Computed tomography provides enhanced techniques for longitudinal monitoring of progressive intracranial volume loss associated with regional neurodegeneration in ovine neuronal ceroid lipofuscinoses

Abstract Introduction The neuronal ceroid lipofuscinoses (NCLs; Batten disease) are a group of fatal neurodegenerative lysosomal storage diseases of children caused by various mutations in a range of genes. Forms associated with mutations in two of these, CLN5 and CLN6, are being investigated in well‐established sheep models. Brain atrophy leading to psychomotor degeneration is among the defining features, as is regional progressive ossification of the inner cranium. Ongoing viral‐mediated gene therapy trials in these sheep are yielding encouraging results. In vivo assessment of brain atrophy is integral to the longitudinal monitoring of individual animals and provides robust data for translation to treatments for humans. Methods Computed tomography (CT)‐based three‐dimensional reconstruction of the intracranial volume (ICV) over time reflects the progression of cortical brain atrophy, verifying the use of ICV measurements as a surrogate measure for brain size in ovine NCL. Results ICVs of NCL‐affected sheep increase for the first few months, but then decline progressively between 5 and 13 months in CLN5−/− sheep and 11–15 months in CLN6−/−sheep. Cerebral ventricular volumes are also increased in affected animals. To facilitate ICV measures, the radiodensities of ovine brain tissue and cerebrospinal fluid were identified. Ovine brain tissue exhibited a Hounsfield unit (HU) range of (24; 56) and cerebrospinal fluid a HU range of (−12; 23). Conclusions Computed tomography scanning and reconstruction verify that brain atrophy ovine CLN5 NCL originates in the occipital lobes with subsequent propagation throughout the whole cortex and these regional differences are reflected in the ICV loss.


| INTRODUC TI ON
The neuronal ceroid lipofuscinoses (NCLs, Batten disease) are a group of mainly autosomal recessively inherited lysosomal storage diseases caused by mutations in a number of genes that lead to degenerative and fatal encephalopathies in children (Haltia, 2006). Affected children progressively develop blindness, motor degeneration, epilepsy, and dementia, eventually reaching total dependency and premature death (Kousi, Lehesjoki, & Mole, 2012;Mole, Williams, & Goebel, 2005, 2011. There are no effective treatments. The NCLs are unified through severe cerebral neurodegeneration, neuroinflammation, and the nearly ubiquitous intracellular accumulation of storage bodies containing specific proteins (Kousi et al., 2012;Mole, Williams, & Goebel, 2005;Palmer, 2015;Palmer et al., 2015). A flock of sheep with a mutation in the CLN5 gene and two with mutations in the CLN6 gene provide well-established large animal models that largely mirror the human pathological changes including the intracellular accumulation of subunit c of the mitochondrial ATP synthase (Cook et al., 2002;Frugier et al., 2008;Jolly et al., 1989;Jolly, Arthur, Kay, & Palmer, 2002;Palmer, 2015). Other common features of the ovine and human NCLs include progressive cerebral atrophy and enlargement of the cerebral ventricles (Frugier et al., 2008;Jolly et al., 2002Jolly et al., , 1989Woods et al., 1993). Neuroimaging techniques, such as magnetic resonance imaging (MRI) and computed tomography (CT), have been used to monitor changes of brain size and intracerebroventricular volumes in both human and ovine NCLs (Dyke et al., 2016;Jadav et al., 2014;Lobel et al., 2013;Sawiak et al., 2015;Valavanis, Friede, Schubinger, & Hayek, 1980;Woods et al., 1993). Viral-mediated gene therapy trials are being conducted in affected sheep (Palmer, 2015). In vivo assessment of brain atrophy is integral to the longitudinal monitoring of individual animals and provides robust data for translation to human clinical trials.
It has been observed at postmortem examinations of NCL-affected animals that their skulls are consistently thicker than those of controls and that the intracranial volume (ICV) reduces with disease development, suggesting that CT-based measurement of the ICV could be a suitable surrogate measure for brain size changes in ovine NCL. Determination of tissue volumes from CT data requires discrimination of tissue boundaries. Tissues exhibit differences in their ability to attenuate radiation; that is, they have differing radiodensities. The CT number or Hounsfield unit (HU) provides quantification of the extent to which a tissue attenuates X-rays (Table 1) (Hounsfield, 1980) and can be used to identify boundaries between adjacent tissues. Human brain HU values range between 20 and 50, and those of cerebrospinal fluid (CSF) vary between −5 and 20 (Arimitsu, Di Chiro, Brooks, & Smith, 1977;Hacker & Artmann, 1978).
Neurodegeneration may also alter the radio-density of brain tissue, as has been reported in sheep and dog models of NCLs (Armstrong, Quisling, Webb, & Koppang, 1983;Woods et al., 1993). The aims of this study were to determine sheep-specific HU thresholds for brain tissue and CSF, to investigate the relation of ICV and brain volume in NCL-affected animals and controls and to perform a longitudinal study of ICV and brain volumes in CLN5-and CLN6-affected sheep and controls.

| MATERIAL S AND ME THODS
All experiments were performed in accordance with the Lincoln University Code of Ethical Conduct for the Use of Animals, under the New Zealand Animal Welfare Act 1999 and were approved by the Lincoln University Animal Ethics Committee. Computed tomography was conducted using a GE Prospeed CT scanner (GE Healthcare, Hyogo, Japan, 1997).
Previously performed CT scans of 56 sheep were analyzed to determine sheep-specific HU thresholds for brain tissue and CSF.
The sheep included both males and females aged between 3.6 and 68.4 months and were CLN5 +/and CLN5 −/− Borderdales, CLN6 +/-CLN6 −/− South Hampshires and unrelated control Coopworth and Romney crosses. HU values of brain (average of grey and white matter) and CSF-filled spaces were determined using ImageJ (Rasband, ImageJ, US National Institutes of Health, Bethesda, MD, USA) in five images selected from each image set based on the image quality and the visibility of the lateral cerebral ventricles. The average HU values of brain tissue were recorded from two 1-cm 2 circular areas selected for their visibility and uniformity of tissue type within the area, and both cortical and subcortical areas were included. Likewise, a 1-cm 2 circular area was selected within each of the lateral cerebral ventricles and the average HU values of these were recorded. Following this, the 99.9% confidence intervals representing the range of HU values of brain tissue and CSF spaces were determined.
For the longitudinal investigations and the comparison of ICV and brain size, scan settings were as follows: helical, 120 kV, 100 mA, 1 mm slice thickness at a 1-mm slice interval and image analyses were performed in 3D Slicer 4.3.1 freeware (https://www. slicer.org; 3D Slicer: RRID:SCR_005619) (Fedorov et al., 2012). The ICVs were measured from the first image containing the olfactory bulbs to the final image containing cerebellum using the previously established HU threshold for both brain tissue and CSF spaces, -12; 56 (see Section 3, Figure 1) using the "Editor" tool. The volumes of the third and lateral cerebral ventricles were measured using the previously established CSF HU threshold (−12; 23 and subarachnoid spaces) were not included in the CSF space measurements because the resolution of the CT scans was insufficient for accurate discrimination between brain tissue and CSF in these regions. Threshold-based selection of the areas was followed by manual correction for beam hardening and other scanning artefacts.
The "MakeModelEffect" tool was used to produce three-dimensional (3D) reconstruction and volumetric measurements of the regions of interest.
The relationship between ICV and brain volume was examined using severed heads from nine NCL-affected male, two NCL-affected female, and 14 unaffected female sheep aged 8-69 months that were euthanized for other reasons than this study. Following exsanguination, the heads were CT scanned and the ICV and lateral cerebroventricular volumes measured using the protocols described above. Following CT scanning, the skulls and meninges were opened dorsally and the brains removed, leaving the pituitary glands, olfactory bulbs, and the dura mater behind. The spinal cords were severed transversely at a point perpendicular to the caudal margin of the cerebellum. Within four hours of collection, brain volumes were measured using a water displacement method. A horizontal line was drawn on the outside of a glass beaker, which was then filled with water at room temperature to this line. Next a brain was placed in the beaker, which caused the water level to rise as the brain sank.
Water was removed until the water level was back to the initial line and the volume removed was recorded. The ratios of ICV to brain volume were calculated for both affected and normal sheep. The specific gravity of each brain was calculated by dividing the fresh brain weight by the water displacement volume, and the effects of disease status on ICV, brain-specific gravity, and ratio of ICV to brain volume were investigated.

All observations indicate that the neuropathology in ovine CLN5
Batten disease is independent of gender and sheep of all genders

| RE SULTS
The mean brain tissue HU value including both affected and unaffected animals after exsanguination was 39.5 with a standard deviation of 5.2, and thus, 99.9% of the ovine brain tissue was represented by the HU interval (24; 56). Likewise, 99.9% of CSF spaces was represented in the HU value interval of (−12; 23; Figure 1). The mean radio-density (HU value) of brain tissue in NCL-affected animals F I G U R E 1 Range of radio-density representing ovine brain tissue and CSF. Normal distribution of the Hounsfield units (HU) representing the radio-densities of ovine brain tissue (HU range of brain tissue) and those of ovine cerebrospinal fluid (HU range of CSF) (n = 72). Mean HU of brain tissue = 39.7; SD = 5. HU range of CSF HU range of brain tissue (n = 42) was greater than that of brain tissue from unaffected controls (n = 14) (mean affected = 40.3 ± 0.3; mean control = 38.4 ± 0.8; There was no such difference in the radio-density of CSFfilled spaces (mean affected = 5.6 ± 0.5; mean control = 5.5 ± 0.6; p = 0.8).
Brain volumes measured after post mortem by water displacement, and the corresponding ICVs measured from CT scans were both lower for NCL-affected animals than those of controls (Table 2).
Both CT scan ICVs and brain volumes of unaffected control animals closely matched, but the ICVs of NCL-affected animals measured by CT scanning were approximately 4 ml larger than the corresponding brain volumes measured by displacement. Based on this, the ratio of ICV to brain volume was larger in NCL-affected animals compared with unaffected controls. Disease status had no effect on the specific gravity of the brain ( Table 2).
The CT-based 3D reconstructions of the ICV of both CLN5 −/− and CLN6 −/− animals showed obvious visible changes compared with controls ( Figure 2). By 7 months of age, the affected ICVs of both affected genotypes were smaller than those of age-matched controls. At 17 months, the regionality of the brain atrophy was reflected in the shape of the ICVs. Whereas the cerebellar area of the ICVs was largely unchanged, the cerebral cortical areas appeared to be considerably thinner than those of controls, particularly in the occipital region ( Figure 2, https://www.tinkercad.com/things/1dlX-ELC0gvp-sheep-brain-cavity-19mth-control, https://www. tinkercad.com/things/5Y1f9AKrEiA-sheep-brain-cavity-19mth-batten-disease). Overall, ICVs from 17-month-old affected animals of both genotypes were smaller than those of controls, and also smaller than that they were at 7 months of age.
The progression of changes in ICVs of Borderdale (CLN5 −/− and CLN5 +/-) sheep is summarized in Figure 3. By 3 months, the mean ICV of CLN5 −/− sheep was smaller than that of unaffected age-and breed-matched controls. This difference became statistically significant by 5 months and remained so until the end of the trial period at 19 months. The accumulative change of ICV in CLN5 +/control animals initially increased between 3 and 5 months, then remained steady until approximately 9 months, followed by another growth phase until leveling off from about 15 months of age. The mean ICV of CLN5 −/− sheep initially increased between the ages of 3 and 5 months, similarly to that of controls. Thereafter, the volume decreased, plateaued between 7 and 11 months, and then decreased rapidly until the end of the trial period at 19 months. On average, CLN5 −/− sheep (n = 6) lost a total ICV of 6.7 ± 2.6 ml between 3 and 17 months, whereas the mean ICV of control sheep (n = 6) increased by 2.7 ± 1.6 ml over that period. The accumulative change of ICV in CLN6 −/− animals was similar to changes in the CLN5 −/− animals. An initial increase was followed by a slight decline between 5 and 7 months, a plateau until approximately 10 months of age and then volumes declined rapidly until the end of the trial period at 17 months. A total mean ICV loss of 5.8 ± 2.5 ml occurred in CLN6 −/− sheep (n = 6) between 3 and 17 months, while controls (n = 6) gained 3.8 ± 1.2 ml ICV on average over the same period.
The mean volumes of the lateral and third cerebral ventricles are summarized in Figure 5. Those of CLN5 −/− animals were significantly larger than those of controls throughout the entire trial period, and although the mean ventricular volumes of CLN6 −/− animals tended to be higher than those of controls, this did not become statistically significant until 17 months of age.

| D ISCUSS I ON
Neuroimaging has been a component of disease monitoring in NCL research for some time but, until recently, the methods were basic and mainly focused on time-point analyses rather than longitudinal monitoring of individuals. This study sets out to investigate the improvements that could be achieved through automated methods of image analysis to establish the relationship between the reduction in ICV and brain atrophy, and describe the patterns of ICV change in CLN5-affected Borderdale, CLN6-affected South Hampshire, and unaffected control sheep.
Measurements of radio-densities of brain tissue and CSF spaces in NCL-affected sheep and unaffected controls led to improved image analysis. An increased radio-density of brain tissue, measured in HU, was observed in affected animals compared with controls. Note. Volumetric brain measurements of NCL-affected (n = 11) and unaffected control (n = 14) sheep. Intracranial volumes (ICV) measured on CT scans and brain volumes (Volume) measured by water displacement. p ≤ 0.05 regarded as significant.

TA B L E 2 Brain and intracranial volume comparisons
This has also been described in other models (ovine and canine) of NCL (Armstrong et al., 1983;Woods et al., 1993). The specific gravity of brains from affected animals was not increased even though the brains of affected sheep have been described as firmer in texture than those of controls (Jolly et al., 1989). The range of HU values representing brain tissue and CSF spaces reported here are in close agreement with findings from similar investigations in humans, where the mean HU value of grey matter was 39, that of white matter 32 and that of CSF ranged between −5 and 20 (Arimitsu et al., 1977;Hacker & Artmann, 1978). The present results have enabled semi-automated analyses of CT scans to be used for longitudinal monitoring of treatment trials in NCL sheep and to create 3D images.
The comparison between CT-based volumetric measurements of the cranial vault and brain volume measured by water displacement confirms the validity of the CT measurements as a surrogate for brain volume. The ratio of ICV to brain volume in normal sheep was close to unity. However, the ratio for NCL-affected animals was 1.06, which is larger (p = 0.03) than that of normal sheep and shows a slight discrepancy between the ICV and brain volume measurements recorded here. The reduction in ICV in NCL-affected animals and the 3D reconstruction thereof presented in this study support the theory of an ongoing ossification of the skull in these animals. However, the larger ratio of ICV to brain volume indicates that the brain atrophy precedes this ossification, suggesting that the ossification is not causative of brain atrophy, but is rather a symptom of it. This is in agreement with an increase in space between the atrophied brain and the cranial vault of CLN6 −/− NCL-affected sheep reported earlier (Cook et al., 2002).
In human NCLs and other forms of dementia, remodeling of the skull is unusual and is mainly seen in diseases that also affect other parts of the skeleton, such as mucopolysaccharidosis and other skeletal dysplasias (Jelin et al., 2017). However, thickening of the cranium, also called hyperostosis cranii ex vacuo, has been described in hydrocephalic children sporadically, usually as a side effect of successful ventricular shunting, and as a rare consequence of brain atrophy (Di Preta, Powers, & Hicks, 1994;Wolf & Falsetti, 2001).
More detailed investigations into the molecular mechanism, and the regional changes are needed to understand the dynamics of ICV changes in ovine NCL as well as in other species. The present ovine NCL models provide an excellent platform from which to investigate this phenomenon longitudinally.
Although it has been long established that CLN5-and CLN6affected sheep exhibit a reduction in brain size (Jolly & West, 1976;Jolly et al., 2002;Jolly, Janmaat, Graydon, & Clemett, 1982), longitudinal monitoring is required to describe the rate of this change. As brain size is related to overall body size (Courchesne et al., 2000;Ebinger, 1974;Edland et al., 2002), large individual variation would be expected in the measurements of ICV, and this might overshadow disease effects. Assessment of the rate of change instead of actual volumes (Figures 3 and 4)  and CLN6 sheep, the present study showed that the ICV of affected animals changed only minimally until the approximate age of 10 months, which can be described as an absence of gain rather than an active loss, while the ICV of control animals steadily increased throughout the same period. After this, the ICV was reduced in both CLN5 −/− and CLN6 −/− animals, while the ICV of control animals plateaued. The 3D reconstructions revealed that the ICV changes follow the regionality of neuropathology that has been previously described (Oswald et al., 2005;Oswald, Palmer, Kay, Barwell, & Cooper, 2008). The volumes of the third and lateral ventricles of all four genotypes were relatively stable throughout the entire study period. The third and lateral ventricular volumes of CLN5 −/− animals were larger than those of controls and a lesser similar trend was observed in CLN6 −/− animals, which did not become significant until the very end of the trial period ( Figure 3). Overall, ventricular enlargement was more pronounced in CLN5 −/− animals than in CLN6 −/− animals, which may indicate a differing pattern of brain atrophy in the different types of ovine NCL. In recent past, a high-resolution sheep brain MRI atlas has been established based on the brains of 18 sheep and it showed that the mean volume of the lateral and third ventricles combined was approximately 4.9 ml (Ella & Keller, 2015;Ella, Delgadillo, Chemineau, & Keller, 2017). The discrepancy between these measures and the measures recorded in the present study is likely consequent to differences in the measurement techniques and issues with the spatial resolution of the CT scanner used, and the present measures reflect the ventricular volumes sufficiently to compare affected and normal sheep within the same trial.
This study was limited by the age of the CT scanner available to the project, which has an inferior resolution compared with modern machines. The low resolution was insufficient to accurately discriminate tissue borders for small CSF-filled spaces, leading to varying measures of the ventricular volumes in this study. Newer measures are already being used as valuable tools for the assessment of disease progression in trials of potential new therapies for the treatment of human NCL .

ACK N OWLED G M ENTS
The authors thank the staff at the Johnstone Memorial Laboratory at Lincoln University for the animal care. This work was supported by the University of Otago-University of Canterbury MARS bioimaging group through a PhD stipend for K.N.R.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interests.