course of the assessment (Fig. 5B) (P < 0.05). However, here, diabetes in the TLR2KO mice significantly decreased the hyperactivity compared with TLR2KO control (P < 0.05). There was no significant difference in the locomotion between either diabetic TLR2KO or diabetic WT mice and WT control mice. TLR2KO control mice were observed to have an apparent decrease in short-term hippocampus-dependent learning, as assessed by the Recognition Index compared with WT controls (Fig. 5C) (P < 0.05). However, the amount of time that these animals interacted with the novel object as a result of their hyperactive phenotype likely influenced this assessment.

Fig. 5.

Novel object recognition (NOR) in WT and TLR2KO nondiabetic and diabetic mice. Testing revealed that diabetes did not have an effect on distance traveled (A) or mean velocity (B) in WT mice. TLR2KO CTL mice showed a hyperactive phenotype similar to that observed in locomotor activity testing, and this increase in locomotion was reduced by diabetes. C: TLR2KO mice had a decrease in NOR index compared with WT CTL. Because NOR assessment depends on locomotion, this test was likely impacted by the hyperactivity in the TLR2KO mice and does not provide a good measure of short-term hippocampus-dependent learning. n = 9–10. *P < 0.05 vs. WT CTL. #P < 0.05 vs. TLR2KO CTL.

Y-maze testing also showed an increase in locomotor activity in the TLR2KO control group compared with the WT control mice, both in total distance traveled (Fig. 6A) and total mean velocity over the duration of the test (Fig. 6B) (P < 0.01). Diabetic TLR2KO mice had a significant reduction in locomotor activity compared with the TLR2KO control mice (P < 0.001). There were no differences observed between the diabetic TLR2KO mice or diabetic WT mice and WT control mice. Diabetes in the TLR2KO mice led to a decrease in total novel arm entries (Fig. 6C), which would indicate an apparent decrease in short-term hippocampus-dependent learning in these mice (P < 0.05). Diabetes in both the WT and TLR2KO mice significantly reduced the number of total arm entries (Fig. 6D) compared with their respective genotype nondiabetic controls (P < 0.001). Interestingly, given that the diabetic animals had fewer total arm entries and that total distance traveled was not different in the WT, with the TLR2KO diabetic mice having a similar distance traveled as the WT groups, this indicates that both diabetic mice groups exhibited an anxious phenotype wherein they traveled comparable amounts but did not leave the confines of the arm in which they were located.

Fig. 6.

Y-Maze testing in WT and TLR2KO nondiabetic and diabetic mice. Diabetes does not change distance traveled (A) or mean velocity (B) in WT mice. Again, TLR2KO CTL mice exhibit a hyperactive phenotype, which is decreased by diabetes. There was no difference in novel arm entries (C) among the WT mice, but in TLR2KO mice, diabetes caused a decrease, which suggests that diabetes in TLR2KO leads to a reduction in short-term hippocampus-dependent learning. WT and TLR2KO diabetic mice had fewer Total arm entries (D), which suggests an anxious phenotype as they remained in the arm. n = 12 or 13. *P < 0.05 vs. WT CTL. #P < 0.05, vs. TLR2KO CTL.

TLR2KO protects against diabetes-induced impairment in long-term hippocampus- and prefrontal cortex-dependent fear learning.

To determine whether diabetes and TLR2KO would have an impact on long-term memory formation, we used two types of fear-conditioning paradigms. Amygdala-dependent long-term memory formation was assed via a cued fear-conditioning paradigm, and both hippocampus-dependent and prefrontal cortex-depending long-term memory formatting were assessed via a contextual fear-conditioning paradigm.

In this training paradigm, both nondiabetic and diabetic WT and TLR2KO groups responded to the shock during training to an acceptable level (above 10% freezing over the 14-min training period), with diabetic TLR2KO mice exhibiting an increased response (P < 0.01) (Fig. 7A). This indicates that all four groups of mice were able to perceive the unconditioned shock stimulus. The cued testing revealed that nondiabetic TLR2KO control mice exhibit a decreased freezing behavior compared with nondiabetic WT control mice, suggesting that they have a genotype-induced impairment in amygdala-dependent memory formation (Fig. 7B). There was no significant effect of diabetes in either genotype compared with their respective nondiabetic controls. In this conservative dual-pairing training paradigm, the diabetic WT mice had a significant deficit in freezing behavior compared with nondiabetic WT control mice in the contextual test at days 1, 4, and 14 (P < 0.05) (Fig. 7C). TLR2KO prevented the effect of diabetes to cause a deficit in freezing behavior at all three time points when compared with nondiabetic TLR2KO control mice, indicating that they were protected from a reduction in both long-term hippocampal (day 1) and prefrontal cortex (days 4 and 14)-dependent memory formation. Both nondiabetic and diabetic TLR2KO mice exhibited a pronounced trend (P = 0.10) toward an increase in the rate of decay of freezing behavior;, i.e., the decrease in freezing behavior over three time points compared with WT groups, indicating that memory formation and recall may be significantly compromised.

Fig. 7.

Fear-conditioned learning in Y-Maze testing in WT and TLR2KO nondiabetic and diabetic mice. A: training indicates that all four groups of mice freeze to an acceptable degree in response to the shock stimulus, with TLR2KO STZ mice exhibiting an increased response. B: diabetes did not have an effect on cued learning in WT, but TLR2KO CTL mice exhibited a decrease compared with WT CTL, suggesting that they have an impairment in amygdala-dependent fear learning. C: diabetes in WT mice lead to a decrease in hippocampal (day 1) and prefrontal cortex (days 4 and 14)-dependent fear learning, and TLR2KO protected against this effect of diabetes, as well as showing an overall trend toward an increased rate of learning decay (not significant, P = 0.10). n = 9–15 for both groups. *P < 0.05 vs. WT CTL. #P < 0.05 vs. TLR2KO CTL.

DISCUSSION

Reductions in cerebral perfusion have been described in both T1D patients, as well as animal models (16, 30, 43, 45, 46). This reduction in cerebral perfusion in diabetes has been linked to the presence of inflammation and microvascular disease (38). It was also postulated that this decrease in cerebral perfusion may be contributing to amplified cognitive decline in the diabetic subjects. Increasingly the role that inflammation plays in the development and progression of diabetic vascular disease has become a critical area of focus in the treatment of this disease (35). A great deal of work has been performed examining how the innate immune system contributes to the overall inflammatory profile in diabetes, particularly TLR pathway signaling in other organ systems (12, 29, 30, 32). However, the role of TLR2 in regulation of cerebral perfusion and cognitive function in diabetes has not been examined. Additionally, while T2D is more common in the United States, T1D is still an important and difficult illness that affects millions of people. The entirety of cerebral complications of T1D and their overall contribution to the health of this patient population are poorly studied. It is likely that cerebral microvascular damage leads to cognitive decline, even in children with T1D. Although the peak age of onset of T1D is 14 years of age, the disease can occur as early as a few years of age, which suggests potential consequences for the effect of hyperglycemia on the developing brain. Studies have shown that those patients who develop diabetes at an early age score worse on cognitive tests relative to their peers, affecting the development of intelligence, attention, executive function, and psychomotor speed (10, 37, 46). Alarming increases in the number of younger patients diagnosed with T1D intensifies this problem (17, 28).

We chose to assess cerebral perfusion in both the early stage at 6 wk and long-term diabetes at 24 wk postinduction. At the early-stage time point, we assessed cortical perfusion using high-resolution laser speckle contrast imaging and observed that diabetic WT mice had decreased CBF in areas supplied by the MCA compared with control WT mice. Diabetic TLR2KO mice were protected from this reduction, maintaining perfusion at levels similar to both control WT and TLR2KO mice. Diabetes is known to affect the microvasculature, as well as the macrovasculature, and contribute to vascular remodeling and dysfunction. Laser-speckle contrast imaging is able to assess cortical perfusion, but it is unable to provide a measure of perfusion to subcortical structures. Arterial spin labeling (ASL) MRI analysis is a well-recognized method for noninvasive assessment of cerebral perfusion and is based on the high diffusibility of water between blood and tissue (11, 57), allowing imaging of cerebral perfusion throughout the brain. Thus, we chose to use ASL to examine how long-term diabetes affects whole brain perfusion. We observed a significant increase in whole brain perfusion in the diabetic WT mice. This increase in cerebral perfusion was prevented in our diabetic TLR2KO mice, which maintained perfusion similar to that seen in both the control WT and TLR2KO mice. While again, TLR2KO prevents the impact of long-term diabetes on cerebral perfusion relative to WT, when compared with our findings in the early stage of the disease, the direction of change in cerebral perfusion due to diabetes was reversed. Several possible explanations for this reversal of decreased cerebral perfusion in diabetes from a state of hypoperfusion to hyperperfusion should be considered. First, the difference in the method of cerebral perfusion determination between LSCI and ASL MRI (i.e., cortical vs. whole brain perfusion) introduce a variable that could affect the results; however, both methods of cerebral perfusion determination are well validated, and we have previously published in our laboratory using ASL MRI that cerebral perfusion is decreased in both the cortex and striatum in early diabetes (24). A similar early reduction in cerebral perfusion and region-specific metabolic alterations were reported in other studies (3, 58). Studies in human T1D patients with long-term diabetes have demonstrated an increase in cerebral perfusion using SPECT and MRI methods (18, 25, 33, 49). Mirroring the results in the clinical setting, STZ-induced diabetes has been shown to increase cerebral perfusion in animal models (52, 56). These previous studies provide support to our findings of reversal from hypoperfusion to hyperperfusion, as diabetes progresses. In fact, hypoperfusion has been shown to lead to impaired cerebral autoregulation (4), and this loss of autoregulation then results in cerebral hyperperfusion due to an inability to respond to alterations in perfusion pressure (21, 43). We have previously shown early decreased perfusion and loss of cerebral autoregulation as disease progresses in a T2D model. Therefore, it is possible that the initial hypoperfusion in early-stage diabetes led to a loss of autoregulation in our diabetic WT mice, which then caused a concomitant increase in cerebral perfusion (41, 42). In our study, it is possible that cerebrovascular autoregulation is preserved in the diabetic TLRKO STZ mice due to a reduction in TLR2-mediated inflammation and subsequent vascular damage.

Diabetic mice exhibit an impairment in endothelium-dependent vasorelaxation (15, 22), which could play a role in the alterations in the cerebral perfusion that we observed. Previous work by Shah et al. (51) has demonstrated that DPP-IV inhibitors, a class of oral antihyperglycemic drugs, exert beneficial pleiotropic effects on the vasculature, including improvement of endothelium-dependent vasodilation. Linagliptin, one of the newest of the DPP-IV inhibitors, was also shown to have a direct vasodilatory effect due to the NO/cGMP signaling pathway, as well as reducing ROS, decreasing vascular expression of inflammatory genes, and decreasing vascular infiltration of inflammatory cells. We chose to test whether a direct vasodilatory response to linagliptin was impaired in diabetic mice, and whether TLR2KO would protect from this effect. Our findings show that TLR2KO confers endothelial protection, likely through a NO-dependent mechanism, and this may contribute to the maintenance of CBF that we observed in diabetic TLR2KO mice.

On the basis of our findings demonstrating TLR2KO-mediated protection from diabetes-induced alterations of cerebral perfusion and endothelium-dependent vascular dysfunction, we sought to determine whether cognitive function in diabetic TLR2KO mice was similarly protected. Given the work of others examining hippocampal function in STZ-induced diabetes (53, 55), we chose to use NOR and Y-maze assessments of short-term hippocampal memory. Both of these tests rely on locomotor and exploratory activity of the animals in assessment of cognitive function, and as such, the locomotor behavior between the four groups of mice in our study needed to be assessed as a potential confounding variable. We performed locomotor activity testing, which revealed that TLR2KO has a hyperactive phenotype. Diabetic TLR2KO mice did not exhibit the same increased locomotor activity, which suggests that diabetes ameliorates the hyperactivity in this genotype.

One method of testing long-term hippocampus and prefrontal cortex-dependent learning that does not depend on locomotor activity for assessment is fear-conditioned learning. STZ-induced diabetes has been shown to lead a reduction in long-term hippocampus-dependent fear learning assessed during contextual fear-conditioning testing (6). We similarly observed impairment in long-term hippocampus-dependent learning at 24 h, as well as impaired prefrontal cortex-dependent learning at 4 and 14 days in the diabetic WT STZ mice. At all three time points, TLR2KO was able to protect against a diabetes-induced reduction in contextual fear-conditioned learning, suggesting that the effects of diabetes on long-term hippocampus and prefrontal cortex-dependent learning are influenced by TLR2. Additionally, there was an overall trend toward an increased rate of learning decay in the TLR2KO mice compared with nondiabetic WT mice, which although it did not reach statistical significance (P = 0.1), suggests that TLR2KO could have an effect on transfer of long-term memory from a hippocampus-dependent state to a prefrontal cortex dependent state.

TLR2KO mice were shown to exhibit an impairment in the cued fear-conditioning task, which suggests an impairment in amygdala-dependent learning (2). The impairment in amygdala-dependent learning implies that the nondiabetic TLR2KO mice have an anxiolytic phenotype, which confirms the reduction in anxiety that we observed in these mice by the increased center-periphery distance traveled ratio in open-field testing. This effect is similar to the anxiolytic effect of TLR2KO observed by Park et al. (40) but contradictory to the lack of an effect on anxiety in the study by Madar et al. (34) The fear-conditioned learning paradigm used by Madar et al. was based on a three-paired stimulus response as opposed to a two-paired approach in our study, making our well-validated approach (9, 14) more conservative given that it is easier to evoke an effect on learning and memory with increasing pairings. Madar et al. (34) found that TLR2KO mice exhibited an increased freezing response in the contextual fear paradigm, as well as an increase in cue-induced fear. It is possible that these conflicting results are due to differences in the design paradigm of fear-conditioned learning, or again due to the discrepant ages used between their study and this current work.

Although our results show that TLR2KO confers protection from diabetic changes in cerebral perfusion and cerebrovascular dysfunction, this may not be the sole explanation for the associated protection from cognitive impairment that we observed. We used a global TLR2KO mouse model, and TLR2 deficiency has been shown to cause a variety of developmental effects, such as suppression of neuronal differentiation and shifting neural progenitor cells toward an astrocytic fate (39). TLR2 deficiency can also exacerbate memory impairments in mouse models of Alzheimer’s disease, as it is proposed that TLR2 may act as an exogenous receptor for the clearance of amyloid-β (7, 44). TLR2 activation has been shown to be important in neural plasticity in the early postnatal period, such that exposure to TLR2 ligand in infants can lead to pronounced learning and memory issues later in life (34). Thus, it is clear that TLR2 signaling impacts cognitive function both positively and negatively, and the effects occur as a result of neural and inflammatory issues. Our studies show that TLR2 activation also has a pronounced effect on vascular function and cerebral perfusion, and this effect begins early in the course of the diabetes. Additionally, these changes are associated with deficits in cognitive function, in particular, long-term hippocampus- and prefrontal cortex-dependent learning.

GRANTS

This work was supported in part by Veterans Affairs (VA) Merit Award (BX000347), VA Research Career Scientists Award, National Institutes of Health Grants (NS-070239, R01NS-083559), a research grant from Boehringer Ingelheim Pharmaceuticals to A. Ergul, and an American Heart Association Predoctoral Fellowship (15PRE25760034) to T. Hardigan.

DISCLOSURES

Boehringer Ingelheim Pharmaceuticals, Inc., provided both the linagliptin and financial support for the study.

AUTHOR CONTRIBUTIONS

T..H., C.H., and A.E. conceived and designed research; T.H., C.H., and M.N.H. performed experiments; T.H., C.H., and M.N.H. analyzed data; T.H., C.H., R.W., M.N.H., and A.E. interpreted results of experiments; T.H. prepared figures; T.H. drafted manuscript; T.H. and A.E. edited and revised manuscript; T.H., C.H., R.W., M.N.H., and A.E. approved final version of manuscript.