Gadolinium deposition in the brain of dogs after multiple intravenous administrations of linear gadolinium based contrast agents

Objective To determine the effect of a linear gadolinium-based contrast agent (GBCA) on the signal intensity (SI) of the deep cerebellar nuclei (DCN) in a retrospective clinical study on dogs after multiple magnetic resonance (MR) examinations with intravenous injections of gadodiamide and LA-ICP-MS analysis of a canine cerebellum after gadodiamide administration. Animals 15 client-owned dogs of different breeds and additionally 1 research beagle dog cadaver. Procedures In the retrospective study part, 15 dogs who underwent multiple consecutive MR imaging examinations with intravenous injection of linear GBCA gadodiamide were analyzed. SI ratio differences on unenhanced T1-weighted MR images before and after gadodiamide injections was calculated by subtracting SI ratios between DCN and pons of the first examination from the ratio of the last examination. Additionally, 1 research beagle dog cadaver was used for LA-ICP-MS (Laser ablation inductively coupled plasma mass spectrometry) analysis of gadolinium in the cerebellum as an add-on to another animal study. Descriptive and non-parametrical statistical analysis was performed and a p-value of < 0.05 was considered significant. Results No statistically significant differences of SI ratios, between DCN and pons, were detectable based on unenhanced T1-weighted MR images. LA-ICP-MS analyses showed between 1.5 to 2.5 μg gadolinium/g tissue in the cerebellum of the examined dog, 35 months after the last of 3 MRI examination with gadodiamide (two examinations at a dose of 1 x 0.1mmol/kg, last examination at a dose of 3 x 0.05mmol/kg). Conclusion and clinical relevance Although the retrospective MRI study did not indicate any visible effect of SI increase after multiple gadodiamide exposures, further studies based on LA-ICP-MS showed that the optical threshold was not reached for a potential visible effect. Gadolinium was detectable at a level of 1.5 to 2.5 μg gadolinium/g tissue by using LA-ICP-MS in the cerebellum 35 months after last MRI examination. The general importance of gadolinium retention of subvisible contents requires further investigation.


Introduction
In 2014 Tomori Kanda [1] published his work about signal intensity (SI) increase on unenhanced T1 weighted magnetic resonance imaging (MRI), causing hyperintensities in the dentate nucleus (DN)-to pons and globus pallidus to thalamus after consecutive injections of gadolinium-based contrast agents (GBCAs). His important work was the beginning of an ongoing debate about the deposition of GBCAs in patient's brains. [1][2][3][4][5][6][7][8][9][10][11][12][13][14][15][16][17][18] It is shown in numerous studies on humans and in a few animal studies that there is a positive correlation between SI increase on unenhanced T1 weighted MRI and the gadolinium concentration in the brain. [19][20][21][22][23][24][25] Currently, the scientific interest focuses on the question whether all GBCAs or only a specific GBCA subtype is causing such hyperintensities. As free ionic gadolinium is toxic, GBCAs are applied as chelates. According to their chemical structure, two subtypes of GBCAs can be classified. GBCAs of the linear subtype partially and GBCAs of the macrocyclic subtype completely enclose gadolinium. On the one hand, the majority of studies provided evidence that linear GCBAs are stronger correlated with SI increase in the DN. [1, 3-7, 13-15, 17, 18, 20, 22-24, 26-43] On the other hand, SI increase was described after macrocyclic GBCA administration in a few studies that were controversial discussed between specialists due to their limitations. [3,5,7,8,17] It seems that the molecular structure of the GBCA ligand, which defines the GBCA subtype, is a crucial factor for the SI increase on unenhanced T1 weighted MRI. [44] In 2017, as a consequence of the scientific debate, regulatory authorities (EMA, European Medicine Agency; FDA Food and Drug Administration) decided about safety issues using GBCAs during clinical work-up, which resulted in divergent actions between EU and USA. [25] Based on a precautionary approach in the European Union nearly all linear GBCAs were removed from the market, while the FDA issued a class warning for all GBCAs.
Until now, the majority of published animal studies about this topic were performed on rodents [23,24,[45][46][47][48][49] the minority on large experimental animals. [25] So far, in veterinary medicine, no studies described a potential SI increase after multiple linear GBCA administrations in client-owned dogs. This species is of special clinical interest, as dogs will potentially face multiple MR examinations in veterinary medicine during treatment of tumors or during neurological diseases. Moreover, dogs play an indispensable role as animal model for translational research approaches. Therefore, the question arises if in dogs, similar to humans, an SI increase in the DCN is detectable after consecutive MR examinations or whether the number of GBCA administration is below the limit of detection for a visible SI increase on unenhanced T1-weighted MR images. This study followed the guidelines for standardized assessment of SI increase for retrospective MRI studies published from the European Gadolinium Retention Evaluation Consortium (GREC). [44] In the current study, we aimed to retrospectively assess clinical data of dogs after multiple MRI examinations with intravenous administration of gadodiamide. Additionally, the brain of one dog, which was euthanized unrelated to this study, underwent an LA-ICP-MS measurement of gadodiamide 35 months after last MRI examination. [50][51][52] We hypothesized that there is an increased hyperintensity on non-enhanced T1 weighted sequences in the deep cerebellar nucleus (DCN) of dogs after multiple gadodiamide administrations and a detectable gadodiamide retention in the cerebellum of dogs with LA-ICP-MS.

Dataset of patients
This restrospective study was performed on MR imaging data sets of 18 client-owned dogs presented between August 2012 and October 2017 at the Clinic for Diagnostic Imaging at the Vetsuisse Faculty of the University of Zurich. Inclusion criteria was that the dogs had multiple (defined as two or more) MR examinations of the brain with intravenous administration of the linear GBCA gadodiamide at a dose of 0.15mmol/kg. Three patients did not undergo unenhanced T1-weighted imaging or were unreadable for study purposes and were excluded from this study. However, 15 client-owned dogs met the inclusion criteria and had a history of neoplastic or neurological disease as reason for MRI examination. Demographic and clinical characteristics of the all animals in the dataset is summarized in Table 1. All animals underwent general anesthesia during MRI examination based on a standard anesthesia protocol. After premedication with butorphanol and continious intravenous lactated Ringer's infusion, general anesthesia was induced with propofol and maintained with isoflurane combined in oxygen and air. One research beagle dog cadaver, which was euthanized unrelated to this study, was used for dissection and sampling of the whole brain. Accordingly, the additional organ sampling, which was an add-on to the main study purpose, applied 3R requirements and maximized the scientific output from the dog. LA-ICP-MS was used for determination of Gd-concentration in the cerebellum.
Image analysis was performed independently by two of the authors (HR, PK). A picture archiving and communication system (Synapse 1 PACS, Fujifilm) was used for all reading sessions. Evaluation of the images, including analysis of ROIs and the mean SI values, was performed with open source medical image viewer (Horos based upon OsiriX TM , 64-bit medical image viewer for OS X, Version 3 (LGPL-3.0)). Pre-and postcontrast images were subjectively compared and analyzed regarding presence of signal alteration in the region of the deep cerebellar nuclei (DCN). Regions of interest (ROI) were drawn on the unenhanced T1-weighted images on the central pons, and the DCN on both hemispheres. (Fig 1) The anatomically correct description of the DN in dogs is Nucleus lateralis cerebelli, which is the most prominent DCN. Accordingly, we use the more generalized term DCN instead of DN or Nucleus lateralis cerebelli to better reflect the canine anatomical situation. The correct placement of the ROI was confirmed by using T2-weighted images at the same section position for identification of the DCN in unclear cases. The DCN-to-pons SI ratio was calculated by using the following formula: DCNÀ toÀ pons SI ratio ¼ ðmean SI of DCNÞ=ðmean SI of central ponsÞ: The first and last MR examination of the dogs was used to calculate the SI ratio difference by substracting the first examination SI ratio from the last examination SI ratio.

Cadaver sample preparation
The research beagle dog was euthananized according to another unrelated study (animal license number ZH057/17) in accordance to Swiss animal welfare act. As an add-on to the main study, the dog's brain was harvested directly after euthanasia. The left brain hemisphere was coronally cut into 0.5-1.0 cm slices and cryopreserved (at − 80˚C). Deep cerebellar nuclei of the left cerebellum were cut into a 50-μm-thick section and fixed on a piece of cork with Tissue-Tek O.C.T. Compound (Sakura Finetek GmbH, Staufen, Germany). For chemical analysis, the deep cerebellar nuclei were cut in thin sections of 10-μm thickness with a cryotome and mounted on microscopic glass slides. Before the ablation process, microscopic images were recorded with a BZ-9000 inverted fluorescence/bright field microscope (Keyence, Osaka, Japan). The right hemisphere of the brain was formalin fixed (4% buffered formalin), paraffin embedded, and stored until further analysis.

Localisation of deep cerebral nuclei by bench-top μXRF
For the localisation of deep cerebral nuclei, an M4 Tornado bench-top μXRFinstrument (Bruker Nano GmbH, Berlin, Germany) was used. The rhodium X-ray tube was supplied with a voltage of 50 kV and a current of 600 μA throughout the measurements. The emitted X-ray fluorescence was detected by a silicon drift detector (XFlash1 5030, Bruker Nano GmbH, Berlin, Germany). Spatial resolution was approximately 25 μm. Each pixel was measured twice for 200 ms at a pressure of 20 mbar. The data was evaluated using the software ESPRIT HyperMap (Bruker Nano GmbH, Berlin, Germany).

Standard preparation for LA-ICP-MS
For calibration, 11 matrix-matched gelatine standards in a concentration range from 0 ng/g to 600 μg/g were created. For this purpose, a 1000 μg/g gadolinium (Gd) ICP-MS standard (Fluka Analytical, St. Gallen, Switzerland) was diluted and combined with gelatine of highest purity (Grüssing GmbH, Filsum, Germany) resulting in standards with a gelatine content of 10% (w/w). The suspension was homogenized by heating the standards to a temperature of 50˚C and the usage of a vortex mixer. Afterwards, 10 μm thin slices of each standard were prepared via a CryoStar NX70 Cryostat (Thermo Fisher Scientific, Waltham, USA). The thin slices of the standards where then mounted onto a microscopic slide. The Gd concentration in the gelatine standards was confirmed by ICP-TQMS (iCAP TQ, Thermo Fisher Scientific, Waltham, USA) analysis of the digested standards. For digestion, 500 μL concentrated nitric acid (Merck Chemicals GmbH, Darmstadt, Germany) and 100 μL 35% (w/w) hydrogen peroxide (Acros Organics, Geel, Belgium) were added to around 50 mg of each gelatine standard. The mixture was then heated to a temperature of 70˚C until complete digestion. Afterwards, the digests were diluted and a rhodium (Rh) ICP-MS standard (SCP Science, Baie-D'Urfe, Canada) was added as an internal standard. The concentration was determined using an external Gd calibration in a range from 0 pg/g to 30 ng/g consisting of 12 standards. Again, Rh was used as an internal standard.

LA-ICP-MS measurements
For elemental mapping of the Gd distribution, the hyphenation of an LSX 213 G2+ (Cetac Technologies, Omaha, USA) laser ablation system equipped with a HelEx II cell (Teledyne Cetac Technologies, Omaha USA) and an ICP-MS 2030 (Shimadzu, Kyoto, Japan) was used. For ablation, a spot size of 10 μm in combination with a stage speed of 30 μm/s was chosen. Laser energy was optimized for each individual sample to allow for quantitative ablation of the sample with minor ablation of the object slide. Ablated particles were washed out by a constant helium flow of 800 mL/min. For more efficient transportation, a daily tuned argon flow was added on the line to the micro torch of the ICP-MS system. A wet argon plasma with an RF power of 1200 W was used for the ionisation. The ionized analytes were led through a nickel interface and were analysed by an SQ-KED setup.
For quantification, the previously described matrix-matched Gd gelatine standards were used. On each standard, 11 lines with an ablation time of 20 s per line were ablated with the same parameters as used for the sample. The first line of each recording was discarded for consistency reasons.
All data evaluation was performed using the software ImaJar (developed by Robin Schmid, Muenster, Germany).

Statistics
Statistical data was analyzed by using SPSS (IBM 1 SPSS 1 Statistics, version 25, 64-bit-version, IBM, Chicago, Ill). Due to the limited number of available cases, data was defined as non-normally distributed and quantitative data was presented with median (range). Non-parametrical tests were used to compare SI ratio differences between 2 independent observers (Mann-Whitney U test). Inter-rater reliability was assessed based on Intraclass-Correlation-Coefficient (ICC) in a range from 0.0 to 1.0, whereby large numbers mean better reliability. One-sample ttests were used to examine if the mean SI ratio differences between first and last examination were different from 0. A p-value of < 0.05 was considered significant.

Cinical and demographic aspects
An overview about the demographic and cinical characteristic of the dataset is available provided as Table 1. The mean age of all dog patients included was 96.20 (31-157) months. The animals gender was divided as followed: male (5), male castrated (5), female castrated (5). As this was a restrospective clinical study, the dogs were of different breeds: Boxer (3), Border Collie (1), French Bulldog (2), Jack Russel Terrier (2), Catalan Scheepdog (1), Labrador Retriever (3), Magyar Vizsna (1), Spitz (1), Yorkshire Terrier (1). Contrast media injection protocols were the same for all animals, as they followed a clinical standard operation procedure with 0.15 mmol/kg gadodiamide for each MRI examination. The time between the MRI examinations was caused by the clinical work-up and was in median 136 (24-582) days.
Additionally, one research beagle dog cadaver was included into this study, which was euthanized for another unrelated study. The research beagle dog was used for different studies including repetitive gadodiamide administrations in MRI before his termination. It was a 52 months old, female castrated beagle dog with 11.4 kg body weight. The contrast media injections of the dog were as followed: November 2013 (0.1 mmol/kg gadodiamide), December 2013 (0.1 mmol/kg gadodiamide), June 2015 (3 x 0.05 mmol/kg gadodiamide). Accordingly, the last gadodiamide injection was 35 months before termination. Following the 3R requirements, the brain of the dog was sampled and used for chemical analysis, as this was not possible to perform with the retrospectively examined clinical patient population.

MRI data
Unenhanced T1-weighted imaging were independenly analyzed by two observers (HR, PK). A summary of all measurements are provided as Table 2. SI values of the DCN and Pons as well as the SI ratios and the SI ratio differences between the first and the last examination of the left and right hemisphere are summarized in Table 3. Based on non-parametric Mann-Whitney U tests, no significant differences were detectable between both observers at a significance level of The ICC showed very good agreement between both observers (Table 3). For all analyses, ICC was between 0.863 (SI ratio difference right hemisphere)-0.999 (SI DCN left, SI Pons left, SI DCN right, SI Pons right).
SI ratio DCN-to-pons showed no significant differences between left and right hemisphere (observer 1 (p = 0.390), observer 2 (p = 0.722)). SI ratio DCN-to-pons at the right hemisphere was measured in median with 0.997 (0.860-1.072) for observer 1 and 0.991 (0.887-1.071) for observer 2, and at the left hemisphere with 0.999 (0.906-1.111)) for observer 1 and 0.989 (0.884-1.089) for observer 2. (Table 4) SI ratio differences between the first and the last MRI examination of the patients showed no significant differences, whether at the left nor at the right hemisphere (observer 1 (p = 0.831), observer 2 (p = 0.163)). SI ratio differences between the first and the last MRI examination of the patients was measured at the right hemisphere in median with -0.011 (-0.125-0.069) for observer 1 and -0.021 (-0.100-0.043) for observer 2, and at the left hemisphere with -0.002 (-0.079-0.138) for observer 1 and -0.010 (-0.096-0.082) for observer 2. (Table 1) Based on a one-sample t-test, the first and the last MRI examination was tested to be significant different from 0. For both observers, no significant difference from 0 was found (observer 1: p = 0.896/0.402; observer 2: p = 0.451/0.098 [left/right]).

LA-ICP-MS data
Visual assessment of LA-ICP-MS analysis as well as a quantitative analysis was performed with a limit of quantification (LOQ) of 220 ng gadolinium/g tissue and a limit of detection (LOD) of 67 ng/g. Gadolinium concentrations in the DCN of the research beagle dog could be determined between 1.5 to 2.5 μg gadolinium/g tissue, 35 months after last gadodiamide injection. In agreement with the visual assessment of the DCN, gadolinium levels were higher in the DCN than in the surrounding area of the DCN. A co-localisation of Gd and zinc (Zn) was detectable, both increasingly detectable in vessels, especially at the DCN. Furthermore, an anti-correlation with phosphorus (P) was detectable, which is visible in μXRF and LA-ICP-MS. The results are summarized in

Discussion
The current restrospective clinical study is the first study observing multiple gadodiamide applications in veterinary patients in a clinical set-up. The background of this study was the ongoing debate about the clinical relevance of hyperintensities of the DCN after multiple GBCA applications in humans. The hypothesis of this study was that there is a detectable increased hyperintensity on non-enhanced T1 weighted sequences in the DCN of dogs after multiple gadodiamide administrations and a measurable gadolinium retention in the cerebellum of dogs after chemical analysis based on LA-ICP-MS. The hypothesis has to be partially rejected. This study showed 1) no hyperintensities on non-enhanced T1 weighted sequences in the clinical dog patients and 2) a Gd concentration of 1.5 to 2.5 μg gadolinium/g tissue in the brain of a research beagle dog 35 months after last gadodiamide injection.  With these findings, our study supports the hypothesis that a specific threshold is needed to identify hyperintensities on unenhanced T1-weighted images. It seems likely that the data of this dog patient population was not exposed to gadodiamide in an amount to exceed this threshold. With a median cumulative dose of 6.83 mmol/animal (range 0.54 to 16.65mmol/ animal) the clinical dataset shows much less cumulative GBCA dose as compared to published reports from human medicine [1] or animal experiments. [53] Even if a superior brain clearance of macrocyclic GBCAs over linear GBCAs is described [54,55] and all dogs of this study got a linear GBCA administration, the cumulative dose did not reach a visible level of hyperintensities after 2-3 administrations. We determined mean SI ratio differences between first and last MRI examination between 0.2% and 6.9% (range -12.5% to 13.8%), which was not significantly different from 0. (Table 1) This is much lower than in the study of Hu et al., were an SI increase between first and most recent MRI examination with an increase of 18.6% ± 12.7% (range 0.5% to 47.5%) for the DCN was reported. [14] In the study of Hu et al., human patients received a significant number of subsequent GBCA examinations, ranging between 5 and 37. Weberling et al. described a significant SI increase in the DCN after at least 5 consecutive GBCA injections of gadobenate dimeglumine. [13] Although the retrospective MRI study did not indicate any visible effect of SI increase after multiple gadodiamide exposures, further studies based on LA-ICP-MS showed that the optical threshold was not reached for a potential visible effect. Most of the dog patients in this study had a history of neoplastic or neurological disease as reason for MRI examination (Table 1). In the patient records of the dogs, no side effects were linked with the contrast media application. This result is of clinical relevance for all current and future patients in veterinary medicine, which have a history of diagnosed brain tumors (such as meningiomas gliomas, nerve sheeth tumors, pituitary adenomas) or meningoencephalitis. Recurrent MRI examination are of interest in veterinary medicine mainly in dog patients after radiation therapy for follow ups and/or recidive pathologies. Our dataset showed that mostly not more than two MRI examinations with GBCA are performed during clinical work-up in veterinary medicine. This is a clear difference in comparison with human medicine, were more than 35 linear GBCA administrations to a single patient are reported. [18] Even if there is no clinically visible hyperintensity on unenhanced T1-weighted MR images, this does not mean there is no deposition of gadolinium through gadodiamide application after multiple contrast media administrations in dogs. From human medicine it is known that approximately 6 injections of linear GBCAs are needed, before hyperintensities in the DCN become detectable on MRI [7,56] On the other hand, gadolinium deposition was detectable after a single injection of linear GBCAs based on LA-ICP-MS measurments in the brain of sheep. [25] This study confirms gadolinium detection in the brain of a research beagle dog based on the chemical analysis of the Gd content. After three examinations, two times with 1x0.1 mmol/kg and once with 3x0.05 mmol/kg gadodiamide, Gd was measurable 35 month after the last administration. Accordingly, it could be shown in this study that the described Gd deposition in the brain of different species similarly occurs in dogs. Therefore, our findings support the hypothesis of irreversible or long lasting Gd deposition at the DCN. The current explanation about the pathomechanism is based on de-chelation effects of linear GBCA and transmetalation in co-localization with other elements, such as Fe, Co, Cu, Zn. [25,57] We could show similar effects in the research beagle dog, which supports the ongoing discussion. As dogs are a widely used species for translational medicine and model for lots of pre-clinical studies during the approval of new medical products for human medicine, the results of our study should be recognized for all translational studies in future.
Even if the pathomechanism is not clarified at the moment and there are ongoing studies about differences between linear and macrocyclic GBCAs, from a scientific and ethical point of view it is mandatory to reduce the GBCA administration to the essential number with diagnostic importance for the patients. At the same time it is important to maintain the benefit of GBCA as long as there are no valuable alternative contrast media available. This should be done with respect to the risk of Gd deposition and its accumulation during recurrent contrast agent administrations.
As one of the limitations of this study, we have to mention that the number of patients and the number of gadodiamide administrations was lower than in comparable restrospective human studies. The reason is that in veterinary medicine dogs, as the most used private owned species, will not be imaged more often in MRI. This can be explained by the costs (payed by the owner) and the shorter life-span of a dog in comparison to humans. As our data reflects the daily business in a large university hospital, this was the maximum available data, which provides a relevant veterinary insight into this topic. Unfortunately, we did not have an MRI of the research beagle dogs brain, who had multiple gadodiamide administrations before euthanasia. This was caused by the fact that the brain of the dog was used as an add-on to another unrelated study which focused on other regions than the animals brain during MRI examination. It is a beneficial circumstance to have the results of the chemical analysis from this dog, even if there is no associated MRI of the brain. Based on the gained knowledge that an increased signal intensity of the brain might only be visible when the number of GBCA injections surpasses a certain threshold [34], it may have been questionable if this threshold would have been exceeded or not. According to the current knowledge, gadolinium first becomes visible in MRI if a threshold of approximately 1 μg gadolinium/g tissue is exceeded. [25] We assessed between 1.5 to 2.5 μg gadolinium/g tissue in the DCN of the research beagle dog, which would have caused a signal intensity change in MRI. Currently published case reports and research studies describe LA-ICP-MS as a useful tool to measure the quantities of retained gadolinium in human [19,58,59] and rodent tissues [47,60]. Potential risk factors with effect on Gd retention described are ongoing neuroinflammation [60] or primary gliomas [59]. Additionally, any type of inflammation can negatively influence the blood brain barrier permeability [61] and therefore could be potential risk factor for gadolinium retention. In this study LA-ICP-MS was performed in the brain of a healthy research beagle dog, which was considered to be not influenced by clinical factors. On the other side the dataset of this study includes patients showing exactly such types of diseases with potential effect on gadolinium retention. The question how and in which quantity the underlying diseases or a reduced blood-brain-barrier influenced the gadolinium retention in the brain can not be estimated and has to be evaluated in further studies.
Another potential confounding factor for all clinical studies is the time between last GBCA administration and the MRI examination to detect hyperintensities in the DCN or until LA-ICP-MS measurement of gadolinium, as this time is crucial to differentiate between the soluble and insoluble form of gadolinium. In this study the median number of days between last GBCA application and MRI examination was 136 (range 24 to 582days). Accordingly there was enough gadolinium-free-time to detect insoluble gadolinium deposits only.
In conclusion, we could show that no hyperintensities on non-enhanced T1 weighted sequences in the clinical dog patients after 2-3 gadodiamide administrations of 0.1 mmol/kg were detactable and that between 1.5 to 2.5 μg gadolinium/g tissue could be determined in the brain of a research beagle dog 35 months after his last gadodiamide injection. The importance and clinical relevance of gadolinium retention of subvisible contents, the pathomechanism and the potential side effects based on Gd deposition still requires further investigation.