Intranasal insulin reverts central pathology and cognitive impairment in diabetic mother offspring

Background Adverse effects in diabetic mothers offspring (DMO) are a major concern of increasing incidence. Among these, chronic central complications in DMO remain poorly understood, and in extreme cases, diabetes can essentially function as a gestational brain insult. Nevertheless, therapeutic alternatives for DMO are limited. Methods Therefore, we have analyzed the central long-term complications in the offspring from CD1 diabetic mothers treated with streptozotozin, as well as the possible reversion of these alterations by insulin administration to neonates. Brain atrophy, neuronal morphology, tau phosphorylation, proliferation and neurogenesis were assessed in the short term (P7) and in the early adulthood (10 weeks) and cognitive function was also analyzed in the long-term. Results Central complications in DMO were still detected in the adulthood, including cortical and hippocampal thinning due to synaptic loss and neuronal simplification, increased tau hyperphosphorylation, and diminished cell proliferation and neurogenesis. Additionally, maternal diabetes increased the long-term susceptibility to spontaneous central bleeding, inflammation and cognition impairment in the offspring. On the other hand, intracerebroventricular insulin administration to neonates significantly reduced observed alterations. Moreover, non-invasive intranasal insulin reversed central atrophy and tau hyperphosphorylation, and rescued central proliferation and neurogenesis. Vascular damage, inflammation and cognitive alterations were also comparable to their counterparts born to nondiabetic mice, supporting the utility of this pathway to access the central nervous system. Conclusions Our data underlie the long-term effects of central complications in DMO. Moreover, observed improvement after insulin treatment opens the door to therapeutic alternatives for children who are exposed to poorly controlled gestational diabetes, and who may benefit from more individualized treatments.


Background
Gestational diabetes affects 3-10% of pregnant women [1] and epidemiological and animal studies have previously shown that the risk of adverse maternal and perinatal outcomes continuously increases with maternal glycaemia [2]. Following this idea, malformation and mortality rates are reportedly two to five fold higher in diabetic mothers offspring (DMO) (for review [3]). Moreover, these foetal and neonatal complications seem to continue at later stages and have been largely studied at metabolic level [4]. Central nervous system (CNS) complications in DMO are also receiving attention in recent years and many studies have focused on hypothalamic alterations in relation with future metabolic disorders [5,6]. Prolonged hyperglycaemia during critical periods of development underlie malformations in the foetal brain [7] and insulin is also an important regulator of developmental and cognitive functions in the CNS [8]. Optimal control of glucose levels during pregnancy is highly pursued, however, short and long-term related complications are still present and in extreme cases, diabetes has been suggested to essentially function as a gestational brain insult [9] resulting in behavioural problems, language impairment or cognitive development deficits [9,10]. However, to our knowledge exhaustive chronic evaluation of the CNS has not been performed, and the study of therapeutic options should also be addressed. Therefore, we have assessed short and long-term metabolic and central complications in DMO. Moreover, we have attempted to reverse observed alterations by treating DMO with insulin. Although previous studies have treated diabetic mothers with insulin to counterbalance disturbances in the offspring, to our knowledge no preceding studies have directly administered intracerebroventricular (ICV) or intranasal (IN) insulin to DMO. In our hands, DMO presented metabolic alterations in the adulthood, including modified glucose and insulin levels in glucose tolerance tests, while insulin administration significantly improved this aspect. When we analyzed the CNS, DMO presented long-term cortical and hippocampal thinning due to neuronal simplification and synaptic loss, accompanied by reduced central proliferation and neurogenesis, while insulin treatments reverted these effects. Also, spontaneous central bleeding in DMO was significantly improved after insulin administration, and markers of neuronal damage, such as tau hyperphosphorylation were controlled by insulin. Cognition impairment was significantly reversed after insulin treatment. An overall improvement was observed after both, IN and ICV insulin administration, nevertheless, IN pathway lead to a more robust recovery, supporting the utility of this approach to guarantee insulin access to the CNS. Altogether our data could help to elucidate the underlying central complications in DMO and open the door to therapeutic alternatives for children who are exposed to poorly controlled gestational diabetes.

Animals and treatments
Two months old CD1 breeders were treated with streptozotozin (STZ) (50 mg/Kg) for 5 consecutive days [11] prior crossing with healthy males. Both female and male offspring from diabetic and control mothers were randomly divided and sacrified immediately after birth or at P7. A set of animals received unilateral ICV insulin (ICV-Ins) injection at P7 [12,13]. Briefly, mice were anesthetized with isoflurane (Astrazeneca, Spain) and placed in a sterotaxic device (David-Kopf Instruments, Tujunga, CA, USA). ICV administration consisted of 1 μl of insulin (5 mIU) in PBS, at the following coordinates: AP -3 mm, ML -1 mm y DV +4 mm from Bregma. Injections were performed with a 5 μl Hamilton syringe (Hamilton Company, Bonaduz, Switzerland) at a constant flow rate of 0.2 μl/min for 5 min. A delay of 5 min was allowed before complete retraction of the needle to minimize aspiration of the toxin. Sham operated mice followed the same procedure but only PBS was injected. A second group received IN insulin (IN-Ins), for 7 consecutive days (P7-P13) as previously described [14,15]. Briefly, mice received 1 IU/day of insulin in PBS. Mice were lightly anesthetized with isoflurane and held gently on their backs. A 10 μl pipette was used for intranasal administration (1.25 μl of insulin every minute) alternating nostrils. Insulin drops were placed in the opening of the nostril, allowing the mouse to snort each drop into the nasal cavity. A total of 5 μl of insulin were delivered over a course of 4 min. Control mice received PBS following a similar approach. Long-term effects of the ICV-Ins and IN-Ins treatments was assessed at 10 weeks of age. For proliferation and neurogenesis studies animals received BrdU ip (70 mg/ Kg) for 3 consecutive days immediately before sacrifice.
DMO were assigned to foster mothers immediately after birth to guarantee regular postnatal feeding and development. Groups and treatments are presented in Fig.  1g. Metabolic and postmortem characterization included 4-6 animals per group at P1 and 4-7 mice per group at P7. Five weeks old assessment included 5-15 mice and 10 weeks old studies included 6-28 mice per group for behavioural and metabolic assessment and 3-10 mice per group in postmortem studies. Individual sample size for each experiment is detailed in figure legends. In order to guarantee the access of insulin to the brain, a set of animals (n = 3-4) was treated with ICV-Ins or IN-Ins, as described above, and scarified to measure brain insulin levels in the olfactory bulb, cortex, hippocampus and striatum. All experimental procedures were approved by Junta de Andalucia, (Guidelines for Care and Use of Experimental Animals, European Commission Directive 2010/63/UE and Spanish Royal Decree 53/2013).

Brain insulin levels
Brain insulin levels were measured in the olfactory bulb, cortex, hippocampus and striatum of a set of mice, after ICV-Ins and IN-Ins, to guarantee the access to the brain after both administration pathways. Briefly, animals received an overdose of pentobarbital (60 mg/Kg) (Sigma, St. Louis, MO, USA Sigma MO). Selected brain regions were harvested and snap frozen until used. Tissue was homogenised in lysis buffer (CellSignaling, USA) supplemented with a protease inhibitor cocktail (Sigma, USA) and insulin levels were determined by human enzymelinked immunosorbent assay (Mercodia, Spain Iso-Insulin ELISA 10-1128-01).

Metabolic characterization
Body weight was determined in the mothers before STZ administration and on the day of birth, and weight gain was calculated. Postprandial glucose levels were measured before STZ treatment, one week after STZ administration, after crossing and the day of birth. Insulin levels were also measured immediately before STZ administration and the day of birth. Body weight, glucose levels and insulin levels were determined in all offspring (both male and female) the day of birth, and on P7. Endpoint experiments were programmed to determine glucose and insulin levels due to the volume of blood required. Briefly, animals were deeply anesthetized with pentobarbital (60 mg/kg) and blood was withdrawn by cardiac puncture. Glucose an insulin levels were also determined after ICV-Ins or IN-Ins administration, at 5 and 10 weeks of age. Additionally, at 10 weeks of age, before sacrifice, intraperitoneal glucose tolerance test (GTT) (2 g/Kg body weight, Sigma, OR; USA) was performed [16]. Blood glucose levels were measured in mg/dl using the glucometer Optium Xceed (Abbott, United Kingdom). Plasma insulin levels were measured in ng/ml using ultrasensitive mouse enzyme-linked immunosorbent assay (Mercodia, Spain Mouse Insulin ELISA 10-1247-01).

Morris water maze (MWM)
Spatial cognition assessment commenced in the same mice the day after concluding the NOD test. Acquisition consisted in 4 trials/day (60 s/trial) for 4 days, with the platform submerged in quadrant 2. If the animal did not find the platform it was placed on it for 10 s. Retention started a day after and consisted in a single trial with the platform removed [17]. Time required to locate the platform in the acquisition phase, percentage of time spent in quadrant 2 during the retention phase and swim speed were analyzed using Smart software, (Panlab, Spain).

Tissue processing
Cortex and hippocampus from the right hemisphere, from both male and female offspring, were dissected and snap frozen for cresyl violet, Prussian blue, caspase activity and western blot studies. Left hemispheres, from both male and female offspring, were fixed in paraformaldehyde for 2 weeks before 30 μm coronal sections were obtained. In case of ICV treated mice, only ipsilateral hemispheres were used and brains from males and females were either dissected or snap frozen. Similarly, P1 and P7 tissue was snap frozen for biochemical determinations.

Cresyl violet and Prussian blue staining
Brain morphology was analyzed after cresyl violet staining in sections (1.5 to −3.5 mm from Bregma). Cortex and hippocampus thickness were measured using Adobe Photoshop and Image J software as previously described [18].
Presence of haemorrhages was conducted using Prussian blue iron staining and neutral red counterstain in consecutive sections. Sections were imaged with an Olympus Bx60 microscope (Japan) and an Olympus DP71 camera. Images were analyzed using Adobe PhotoShop and Image J softwares to quantify number of haemorrhages, individual haemorrhage size, and haemorrhage burden in the cortex and hippocampus [18].

Caspase activation
Caspases 3/7 activity was analyzed in cortical homogenates from all groups in this study using the Caspase-Glo 3/7 assay (Promega, Madrid, Spain), following manufacturer's indications, as previously described [19].
NeuN and microglia immunohistochemistry was performed as described [20]. Anti-Iba1 (Wako, Osaka,Japan) was quantified as previously described [21]. Number of microglia cells, individual microglia size and burden were quantified in the cortex and hippocampus using Image software [20].

Golgi-Cox staining
Neuronal complexity was measured by Golgi-Cox staining, using Rapid Golgi Stain Kit (FD Neurotechnologies, USA. Ref: PK401). Kit instructions were followed as previously described [20] in all of our animals (both male and female). Neuronal complexity was analyzed by sholl analysis in 10 μm concentric circles from neuronal soma. Spine density was calculated (spines/10 μm) [20]. Ratios of curvature were calculated by dividing the end-to-end distance of a dendrite segment by the total length between the two segment ends. Analysis was completed using Image J software, as previously described [22].

Statistical analysis
Student t test for independent samples or one-way ANOVA, followed by Tuckey b test or Tamhane tests as required, were used. Two-way ANOVA (groupXday) was used to analyze the the MWM test. SPSS v.15 software was used for all statistical analysis.

Metabolic characterization
No differences were detected among mothers before STZ treatment when body weight (p = 0.850), glucose levels (p = 0.798) or insulin levels were compared (p = 0.382) ( Table 1). We detected a significant increase of glycaemia one week after STZ treatment (**p = 0.005 vs. Control). At the date of birth body weight was significantly lower in STZ-treated mothers (**p = 0.003 vs. Control) and the increase of body weight was also significantly compromised (**p = 0.003 vs. Control). The day of birth insulin levels were lower in STZ-treated mothers (**p = 0.005 vs. Control) and glucose levels were highly increased (**p < 0.001 vs. Control) ( Table 1).
Insulin levels in the brain were also tested after IN and ICV delivery. An overall increase of central insulin levels was observed in all regions under study. As it could be expected, higher insulin levels were observed in the olfactory bulb after in intranasal administration, while insulin levels were higher in the hippocampus and the striatum after ICV administration (Fig. 1f ).   (Fig. 2d).

Proliferation and neurogenesis
In the SVZ the number of BrdU+ cells was significantly compromised in DMO (Fig. 4a and d) (Fig. 4c).

Akt levels
Phopho-Akt/total Akt levels were reduced in the cortex of DMO at P1 (**p = 0.001 vs. Control) and P7 (**p < 0.001 vs. Control). By 10 weeks of age, phospho-Akt levels were still reduced in DMO and insulin administration slightly increased the ratio, although differences were not statistically significant [F (5,29) = 2.34, p = 0.066] (Fig. 5a and b). A similar profile was observed in the hippocampus at P1 (**p < 0.01 vs. Control), P7 (**p < 0.01 vs. Control) and by 10 weeks of age [F (5,30) = 1.98, p = 0.390] (Fig. 5a). Since it is feasible that the time selected after insulin delivery might be too long we also measured Phopho-Akt/total Akt ratio 4h after ICV or intranasal administration. Phospho-Akt/total Akt ratios were significantly increased in those regions located in the proximity of the administration site: the hippocampus and the striatum after ICV administration  (Fig. 5c and d).

Discussion
Increased risk of metabolic and cardiovascular diseases in DMO have been reported in animal and human [24] studies. While cognitive alterations in DMO have also been observed [9,10], it is noteworthy that most of previous studies are based on observational cohorts. Therefore a direct causal influence of intrauterine hyperglycemia remains uncertain [25] and therapeutic options have not been completely addressed. To help elucidate some of these aspects, we have analyzed the short and long-term central complications in a murine model of DMO, as well as the effects of ICV and non-invasive IN insulin treatments. Our approach included administering STZ to mothers before crossing [26], and central complications in DMO have been also observed when mothers are mated and diabetes is induced immediately afterwards [27]. However we cannot exclude that final outcomes might be slightly different depending on the protocol used, time selected to induce diabetes or endpoints under study. The fact that glucose levels recovered 7 days after birth might be due to regularization of basal levels once the pups are no longer in intrauterine environment. Increased insulin levels did not reach statistical significance, however it remains feasible that the slight a hyperglycemic increase observed might be enough to control glycaemia in our population. In DMO we also observed growth inhibition, impaired GTT and insulin response, as previously reported in DMO, that show decreased insulin sensitivity and low insulin secretion [28]. The fact that DMO do not regularly release insulin in the GTT suggests a pancreatic exhaustion, over insulin resistance. Nevertheless, we cannot unequivocally point towards a single cause leading to observed metabolic alterations. On the other hand, central insulin administration reversed all these effects, supporting the role of insulin in the regulation of peripheral metabolic complications [29,30].
Central pathology in young adults from diabetic mothers revealed the long term effects of severe maternal hyperglycemia in learning and memory. Epidemiological studies have shown that maternal diabetes is negatively associated with offspring's cognitive development [10] and gestational diabetes has been related to lower general intelligence, language impairment, attention weakness, impulsivity or behavioural problems (for review [9]). A few studies have assessed the ongterm effects of maternal diabetes in the descendants [31] and to our knowledge this is the first one trying to reverse observed deficits by administering insulin to the offspring. Spatial and episodic memory impairment in DMO improved after ICV insulin administration, and were completely reversed after IN insulin. While ICV administration provides useful tool to guarantee access to the CNS in animal models [30], its translational approach is not feasible. On the other hand, IN insulin uses a transport system, via the nasal epithelium into the brain, that avoids high levels of insulin in the periphery. Moreover, previous studies with IN insulin support that insulin signal is crucial in neuronal function and cognition, improving cognitive function in healthy subjects [32], diabetic patients [33] and murine models [34], however to our knowledge no previous studies have tested insulin administration to restore cognitive deficits associated to intrauterine hyperglycaemia. Significant brain atrophy was observed in DMO up to early adulthood, associated with increased caspase activity, cortical and hippocampal thinning, in line with previous studies [35]. Likewise, NeuN/DAPI ratios were reduced in DMO, while insulin treatment completely reversed these observations. In order to further characterize brain atrophy, we examined neuronal morphology, since it has been shown to correlate with neuronal function [36]. Neurite curvature was not significantly altered in DMO, however, severe reduction of neuronal ramifications, spine density and synaptophysin levels were observed, in accordance with previous studies [37]. Synaptogenesis is a relevant event during development that requires a fine and precise pre-and postsynaptic specialization [35]. Therefore, interferences at this level may result in abnormal layer development and related cognition and behavioural problems. Insulin treatment fully recovered neuronal complexity and spine density in DMO, in line with previous studies showing that altered maternal metabolism may impair offspring neuronal projections normal development [38]. It has been already reported that foetal hyperglycaemia alters the expression of genes involved in proliferation and differentiation of neural cells [35]. Insulin also plays a crucial role in proliferation and neurogenesis in the CNS and in our hands, DMO presented a significant reduction of both processes, especially relevant in a neurogenic niche as the SVZ. Insulin signaling in the brain can respond to changes in systemic metabolic state while local insulin signaling promotes neurogenesis [39], as observed in our DMO after insulin treatment. Since Akt plays a crucial role in insulin signaling, as well as in cell proliferation and cell survival, it has been previously suggested that variations in Akt may underlie the observed alterations in DMO [40]. We also observed a reduction in phospho-Akt levels in young DMO, supporting that diabetes during pregnancy strongly influences the regulation of Akt in the developing brain, and that strict maternal metabolic control might be crucial [26]. Insulin treatment slightly increased phospho-Akt levels. Limited differences observed might be due to the fact that determinations were performed days and weeks after insulin administration. Therefore, we also performed determinations 4 h after insulin administration and we detected that phospho-Akt/total AKt ratios were significantly increased in those regions located in the proximity of the administration site: the hippocampus and the striatum after ICV administration, and the olfactory bulb after intranasal administration, in line with previous observations [41].
Tau hyperphosphorylation exerts toxic effects and it is a pathological hallmark of neurodegenerative disorders, also described in different diabetic animal models [17,19]. It has been suggested that hyperglycaemic conditions induce tau deregulation [42]. We observed a significant increase of hyperphophorylated tau in DMO and similar outcomes have been observed in other models [43]. However, to our knowledge, the detected improvement after insulin treatment has not been previously reported.
Small vessel disease is a more common cause of ischemic stroke in people with diabetes [44]. It has been also described that embryos from diabetic mothers present generalized vascular lesions [45], however we believe no previous studies have assessed spontaneous central bleeding in DMO. Haemorrhage burden was increased in P7 DMO, and this effect was still detectable in the adulthood, supporting the long-term effects of maternal diabetes in the offspring. Previous studies have also associated the presence of haemorrhages with an increase in tau phosphorylation [46,47]. Moreover, insulin treatment can effectively reverse these effects. We also detected a significant increase in microglia burden up to early adulthood in DMO, in line with previous studies showing that maternal prediabetes is enough to increase microglia activation and cytokines involved with trafficking across the blood-brain barrier [48], suggesting an activation of the local inflammatory response that may ultimately contribute to observed spontaneous bleeding. Furthermore, insulin treatment reduced microglia activation, supporting its role as an anti-inflammatory agent (for review see [49]).

Conclusions
These data suggest that maternal diabetes not only affects the CNS development but it also has long-term effects, that are still evident in the adulthood. Maternal diabetes affects neuronal complexity and synaptic density, tau hyperphosphorylation, central inflammation or spontaneous bleeding, which may altogether compromise learning and memory abilities. Moreover, ICV insulin administration counterbalances many of these aspects and non-invasive IN administration robustly reverses detected alterations. This could lead to studies in greater depth on the use of IN insulin as a therapeutic alternative for those infants from diabetic mothers, who may benefit from more exhaustive follow-up assessments [9] and individualized treatments.