Diffusion tensor imaging with direct cytopathological validation: characterisation of decorin treatment in experimental juvenile communicating hydrocephalus

Three-week-old Sprague–Dawley rats (Charles River, Massachusetts, USA) were housed in litters in individual cages, kept under a 12 h light/dark cycle with free access to food and water. Animals were monitored for adverse effects of treatments, such as distress, lethargy, weight loss and seizures, and any animals showing severe adverse effects were euthanised. Experiments were conducted at the University of Utah in accordance with the guidelines of the National Institutes of Health Care and Use of Laboratory Animals and approved by the University of Utah Ethics Committee.

The experimental design and surgical techniques are described in detail elsewhere [12]. Using a ventral approach, the interval between the occipital bone and the C-1 vertebral body was exposed and a 30 gauge angled needle was inserted into the prepontine (basal cistern) subarachnoid space. 30 µl of 20 % kaolin solution (200 mg/ml in 0.9 % sterile saline; Fisher Scientific, Massachusetts, USA) was injected to induce communicating hydrocephalus and the rat was either allowed to recover or underwent osmotic pump and intraventricular cannula implantations. The cannulae were inserted into the right lateral ventricle and fixed in place with glue and bone cement (Biomet UK Ltd, Bridgend, UK) to a stabilising screw, and connected to subcutaneously implanted mini osmotic pumps. Osmotic pumps (model 2002 adapted for use in MRI scanners with PEEK tubing, Alzet, Durect Corporation, California, USA) were filled with either 5 mg/ml human recombinant decorin (Galacorin^TM, Catalent/Pharma Solutions, New Jersey, USA) or 10 mM phosphate buffered saline (PBS) pH 7.4 (Sigma-Aldrich, Missouri, USA). Over the subsequent 14 days, human recombinant decorin was infused at a rate of 2.5 mg/0.5 ml/h.

Rats were randomly assigned to four groups: (1) Intact age-matched controls (Intact group, n = 4); (2) basal cistern kaolin injections only (kaolin group, n = 4); (3) kaolin injection with intraventricular infusion of PBS (kaolin + PBS group; n = 6); and (4) kaolin injection with intraventricular infusion of decorin (kaolin + decorin group; n = 5). Magnetic resonance imaging (MRI) and diffusion tensor imaging (DTI) were conducted after 14 days of treatment to assess the extent of hydrocephalus before sacrifice, then the brains were removed and processed for histology.

Imaging experiments were conducted 14 days post injury using a 7-Tesla horizontal-bore Bruker Biospec MRI scanner (Bruker Biospin, Ettlingen, MA, USA) interfaced with a 12-cm actively shielded gradient insert capable of producing magnetic field gradient up to 600 mT/m. Animals were anesthetised using 1–3 % Isoflurane and 0.8 L/min O₂ and their vital signs (respiration, temperature, heart rate and oxygen saturation percentage) were continuously monitored using a MR-compatible physiological monitoring system (SA Instruments, Stony Brook, NY, USA). Animals were placed in a 72-mm volume coil for signal transmission, and a quadrature surface coil was placed on the head for signal reception. Acquisition of T₂-weighted MRI scans and ventricular volume analysis has been described previously [12]. DTI scans were conducted using spin echo diffusion-weighted sequences with single-shot EPI readout, with the following parameters (TR of 3760 ms, TE of 44 ms, 15 coronal 1 mm-thick slices, a field of view of 2.5 × 2.5 cm, and an in-plane resolution of 195 × 195 µm). Thirty uniformly-spaced over unit sphere diffusion-weighted gradient directions and five non-weighted images were acquired with two signal averages and the following diffusion parameters: diffusion gradient duration 7 ms, separation 20 ms, diffusion encoding sensitivity 700 s/mm². Scan time was 4 min. For ventriculomegaly analysis, one MRI scan image was chosen from the rostral cerebrum (−0.36 mm from bregma) and the caudal cerebrum (−3.72 mm from bregma) for each rat, and the ventricular area was determined in each scan using ImageJ.

Prior to commencing voxel-based analysis, double blinding was introduced to prevent group identification. Using the software, DSI Studio (DSI Studio, Pittsburgh, PA), DTI images were reconstructed and processed to produce voxel based maps, from which regions of interest (ROIs) could be analysed. The seven ROIs selected include: corpus callosum, periventricular white matter, caudal and rostral internal capsule, outer parietal cortex, inner parietal cortex and CA1 hippocampus. DTI parameter values from four serial sections (1.28 mm anterior to Bregma to 3.72 mm posterior to Bregma) of the corpus callosum and periventricular white matter were analysed in a total of 17 rats [Intact (n = 4), kaolin (n = 4), kaolin + PBS (n = 5), kaolin + decorin (n = 4)]. The rostral corpus callosum and periventricular white matter sections were defined as 1.28 and −0.36 mm from Bregma. The caudal corpus callosum and periventricular white matter sections were derived from −2.76 to −3.72 mm from Bregma (Fig. 2a). The remaining five ROIs were analysed in 16 rats, with four animals being examined in each experimental group. Three sections were independently analysed for the CA1, caudal internal capsule, outer and inner parietal cortex from 0.36 to 3.72 mm posterior to Bregma. An average of two sections from 1.28 mm anterior to Bregma to 0.36 mm posterior to Bregma were individually analysed for the rostral internal capsule. ROIs were identified using a FA voxel based map and an analogous DTI image (Fig. 1). The mean FA, MD, AD and RD values were calculated for each ROI of each animal.

Rats were euthanised and immediately perfused transcardially with PBS followed by 4 % paraformaldehyde (Alfa Aesar, Ward Hill, MA, USA) in PBS. Brains were immersed in 4 % paraformaldehyde overnight at 4 °C, cryoprotected by sequential immersion in 10, 20 and 30 % sucrose solutions in PBS at 4 °C and embedded in optimum cutting temperature embedding matrix (Fisher Scientific). Subsequent sectioning and staining of the tissue was conducted at the University if Birmingham. Coronal sections 15-µm thick were cut on a Bright cryostat (Bright Instrument, Huntingdon, UK), serially mounted and stored at −20 °C before staining.

Myelin integrity was assessed with an antibody against myelin basic protein (MBP; rat, Merck Millipore, Watford, UK, MAB386). Antibodies against glial fibrillary acidic protein (GFAP; mouse, Sigma-Aldrich, G3893) and OX-42 (CD-11b; mouse, Serotec, Kidlington, UK, MCA527R) were used to assess gliosis and the extent of oedema resolution was examined by aquaporin-4 (AQP4) antibody staining (chicken, Genway, San Diego, CA, USA, 07GA0175-070718).

Immunohistochemistry was conducted on the caudal cerebrum of 19 rats [Intact (n = 4), kaolin (n = 4), kaolin + PBS (n = 6), kaolin + decorin (n = 5)]. All selected sections were at least −2.5 mm posterior to Bregma and corresponded with the location of the DTI sections. Sections were washed in PBST (10 mM PBS pH 7.4 containing 0.3 % Tween20) and blocked in 2 % bovine serum albumin (BSA) and 15 % normal goat serum in PBST at room temperature for 1 h. Subsequently, sections were washed in PBST, before being incubated at 4 °C overnight in primary antibody diluting buffer containing PBST and 2 % BSA. After washing in PBST the sections were incubated for 1 h in secondary antibody solution (Alexa Fluor^® 488 or 594 labelled secondary antibodies (Life Technologies, Paisley, UK) in PBST with 2 % BSA and 1.5 % normal goat serum) at room temperature, in the dark. After further PBST washes, sections were mounted in Vectashield containing DAPI (Vector Laboratories, Peterborough, UK). The Zeiss Axioplan 2 imaging epifluorescent microscope (Carl Zeiss, Germany) and the AxioCam Hrc (Carl Zeiss, Jena, Germany) were used to view and capture images under the same conditions for each antibody at ×400 magnification.

Quantitative analysis was undertaken using the software, Image J and all analyses were undertaken with the operator masked to the experimental group. Images for each immunofluorescent stain were processed identically before being analysed. For each image analysed, four randomly placed regions of interest (ROIs) were drawn with each ROI being 2.96 mm wide and 1.57 mm in height. For the corpus callosum, periventricular white matter and CA1 and CA3 hippocampal regions, a mean of 16 ROIs (four regions of interest × four coronal sections) were chosen per rat per stain. An average of 8 ROIs (four regions × two coronal sections) were selected for the internal capsule, caudate-putamen, parietal cortex and occipital cortex. All areas were analysed for GFAP, OX-42 and AQP4 staining however, as MBP is a marker of white matter integrity, only the corpus callosum, periventricular white matter and internal capsule were analysed for this antibody.

GFAP and OX-42 image processing included the conversion of images into a gray scale format prior to spatial filtering, thresholding and despeckling of the images using Image J. Images stained for AQP4 and MBP were identically processed to the GFAP and OX-42 images except thresholding was not performed. The mean percentage area of GFAP, OX-42, AQP4 and MBP positive staining, for each experimental group was calculated.

In order to assess hippocampal size, one cerebral section, at least 2.5 mm posterior to Bregma from each experimental animal was examined at ×10 magnification using the Nikon SM21500 dissecting microscope (Nikon, Tokyo, Japan). Images were captured with a Nikon ds-2mv high-resolution camera (Nikon). Hippocampal area was assessed by using the Image J software analyze area tool.

Statistical analysis was conducted using SPSS software, version 22 (IBM, Armonk, NY). Normally distributed data were analysed using a one-way ANOVA followed by a post hoc Tukey test. In the absence of normality, data were analysed using the Kruskal–Wallis test and tested for significant pairwise comparisons. Normally distributed data were expressed as the mean ± standard error of the mean (SEM). Correlation analysis was performed using a two-tailed Spearman’s correlation test. As immunohistochemistry analysis was performed on caudal sections, mean DTI data from Section 2.76 and 3.72 mm posterior to Bregma were used for correlation analysis. Correlation analysis was not undertaken for the inner parietal cortex, caudate-putamen, occipital cortex or CA3 region of the hippocampus because DTI analysis was not being performed in these areas. Values were considered statistically significant when p values were *p < 0.05, **p < 0.01, ***p < 0.001 and ****p < 0.0001.

In the caudal periventricular white matter (−2.76 and −3.72 mm from Bregma), compared to Intact controls (0.83 ± 0.00 and 1.20 ± 0.02, respectively; Fig. 2b), the kaolin and kaolin + PBS groups displayed a significant increase in the MD (1.48 ± 0.22, p = 0.023 and 1.56 ± 0.15, p = 0.031 respectively) and AD (2.04 ± 0.27, p = 0.020 and 2.16 ± 0.17, p = 0.018, respectively). By contrast, the MD of decorin treated rats (0.85 ± 0.00) was significantly lower in comparison to kaolin (1.48 ± 0.22, p = 0.032) and kaolin + PBS (1.56 ± 0.15, p = 0.044) rats. No significant differences were observed in any DTI parameters in the rostral periventricular white matter (1.28 and −0.36 mm from Bregma) between the four experimental groups.

In the caudal corpus callosum (−2.76 and −3.72 mm from Bregma), the AD for kaolin (1.80 ± 0.15, p = 0.010) and kaolin + PBS (1.92 ± 0.12, p = 0.001) groups were significantly higher than Intact controls (1.56 ± 0.05; Fig. 2c). Moreover, kaolin + PBS rats displayed a significant elevation in the MD (1.17 ± 0.08, p = 0.030) and RD (1.00 ± 0.06, p = 0.025), compared to Intact animals (0.88 ± 0.02 and 0.54 ± 0.02, respectively). Furthermore, decorin treatment significantly reduced the AD (1.49 ± 0.02, p = 0.004) and RD (0.62 ± 0.03, p = 0.034) compared to the kaolin + PBS animals (1.92 ± 0.12 and 1.00 ± 0.06, respectively) in the caudal corpus callosum. No significant differences existed between decorin treated rats and Intact controls for all DTI parameters in the caudal corpus callosum. Similar to the rostral periventricular white matter, in the rostral corpus callosum (1.28 and −0.36 mm from Bregma), there were no significant differences in the DTI parameters between the experimental groups.

Alongside the corpus callosum and the periventricular white matter, five other regions of interest were examined by DTI voxel based analysis (Additional file 2: Figure S2). Significant differences in the four DTI parameters were not observed in the outer or inner parietal cortex (Additional file 2: Figure S2a, b). However, the FA of the kaolin group was significantly lower (0.11 ± 0.02, p = 0.036) than Intact controls (0.16 ± 0.01) in the CA1 region of the hippocampus (Additional file 2: Figure S2c). Moreover, in the caudal internal capsule (Additional file 2: Figure S2d), decorin (0.83 ± 0.01, p = 0.003) reduced the decrease in the MD observed in kaolin + PBS animals (0.77 ± 0.00). In the rostral internal capsule (Additional file 2: Figure S2e), kaolin + PBS rats displayed significantly greater FA (0.16 ± 0.01, p = 0.022) and AD (1.70 ± 0.28, p = 0.039) compared to Intact controls (0.23 ± 0.00 and 0.68 ± 0.23, respectively).

As DTI parameter abnormalities were predominantly observed in the caudal cerebrum, we investigated whether non-uniform ventriculomegaly occurs in the basal cistern model of communicating hydrocephalus. In the kaolin and kaolin + PBS groups, the ventricles expanded significantly (p < 0.05) rostrally (8.04 ± 1.74 and 10.12 ± 3.83 mm³, respectively) and caudally (16.22 ± 2.76 and 21.00 ± 5.43 mm³, respectively) compared to the Intact controls (rostral = 1.26 ± 0.11 and caudal = 0.93 ± 0.10 mm³). Furthermore, significant changes in the mean differences between the rostral versus caudal ventricular volume were discovered (Table 1).

Rostrally, ventricular volume significantly correlated with the FA, MD, AD and RD measurements of the corpus callosum and periventricular white matter (Table 2). Likewise, the caudal ventricular volume correlated with all DTI parameter measures in the corpus callosum and all except the AD in the periventricular white matter (Table 2).

As DTI parameter abnormalities were observed in the caudal periventricular white matter, corresponding immunohistochemistry analysis was conducted to aid hydrocephalic cytopathology characterisation of these tissues. We determined that the levels of GFAP positive immunostaining (Fig. 3a) were significantly increased in kaolin rats (2.49 ± 0.23 %, p = 0.048) compared to Intact controls (0.82 ± 0.11 %) indicating the presence of astrogliosis in the caudal periventricular white matter. Furthermore, the GFAP positive astrocytes of hydrocephalic animals exhibited features typical of reactive astrocytic morphology; cytoplasmic processes underwent thickening in kaolin and kaolin + PBS rats. The increase in GFAP positive staining observed in the kaolin rats (2.49 ± 0.23 %) was prevented with decorin treatment (0.49 ± 0.11 %, p = 0.002). Additionally, kaolin (1.77 ± 0.14 %, p = 0.040) and kaolin + PBS rats (1.34 ± 0.29 %, p = 0.056) displayed a significant and non-significant increase respectively in periventricular white matter AQP4 positive immunostaining (Fig. 3b) compared to Intact controls (0.80 ± 0.12 %). Importantly, AQP4 immunostaining in kaolin + decorin treated rats was significantly lower (0.56 ± 0.09 %, p = 0.006) than in kaolin animals (1.77 ± 0.14 %). No significant difference existed in periventricular white matter AQP4 immunostaining between decorin treated and Intact control rats (p = 0.860). In contrast, significant differences in OX-42 and MBP levels were not present between the experimental groups in the periventricular white matter (Fig. 3c, d). Furthermore, significant differences in GFAP, OX-42, MBP and AQP4 immunostaining were not present between the experimental groups in the caudal corpus callosum (Additional file 3: Table S1).

Although significant differences in myelin levels were not present in the caudal corpus callosum and caudal periventricular white matter, loss of myelin (assessed by measurement of MBP) in the caudal internal capsule was present in kaolin and kaolin + PBS animals (Fig. 4); compared to Intact controls (8.11 ± 0.49 %), a decrease in MBP immunostaining was present in kaolin (3.13 ± 0.28 %, p < 0.001) and kaolin + PBS (5.15 ± 0.47 %, p = 0.001) rats. Furthermore, decorin treated rats displayed higher MBP levels (5.87 ± 0.29 %) compared to kaolin (3.13 ± 0.28 %, p = 0.002) and kaolin + PBS rats (5.15 ± 0.47 %, p = 0.018), although the levels did not quite reach the same as in intact rats (p = 0.009), indicating some myelin protection. Qualitatively, the longitudinal organisation of myelin was disrupted, with discontinuity present along the length of the myelin fibres in animals receiving kaolin and kaolin + PBS compared to Intact controls. The regular parallel arrangement of MBP staining was protected with decorin treatment.

Upon examining total hippocampal area, significant differences were present between the four experimental groups; a decrease in normalised hippocampal area was identified in the kaolin (47 ± 9 %, p = 0.006) and kaolin + PBS rats (69 ± 9 %, p < 0.001) compared to Intact controls (100 ± 6 %). Decorin treatment attenuated the hippocampal atrophy (89 ± 7 %, p = 0.008) compared to kaolin rats but failed to maintain the hippocampal size to that in Intact controls (100 ± 6 %, p < 0.001). In the CA1 region of the hippocampus, similar levels of GFAP, OX-42 and AQP4 were observed between all the experimental groups. In the CA3 region, GFAP levels were comparable among all four groups however kaolin rats (0.43 ± 0.02 %, p = 0.057) demonstrated a trend towards reduced OX-42 levels compared to Intact controls (0.89 ± 0.03 %), and kaolin + PBS rats (1.04 ± 0.13 %, p = 0.043) displayed a significant increase in AQP4 levels compared to Intact controls (1.60 ± 0.10 %; Additional file 4: Table S2).

No significant differences in AQP4, OX-42 and GFAP immunostaining were present between the experimental groups in the caudate putamen, parietal cortex and occipital cortex (Additional file 4: Table S2). However, in the corpus callosum, periventricular white matter, caudate-putamen, parietal cortex and occipital cortex, GFAP immunostaining positively correlated with AQP4 levels (Table 3). No significant correlations were present between OX-42 and AQP4 levels in any of the regions of interest.

In the caudal corpus callosum, increased astrocyte (GFAP) and AQP4 levels positively correlated with the AD (Table 4). Furthermore, in the caudal periventricular white matter (Table 4), GFAP and AQP4 positively correlated with AD, MD and RD. Moreover, the presence of cytopathology discouraged anisotropic water diffusion in the caudal periventricular white matter as the FA negatively correlated with astrocyte (GFAP), microglial (OX-42) and AQP4 immunostaining. A negative correlation was also present between myelin levels (MBP) and the MD of the caudal periventricular white matter.