Differences between normal and diabetic brains in middle-aged rats by MRI

Abstract

Normal aging is a risk factor for metabolic disorders such as diabetes, and diabetes is also a recognized cause of accelerated aging. Being able to distinguish changes caused by normal aging from those caused by diabetes, would provide insight into how the aging brain interacts with diabetes. Eight types of MRI metric maps (magnetization relaxation time constants of T₁ and T₂, cerebral blood flow, cerebrovascular permeability, mean diffusivity, diffusion fractional anisotropy, mean diffusion kurtosis and diffusion directional entropy) were generated for all rats from the three groups of normal young, healthy and 1.5-month diabetic middle-aged rats under investigation. Measurements of multiple MRI indices of cerebral white and gray matter from animals of the three groups provide complementary results and insight into differences between healthy and diabetic white / gray matter in the mid-aged rats. Our data indicate that MRI may distinguish between the normal and diabetes in mid-aged rat brains by measuring either T₁ and T₂ of gray matter, or fractional anisotropy of white matter and gray matter. Therefore, MRI can distinguish changes of cerebral tissue due to the normal aging from diabetic aging, which may lead to be able to better understand how diabetes accelerates aging in normal brain.

1. Introduction

Normal aging is a complex and heterogeneous process and is a risk factor for metabolic disorders such as type 2 diabetes mellitus (T2DM). The vast majority (90–95%) of diabetic patients in the U.S., i.e., 24 million Americans, are T2DM. In T2DM the production of insulin is either insufficient for metabolism or the cells of the body do not properly respond to the insulin produced (insulin resistance) (Ergul et al., 2009). Although T2DM may first appear at any age, its prevalence rises dramatically in older populations (Horwitz, 1982). Overall, nearly one in three older adults meets current American Diabetes Association criteria for diabetes, and fewer than 30% of the U.S. population over age 65 have normal glucose levels by American Diabetes Association criteria (Halter, 2012).

Diabetes is also a recognized cause of accelerated aging. Central nervous system (CNS) changes observed in aging diabetic patients appear to accelerate degenerative brain dysfunction. The commonalities between diabetes-induced and age-related CNS changes have led to the hypothesis of advanced brain aging in diabetic patients (Wrighten et al., 2009), although the mechanisms linking diabetes and aging are not well understood (Ahima, 2009; Minamino et al., 2009). Therefore, being able to distinguish CNS changes caused by normal aging from those caused by diabetes, would provide insight into how the aging brain interacts with diabetes, which may also impact the management of older patients with diabetes.

Magnetic resonance imaging (MRI) has been employed to characterize brain changes with aging (Breger et al., 1991; Chen et al., 2011; Chun et al., 2000; Das et al., 2016; Ge et al., 2002) and diabetes (Brundel et al., 2014; Last et al., 2007; Reijmer et al., 2013; Starr et al., 2003). MRI structural neuroimaging has shown that brain atrophy is an important feature of T2DM patients (Schmidt et al., 2004). The differences of MRI indices of cerebral tissue attributed to aging versus diabetes appeared to be dependent on region of brain tissue, stage of aging and duration of hyperglycemia (Breger et al., 1991; Brundel et al., 2014; Chen et al., 2011; Chun et al., 2000; Das et al., 2016; Ge et al., 2002; Last et al., 2007; Reijmer et al., 2013; Starr et al., 2003). Previous investigations have also demonstrated that conventional MRI measurements, such as diffusivity and relaxation time constant T₁ (or relaxation rate constant R₁ = 1/T₁) of water in cerebral tissues, were not monotonically correlated with age in healthy human brain, where a maximum diffusivity and minimum T₁ were reached at onset of mid-age (about 40 years old), respectively (Yeatman et al., 2014). However, existing MRI studies have not systematically evaluated multiple MRI metrics between healthy and diabetic brains, and have been unable to distinguish changes between normal aging brain and age-matched diabetic brain (Das et al., 2016; Falangola et al., 2008; Gong et al., 2014; Yeatman et al., 2014). In the present study, we therefore, performed multiple MRI measurements of white matter and gray matter in healthy young adult, healthy mid-aged and non-long-standing diabetic mid-aged rats. We propose that MRI can be employed to distinguish effects of age-related biomedical changes from diabetes-induced biomedical changes in brain.

2. Results

Eight types of MRI metric maps were generated, i.e., longitudinal (spin-lattice) relaxation time constant T₁, transversal (spin—spin) relaxation time constant T₁, relative cerebral blood flow (relCBF), cerebrovascular permeability (by using contrast-enhanced T₁-weighted imaging: CE-T1WI), mean apparent diffusion coefficient (i.e., mean diffusivity: MD), diffusion fractional anisotropy (FA), mean diffusion kurtosis (MK) and diffusion directional entropy (En), for all rats from the three groups of healthy young adult, healthy mid-aged and non-long-standing diabetic mid-aged rats under investigation. Fig. 1 shows representative regular non-diffusion related maps, and Fig. 2 shows representative conventional diffusion related maps. An intuitive analysis of Fig. 1 demonstrated that healthy young adult rats had longer T₁, T₂ and relCBF values with redder pseudo-color images, but lower cerebrovascular permeability (per CE-T1WI) with bluer pseudo-color image, compared to healthy mid-aged and non-long-standing diabetic mid-aged rats. As to diffusion related maps of MD, FA and En (entropy) in Fig. 2, the differences among these rats were not apparent by reviewing the grayscale images, except that the MK maps, exhibited bright spots with high MK values scattered on the MK map of mid-aged rats while the bright spots were absent on the MK map of the young healthy adult rat.

Fig. 1.

Regular non-diffusion related MRI pseudo-color maps or images from representative rats in the present investigation. From top to bottom, there are T₁ maps (the first row), T₂ maps (the second row), CBF maps (the third row) and CE-T1WI images of blood-to-brain permeability (the fourth row) in order by the representative healthy young adult rat (the left column), healthy mid-aged rat (the middle column) and non-long-standing diabetic mid-aged rat (the right column).

Fig. 2.

Conventional diffusion related MRI maps from representative rats in the present investigation. From top to bottom, there are mean diffusivity maps (the first row), fractional anisotropy maps (the second row), mean kurtosis maps (the third row) and directional entropy maps (the fourth row) in order by the representative healthy young adult rat (the left column), healthy mid-aged rat (the middle column) and non-long-standing diabetic mid-aged rat (the right column).

Results of ANOVA for all three rat groups were listed in Table 1: eight types of MRI measurements from white matter (corpus callosum) and gray matter (cortex).

Table 1.

ANOVA tests of MRI measurements.

Corpus callosum			Cortex
MRI metrics	F value	p value	F value	p value
T1	15.97	2.65E-05	17.39	1.40E-05
T2	50.87	6.95E-10	64.62	5.10E-11
CBF	10.83	3.52E-04	0.41	6.67E-01*
GdT1	10.97	3.26E-04	8.71	1.21E-03
FA	13.46	8.81E-05	4.35	2.30E-02
Entropy	32.69	6.14E-04	1.08	3.55E-01*
MD	4.32	2.37E-02	1.79	1.87E-01*
MK	88.97	1.31E-12	71.72	1.57E-11

^*p > 0.05, no statistical significant differences.

Fig. 3 shows MRI measurements of the corpus callosum (i.e., white matter) for each of the animal groups in the sequence of healthy young adult, healthy mid-aged and non-long-standing diabetic mid-aged rats. As indicated by the histograms, MRI measurements of T₁ (1544 ± 17 ms vs 1511 ± 38 ms vs 1472 ± 27 ms for young healthy adult vs healthy mid-aged vs non-long-standing diabetic mid-aged rats, respectively), T₂ (54.7 ± 1.0 ms vs 52.8 ± 1.9 ms vs 46.0 ± 2.8 ms), relCBF (98.0 ± 7.0 vs 91.7 ± 6.3 vs 79.7 ± 12.3), MK (1.06 ± 0.03 vs 1.29 ± 0.06 vs 1.35 ± 0.05) and entropy (2.84 ± 0.02 vs 2.93 ± 0.05 vs 2.97 ± 0.04), respectively, detected significant white matter changes caused by either aging or diabetes, or both. However, CE-T1WI featured as cerebrovascular permeability for image contrast agent of gadolinium diethylenetriamine pentaacetic acid (Gd-DTPA, Magnevist®, Berlex Inc, Montville, NJ) detected significant white matter changes (−0.86 ± 2.07% vs 2.81 ± 1.71% vs 2.85 ± 2.28%) because of aging only. There were no apparent differences of white matter caused by diabetes (between mid-aged normal and diabetic rats). Conversely, MD (9.7 ± 0.3 × 10⁻⁴ mm²/s vs 9.7 ± 0.5 × 10⁻⁴ mm²/s vs 9.2 ± 0.5 × 10⁻⁴ mm²/s) and FA (0.38 ± 0.01 vs 0.39 ± 0.02 vs 0.42 ± 0.02) successfully identified significant white matter changes specifically due to diabetes, rather than aging.

Fig. 3.

MRI measurements of the corpus callosum to represent white matter. For each of the animal groups in the sequence of healthy young adult, healthy mid-aged and non-long-standing diabetic mid-aged, as indicated by the histograms, MRI measurements (mean ± SD) of T₁ (a): 1544 ± 17 ms vs 1511 ± 38 ms vs 1472 ± 27 ms; T₂ (b): 54.7 ± 1.0 ms vs 52.8 ± 1.9 ms vs 46.0 ± 2.8 ms; relCBF (c): 98.0 ± 7.0 vs 91.7 ± 6.3 vs 79.7 ± 12.3; CE-T1WI (d): −0.86 ± 2.07% vs 2.81 ± 1.71% vs 2.85 ± 2.28%; MD (e): 9.7 ± 0.3 × 10⁻⁴ mm²/s vs 9.7 ± 0.5 × 10⁻⁴ mm²/s vs 9.2 ± 0.5 × 10⁻⁴ mm²/s; FA (f): 0.38 ± 0.01 vs 0.39 ± 0.02 vs 0.42 ± 0.02; MK (g): 1.06 ± 0.03 vs 1.29 ± 0.06 vs 1.35 ± 0.05 and entropy (h): 2.84 ± 0.02 vs 2.93 ± 0.05 vs 2.97 ± 0.04.

MRI measurements appeared to be more sensitive in detecting changes related to age and diabetes in white matter than in gray matter (i.e., cerebral cortex). As indicated by the histogram charts in Fig. 4, MRI measurements (young healthy adult vs healthy mid-aged vs non-long-standing diabetic mid-aged rats, respectively) of relCBF (252.8 ± 54.9 vs 235.6 ± 49.2 vs 197.9 ± 51.5), MD (9.5 ± 0.2 × 10⁻⁴ mm²/s vs 9.7 ± 0.3 × 10⁻⁴ mm²/s vs 9.5 ± 0.3 × 10⁻⁴ mm²/s) and entropy (2.53 ± 0.01 vs 2.56 ± 0.06 vs 2.57 ± 0.09) failed to detect any differences of gray matter among young healthy adult, healthy mid-aged and non-long-standing diabetic mid-aged rats, respectively. Significant gray matter changes caused by aging factor were found for measurements of T₁ (1631 ± 19 ms vs 1581 ± 31 ms vs 1562 ± 29 ms), CE-T1WI (−0.88 ± 4.05% vs 3.67 ± 2.24% vs 3.82 ± 1.77%), FA (0.18 ± 0.01 vs 0.17 ± 0.01 vs 0.16 ± 0.02) and MK (1.00 ± 0.02 vs 1.20 ± 0.04 vs 1.21 ± 0.06). T₂ maps (64.8 ± 0.9 ms vs 64.7 ± 1.7 ms vs 58.5 ± 1.5 ms) identified significant changes in gray matter due to diabetes (p < 0.002 between mid-aged rats with or without diabetes) and not aging (p > 0.89 between young and mid-aged healthy rats).

Fig. 4.

MRI measurements of the cortex to represent gray matter. For each of the animal groups in the sequence of healthy young adult, healthy mid-aged and non-long-standing diabetic mid-aged, as indicated by the histograms, MRI measurements (mean ± SD) of T₁ (a): 1631 ± 19 ms vs 1581 ± 31 ms vs 1562 ± 29 ms; T₂ (b): 64.8 ± 0.9 ms vs 64.7 ± 1.7 ms vs 58.5 ± 1.5 ms; relCBF (c): 252.8 ± 54.9 vs 235.6 ± 49.2 vs 197.9 ± 51.5; CE-T1WI (d): −0.88 ± 4.05% vs 3.67 ± 2.24% vs 3.82 ± 1.77%; MD (e): 9.5 ± 0.2 × 10⁻⁴ mm²/s vs 9.7 ± 0.3 × 10 ⁻⁴ mm²/s vs 9.5 ± 0.3 × 10⁻⁴ mm²/s; FA (f): 0.18 ± 0.01 vs 0.17 ± 0.01 vs 0.16 ± 0.02; MK (g): 1.00 ± 0.02 vs 1.20 ± 0.04 vs 1.21 ± 0.06 and entropy (h): 2.53 ± 0.01 vs 2.56 ± 0.06 vs 2.57 ± 0.09.

3. Discussion

We would like to emphasize that our MRI findings in the present investigation may not be generalizable to other animals and humans, since we only used one strain (Wistar) of mid-aged (15 months old) male rats for the study of normal brain aging, and induced diabetes via administration of streptozotocin-nicotinamide in rats with 1.5 months standing. Another limitation of the present study is the ROI selection for quantitative MRI measurements with a 7 Tesla system. Corpus callosum is a major representative anatomical white matter structure in rat cerebrum. However, limited by the MRI spatial resolution, the partial volume effect cannot be avoided when selecting the corpus callosum ROI in rat brain from MRI images. Cortex mostly contains cell bodies, but still contains minor axonal fiber content. Thus cortex consists only of relatively pure gray matter, with a component of white matter. Most other cerebral regions are mixture of white and gray matter, MRI changes of these regions due to aging and diabetes may be expressed as a combination of changes of corpus callosum and cortex.

The process of normal cerebral aging is highly individual-dependent, complex and heterogeneous (Gunning-Dixon et al., 2009), and is characterized by structural and morphological brain changes with a trend for a slight decrease in brain weight and volume (brain atrophy), decline in cognitive performance and modifications in electro-physiological properties (Wrighten et al., 2009). Thus, aging brain with or without diabetes may not be characterized by any single index. In the current study, multiple MRI measurements were simultaneously employed to characterize brain changes due to aging or diabetes in rats. Our results demonstrated that metrics of T₁, T₂, relCBF, MK and entropy measured in the corpus callosum failed to discriminate between the normal aged brain and diabetic aged brain because of significant differences of these measured MRI values either between the young and mid-aged healthy rats, or between the mid-aged healthy and diabetic rats, as well as between the healthy young adult and non-long-standing diabetic mid-aged rats (Fig. 3a, b, c, g and h). Cerebrovascular permeability specified to Gd-DTPA by CE-T1WI was able to distinguish between the young and mid-aged brains and not between the mid-aged healthy and non-long-standing diabetic brains (Fig. 3d), and MD/FA only distinguished between the mid-aged healthy and non-long-standing diabetic brains (Fig. 3e and f). Thus, the combination of CE-T1WI using Gd-DTPA and FA (or MD) measurements were able to white matter changes between the age and diabetes factors. For cortex, metrics of relCBF, MD and entropy failed to detect any changes of gray matter in mid-aged healthy and non-long-standing diabetic rat brains (Fig. 4c, e and h); T₁, CE-T1WI, FA and MK discriminated changes between the healthy young and mid-aged brains rather than between the mid-aged healthy and non-long-standing diabetic brains (Fig. 3a, d, f and g), and only T₂ can discriminate changes between the healthy and diabetic brains rather than between the healthy young and mid-aged brains (Fig. 3b). Therefore, combination of T₂ and T₁ (or CE-T1WI/FA/MK) measurements appear to distinguish between age and diabetes effects within gray matter.

Consideration of the relative ease of acquiring MRI data and the limitations of image contrast (e.g., Gd-DTPA) injection in patients, using T₁ and T₂ combination appears superior to the CE-T1WI and FA combination to identify distinguishing characteristics between cerebral changes attributed to aging and diabetes factors. In addition, analysis of the current results across the white and gray matters, indicates that FA interestingly exhibited significant changes that appeared specifically dependent on diabetes for white matter, and dependent only on aging for gray matter in mid-age brain. Based on the different natures of FA in white and gray matter, a combination of FA measurements in both white and gray matters may also discriminate between aging and non-long-standing diabetes factors that evoke cerebral changes during mid-age.

Prior studies demonstrated that changes of MRI indices in aging or diabetic cerebral tissue were strongly dependent on stage of age (infant, adolescent, adult or older) (Agartz et al., 1991; Badve et al., 2015; Breger et al., 1991; Callaghan et al., 2014; Cho et al., 1997; Ding et al., 2004; Hedman et al., 2012; Saito et al., 2009), region of brain tissue (Chen et al., 2011; Hsu et al., 2012; Kumar et al., 2012; Raz et al., 2005; Steen et al., 1995) and duration of hyperglycemia (Banks et al., 1997; Brundel et al., 2014; Last et al., 2007). Here, we perform measurements of multiple MRI indices of corpus callosum, a white matter region of brain, and of cortex, a mostly gray matter enriched brain region in healthy young adult (2–3 mo), healthy mid-age (14–15 mo) and mid-age (14–15 mo) with approximate 2 months T2DM induced by streptozotocin-nicotinamide administration in male Wistar rats. Our study is the first to systematically measure multiple MRI metrics to examine the effects of aging and non-long-standing diabetes on brains, specific to white matter and gray matter in rats.

Relaxation time constants T₁ and T₂ are two basic MRI parameters, which can be readily obtained using conventional MRI system and sequences. It has been known for a long time that changes of relaxation times are associated with changes of structure and metabolism in brain, which suggests that T₁ and T₂ should reflect changes in aging or diabetic brain. For instance, brain structure differs in adult T2DM patients with global brain atrophy compared to age-matched people without diabetes (Brundel et al., 2014; Last et al., 2007), or from increased hippocampal volumes from a large cohort of children with type 1 diabetes (Seaquist, 2010). The global brain atrophy with age compresses the space of free water which reduces the ratio of free water versus restricted water and results in the reduction of T₂ value (Peters, 2006), since T₂ value of brain tissue is governed mainly by exchange diffusion of water between the bound layer and a free water phase. In addition to accelerated brain atrophy induced by diabetes (Last et al., 2007), elevated levels of iron have been found in blood plasma in diabetics (Thomas et al., 2004). Both factors may reduce the T₂ value of brain tissue. Thus, it is interesting that T₂ can be employed to detect diabetic changes in brain tissue. However, the association of T₁ or T₂ values of brain with age or diabetes is not completely determined. Existing data from infant (3 weeks) to mid-age adult (39 years), show that T₂ values of human brain continuously decrease with increasing age (Ding et al., 2004). However, except for a few regions, T₂ values typically increased with age in multiple human brain sites from mid-age adult (31 years) to elderly healthy subjects (66 years) (Kumar et al., 2012). Type 1 diabetes accelerated T₂ reduction observed in the young patients (20.3 ± 4.2 years) (Pell et al., 2012). In mice, a certain degree of aging-associated decrease in T₂ values is seen in the cortex and hippocampus (Falangola et al., 2007), like in human. However, the temporal profiles of T₁ values of human brain white matter are similar to T₂ values, both sharing a cradle-shaped relation to age with minimal values at age around 40 years (Agartz et al., 1991; Yeatman et al., 2014). Unfortunately, the existing clinical data are not always consistent in MRI measurements of T₁ and T₂ (Agartz et al., 1991; Badve et al., 2015; Breger et al., 1991; Cho et al., 1997; Ding et al., 2004; Steen et al., 1995).

In the present animal study, both T₁ and T₂ values of white matter significantly decreased at mid-age compared to young adult (Fig. 3a and b). T2DM significantly exaggerated the reduction of white matter T₁ and T₂, compared to age-matched non T2DM rats. The results suggest that diabetes accelerates the structural changes of white matter during brain aging at least to middle age. The T₁ and T₂ properties of gray matter were different (Fig. 4a and b), where aging and diabetes dominated changes of T₁ and T₂, respectively. In gray matter, compared with non T2DM rats, non-long-standing diabetes did not significantly lower the T₁ values, but did reduce the T₂ values; conversely, aging significantly lowered the T₁ values while it did not reduce the T₂ values, compared the healthy young rats with the healthy mid-aged rats. This difference between T₁ and T₂ values of gray matter in response to diabetes and aging indicates that aging and diabetes may differentially affect the structure of cerebral tissues. Our data suggest that we can discriminate between healthy and non-long-standing diabetic brains in mid-aged rat by measuring cortical T₁ and T₂ values: if only the T₁ value is reduced, this change would represent normal aging brain; if both cortical T₁ and T₂ values are reduced, this indicates an early stage diabetic aging brain.

Diabetes mellitus is associated with vascular pathology, since diabetes affects the circulatory system. Vascular disease progresses with aging, and blood-to-brain barrier (BBB) permeability increases with normal aging (Farrall and Wardlaw, 2009). Diabetes increases BBB permeability via a loss of tight junction proteins (Farrall and Wardlaw, 2009; Hawkins et al., 2007), and increased BBB permeability of white matter was detected using MRI (CE-T1WI with Gd-DTPA) in aged patients (65–70 years) with T2DM (Starr et al., 2003). CE-T1WI using Gd-DTPA readily permits measurement of BBB permeability in rat brain, which is significantly correlated with the quantitative MRI permeability parameter of blood-to-brain transfer rate constant (K_i) of Gd-DTPA using Look-Locker sequence for T₁ measurement (Ding et al., 2016; Ewing et al., 2006). Our data demonstrated that normal aging increased BBB permeability for small size molecules (e.g., Gd-DTPA) in both white and gray matter, and T2DM present for less than two-months at mid-age did not exaggerate BBB permeability for the small size molecules in white and gray matter (Fig. 3d and 4d). Our data appear consistent with a prior study on BBB permeability in T2DM mid-aged animals subjected to stroke (Zhang et al., 2016). Immunohistochemical analysis showed that density of fibrin/fibrinogen (—340 kDa molecular weight) immunoreactive vessels (6.4 ± 1.9/mm²), including cortex and corpus callosum in the contralateral hemisphere, was not altered compared to the non-T2DM rats (4.9 ± 1.4/mm²), indicating no increase in BBB permeability (p > 0.05) (Zhang et al., 2016). However, in the present study T2DM did increase BBB permeability for small size molecules (Gd-DTPA: 0.94 kDa molecular weight) in the hippocampus and thalamus (data not shown), consistent with a prior report of fibrin leakage in hippocampus (Zhang et al., 2016). Thus, the MRI measures of BBB permeability of white and gray matter may not be a good biomarker for monitoring the influence of non-long-standing diabetes at mid-age, while the BBB permeability of hippocampus appears sensitive to diabetes.

In addition to BBB permeability, relCBF or CBF is an important index of vascular function. CBF in most cortical and subcortical regions is not significantly reduced between the healthy mid-aged and healthy young adults (Chen et al., 2011). However, CBF is reduced in the frontal region in elderly patients with diabetes (Last et al., 2007). Our animal results are consistent with those in humans (Chen et al., 2011; Last et al., 2007), as noted above. Neither mid-aged healthy nor early stage diabetic rats exhibited hypoperfusion in gray matter compared to healthy young adult or healthy mid-aged rats (Fig. 4c), respectively. However, compared to healthy young adult, significant hypoperfusion in white matter was detected in healthy mid-aged rats, and blood perfusion of corpus callosum further significantly decreased in the non-long-standing diabetic mid-aged rats (Fig. 3c). Therefore, the present MRI cerebrovascular results primarily indicate that both aging and diabetes have effects on cerebral hypoperfusion, which influence blood perfusion more significantly in the white matter than in gray matter.

FA and MD derived from diffusion tensor imaging (DTI) scans have been widely employed in MRI diffusion studies of the aging brain (Falangola et al., 2008; Hsu et al., 2012). Except for the elderly, no significant differences were detected by FA and MD from adolescence to young adult (26–47 years old, including middle age) healthy people (Falangola et al., 2008). For elderly or long-standing T2DM patients compared with healthy control subjects, significant decrease of FA and increase of MD were generally observed in white matter and/or gray matter (Frøkjӕr et al., 2013; Hsu et al., 2012; Reijmer et al., 2013). However, clinical results derived from DTI are not completely consistent (Chun et al., 2000; Falangola et al., 2008; Gong et al., 2014; Hsu et al., 2008; Salat et al., 2005; Westlye et al., 2010). Our present results that demonstrated that no changes of FA and MD in white matter were acquired in young adult to middle aged healthy rats (Fig. 3e and f), and non-long-standing diabetes significantly decreased MD and increased FA of white matter (Fig. 3e and f) in the middle aged rats. For gray matter, both middle aging and non-long-standing diabetes did not influence MD values (Fig. 4e); however, from young adult to mid-age, aging significantly decreased FA, while the approximate two-month standing T2DM rats exhibited no effects on FA (Fig. 4f). These new findings call for further investigations to consolidate results derived from DTI.

Based on our FA results in the current study, we may be able to discriminate between healthy and non-long-standing diabetic brains in mid-aged rats by measuring FAs of white and gray matter. Normal aging from youth to mid-aged would be indicated if the FA value of gray matter decreases without an increase in the FA value of white matter. If the FA value of gray matter decreases with an elevated FA value of white matter, this would be an index of the non-long-standing diabetic aging brain.

MK and entropy measures have expanded MRI diffusion indices from FA and MD to quantitative measures of the non-Gaussian diffusion process and anisotropy (Fieremans et al., 2011), which are more suitable for brain tissue (Ding et al., 2017). Age dependence was found for MK in several anatomical regions, including white and gray matter, which increased from adolescence to adulthood in healthy volunteers (Das et al., 2016; Gong et al., 2014). Few investigations, however, employ MK analysis of the diabetic brain (Xie et al., 2017; Xiong et al., 2019). Our results demonstrated that MK and entropy of white matter in rat brain significantly increased with aging, and further increased with T2DM (Fig. 3g and h). In gray matter region, significant differences of MK were found with aging, rather than early stage diabetes (Fig. 4g), while no differences of entropy were observed among all groups of rats (Fig. 4h). These animal results of MK are consistent with measures of MK in human (Das et al., 2016; Gong et al., 2014). Based on our data, MK appears superior to MD in investigating brain changes affected by aging with or without diabetes. Entropy seems sensitive to white matter, which detected differences caused by either aging or diabetes (Fig. 3h). However no changes were evident with aging and non-long-standing diabetes in gray matter (Fig. 4h). Since white matter is composed of bundles of directionally organized neuronal fibers, the present results of entropy appear reasonable.

In summary, we present multiple MRI indices of white and gray matter in brain from healthy young adult, healthy mid-aged and non-long-standing diabetic mid-aged rats. These data provide insight into differences between healthy and/or early stage diabetic white and gray matter in young and mid-age rat. Our data indicate that we may distinguish between the healthy and non-long-standing diabetes in mid-aged rat brains by measuring: 1) T₁ and T₂ of gray matter, where normal aging only reduces T₁ value (Breger et al., 1991; Cho et al., 1997; Kumar et al., 2012; Steen et al., 1995) and non-long-standing diabetic aging reduces both T₁ and T₂ values (Pell et al., 2012); 2) FAs of white and gray matters, where normal aging only decreases FA value of gray matter (Falangola et al., 2008) while non-long-standing diabetic aging not only decreases FA value of gray matter (Rofey et al., 2015) but also elevates FA value of white matter (Yau et al., 2014; Yeatman et al., 2014). Since FA, T₁ and T₂ are conventional MRI measurements performed in the clinic, this method of discriminating age-matched healthy and early stage diabetic brain may be readily applied in the clinic trial and in human research. Once we can distinguish changes of cerebral tissue due to the normal aging from diabetic aging, we may be able to better understand how diabetes accelerates aging in normal brain.

4. Experimental procedure

All experimental procedures were conducted in accordance with the National Institutes of Health (NIH) Guide for the Care and Use of Laboratory Animals and were approved by the Institutional Animal Care and Use Committee of Henry Ford Health System.

4.1. Animal model and groups

In the present study, all animals were male Wistar rats (Charles River, Wilmington, MA). T2DM was induced in rats at 13 months of age by a single intraperitoneal injection of both 210 mg/kg of nicotinamide (NA, Sigma Chemical Co., St. Louis, MO), a water-soluble vitamin with anti-inflammatory actions, and 60 mg/kg of streptozotocin (STZ, Zanosar®, Sigma Chemical Co., St. Louis, MO), a naturally occurring chemical that is particularly toxic to the insulin-producing beta cells of the pancreas in mammals. This rat model of mid-aged T2DM has been demonstrated to produce non-insulin dependent diabetic syndromes that resemble human T2DM (Masiello et al., 1998; Su et al., 2006), and this model mimics pathological manifestations of human T2DM, including increase in blood glucose, insulin secretory dysfunction, and glucose intolerance, but not hyperinsulinemia and obesity (Kahn, 2003; Szkudelski, 2012). At 2 weeks after STZ-NA administration, the diabetic state of rats was assessed by measurements of non-fasting plasma glucose concentration using test strips for glucose (Polymer Technology System, Indianapolis, IN). Rats that developed hyperglycemia with non-fasting plasma glucose concentrations > 250 mg/dL are considered T2DM, and were selected for the study (n = 10). MRI scans were performed on these rats (non-long-standing diabetic mid-aged rats) at an average of 1.5 months after STZ-NA administration (approximately 1 month after confirmation of T2DM). Another group of healthy mid-aged rats without T2DM (n = 10, 14–15 months) were age-matched with the diabetic rats (Szkudelski, 2012). Together with the above two mid-aged groups of rats with or without diabetes, a third group (n = 10) of healthy young adult (250–300 g, 2–3 months) rats were used in the MRI experiments, as an age control group.

4.2. MRI measurements

MRI scans were performed using a 7 T system (ClinScan, Bruker-Biospin, Ettlingen, Germany). A birdcage transmitter and a quardrature half-volume receiver were used in the MRI scans for all rats. During MRI scans, rats were anesthesized using medical air (1.0 L/min) with isoflurane (1.0–1.5%). Stereotactic ear bars were used to minimize movement, and rectal temperature was maintained at 37.0 ± 0.5 °C using a feedback controlled water bath (YSI Inc, Yellow Springs, OH). At the beginning of each MRI scan, a fast gradient echo imaging sequence was used as scout for reproducible positioning of all animals in the magnet. Various MRI sequences were scanned and multiple MRI parametric maps were obtained.

Three dimensional variable flip angle spoiled gradient recalled echo sequences were used for in vivo T₁ mapping (Trzasko et al., 2013). A matrix 128×64×16 fit for a 32×32×16mm³ field-of-view (FOV); with time of repetition (TR) and time of echo (TE) were 30 ms and 3.8 ms, respectively; six different flip angles (2°, 5°, 10°, 15°, 20°, 25°) were used with the bandwidth of 200 Hz per pixel.

T₂-weighted imaging (T2WI) was acquired using a multi-slice (13 slices) and multi-echo (6 echoes) sequences, with TE as 15 ms, and equal interval to 90 ms, TR of 4.5sec, a 32×32mm² FOV, 1 mm slice thickness and 128×96 matrix.

CE-T1WI consisting of two T₁ weighted imaging acquisitions were performed before and 6 min after injection of Gd-DTPA, into a tail vein at a dose of 0.4 mL/kg. T1WI was acquired using a conventional multi-slice single spin-echo sequence with TE of 8 ms, a TR of 500 ms and with the same other parameters as for T2WI.

CBF was estimated using perfusion MRI by employing a pulsed arterial spin labeling (PASL) technique, a second version quantitative imaging of perfusion using a single subtraction, with thin-slice periodic saturation modified from a proximal inversion with a control for off-resonance effects labeling scheme (Luh et al., 1999; Wong et al., 1998). Single-shot of echo-planar readout was used for each 2 mm-thickness slice for a total 5 slices perpendicular to the magnetic field direction with a 0.5 mm off-center to match every two slices of T2WI. Complementary PASL imaging parameters were: FOV 32×32 mm², matrix 128×128, flip angle 90°, frames 201, TR 3000 ms, TE 18 ms, start of label saturation time 500 ms, end of label saturation time 800 ms; inversion time 1000 ms, pixel bandwidth 2605 Hz. Bipolar crusher gradients were used with a cutoff speed of 60 cm/s. The 80 mm labeling slab was positioned with a 6.1 mm gap caudal to the imaging slab.

Diffusional measures included q-ball DTI and diffusion kurtosis imaging (DKI) (Jensen et al., 2005; Tuch et al., 2003), using a pulsed gradient (duration δ = 10 ms; interval Δ = 18 ms) spin-echo sequence and a one-shot echo-planar readout with a baseline of b = 0 s/mm². The geometric parameters of diffusion MRI were matched to T2WI, with repetition time of 10 s and echo time of 50 ms. Additional MRI parameters were 64 directions of diffusion gradients with b = 1500 s/mm² in q-ball acquisition; as for DKI, 20 directions of diffusion gradients with b = 900 and 1800 s/mm², respectively, and 2 averages were employed.

4.3. Image processing

Some images and maps, such as T1WI images for CE-T1WI, T₂ and relCBF maps, were immediately generated by the MRI system after data acquisition and then directly converted to planar resolution of 128×128 matrix which was employed for data measurement and analysis. The CE-T1WI map was produced by subtracting the pre contrast T1WI image from the corresponding post contrast T1WI image (Ding et al., 2014), performed using homemade software, Eigentool (Ding et al., 2008). All T₁ estimations were executed by Matlab (Mathworks, Natick, MA, USA). T₁ fitting was performed to produce T₁ maps via nonlinear least squares regression (Wang et al., 1987). Final T₁ maps are equivalent to 128×128 matrix in plane of 16 slices with a 1 mm slice thickness.

The diffusion MRI data analysis was first performed using Camino software (Alexander, 2005), and then Eigentool (Peck et al., 1992) was employed after format transformation by using homemade software developed under Matlab (Ding et al., 2017). Using the linear least-square fitting, a diffusive ellipsoid for each pixel was calculated from the q-ball DTI data. The apparent diffusion coefficient of water along the longest ellipsoidal principal axis (eigenvalue λ₁) was noted as the axial diffusivity; whereas the perpendicular or radial diffusivity was the mean of the other two eigenvalues (λ₂ and λ₃) (Alexander et al., 2007). The maps of MD and FA of water diffusion were then calculated from these three eigenvalues (Alexander et al., 2007). The entropy was calculated pixel-by-pixel using the probability of the MRI signal attenuation value being repeated due to water displacement along a certain spatial orientation for each pixel (Ding et al., 2017). With a two compartment model of the intra- and extracellular space used for cerebral tissue, MK maps were deduced based on the DKI data using the biexponential fitting (Fieremans et al., 2011).

All metric maps other than the T₂ map (i.e. T₁, relCBF, CE-T1WI, FA, entropy, MD and MK) were then co-registered with the T₂ map slice by slice and were finally transformed into the same image format for our software, Eigentool. Since the geometric parameters of all pulse sequences in the experiment were set to match the T2WI, only scan-to-scan misregistration caused by head movement required correction by rigid-body alignment of each scan to the T₂ map (Ding et al., 2008; Nestares and Heeger, 2000). Distortion in echo planar readout maps were corrected by registration.

4.4. Data and statistical analysis

Existing MRI segmentation results of rat brain, as shown in Fig. 5a, demonstrated that cortex and corpus callosum consist of relatively pure gray matter and white matter, respectively (Valdes-Hernandez et al., 2011). In the present study, hence, the MRI measurements from either cortex or corpus callosum were accordingly referred to as metrics of either gray matter or white matter, respectively. According to T₂-weighted MRI images, as indicated in Fig. 5b, with the complementary colored FA map (Fig. 5c), regions-of-interest (ROIs) of gray matter and white matter were outlined from cortex and corpus callosum, as shown in Fig. 5d, where the cingulum was withdrawn from the white matter ROI. Thus, MRI measurements of such gray and white matter ROIs were performed for all rats in this study.

Fig. 5.

Regions-of-interest of MRI measurements in the present investigation. Existing MRI segmentation results of rat brain (a), demonstrated that cortex and corpus callosum consist of relatively pure gray matter and white matter, respectively. According to T₂-weighted image (b), with the complementary colored FA map (c), regions-of-interest of gray matter and white matter were outlined from cortex and corpus callosum (d).

The relCBF values were measured for white matter and gray matter, since the MRI system directly provided the relCBF map immediately after acquisition scans. The conversion from relCBF to CBF (mL/100 g/min) depends on sequence parameters, and biological parameters related to features of cerebral tissues. With the blood water partition coefficient as 0.9 mL/g in the cerebral tissue, the inversion efficiency of equilibrium magnetization is 0.95 using the current pulse sequence, T₁ of the water in arterial blood is 2200 ms, and the conversion coefficient is 0.24 using current sequence parameters.

Data analysis was performed in a blinded fashion. Measurements are summarized as mean and standard deviation (SD). Analysis of variance was performed at first for all three groups. Student’s t-test was then applied to pairs of animal groups of every MRI measurement for individual statistical differences. The effect was detected at the 0.05 level.

HIGHLIGHTS.

• Multiple MRI measurements of brain for young, mid-aged and diabetic rats.

• Aging only reduces T₁ value, diabetes reduces T₁ and T₂ values, of gray matter.

• Aging only decreases gray matter FA, besides diabetes elevates white matter FA.

• These findings may be readily applied in the clinic trial and in human research.