Ex vivo visualization of the trigeminal pathways in the human brainstem using 11.7T diffusion MRI combined with microscopy polarized light imaging

Classic anatomical atlases depict a contralateral hemispheral representation of each side of the face. Recently, however, a bilateral projection of each hemiface was hypothesized, based on animal studies that showed the coexistence of an additional trigeminothalamic tract sprouting from the trigeminal principal sensory nucleus that ascends ipsilaterally. This study aims to provide an anatomical substrate for the hypothesized bilateral projection. Three post-mortem human brainstems were scanned for anatomical and diffusion magnetic resonance imaging at 11.7T. The trigeminal tracts were delineated in each brainstem using track density imaging (TDI) and tractography. To evaluate the reconstructed tracts, the same brainstems were sectioned for polarized light imaging (PLI). Anatomical 11.7T MRI shows a dispersion of the trigeminal tract (tt) into a ventral and dorsal portion. This bifurcation was also seen on the TDI maps, tractography results and PLI images of all three specimens. Referring to a similar anatomic feature in primate brains, the dorsal and ventral tracts were named the dorsal and ventral trigeminothalamic tract (dtt and vtt), respectively. This study shows that both the dtt and vtt are present in humans, indicating that each hemiface has a bilateral projection, although the functional relevance of these tracts cannot be determined by the present anatomical study. If both tracts convey noxious stimuli, this could open up new insights into and treatments for orofacial pain in patients.


Introduction
Classically, orofacial nociceptive afferents, running through the trigeminal nerve and the trigeminal tract (TT), are believed to synapse in the ipsilateral trigeminal sensory nucleus complex (Sessle 2000). Second-order neurons then cross and ascend as the trigeminothalamic tract to the contralateral ventro-posterior medial nucleus of the thalamus. From there, projections ascend to the primary and secondary somatosensory cortices and other cortical regions, such as the insula (Ralston 2005). However, the exact anatomy of the central portion of the trigeminal pathways in humans remains a point of discussion. Knowledge of the exact neuroanatomy of the trigeminal tracts will contribute to the understanding of orofacial pain and its treatment. Preclinical experiments provide evidence for a bilateral projection of the orofacial region causing activation of both thalami and both primary sensory cortices in healthy subjects, as well as patients suffering from chronic orofacial pain (Nash et al. 2009(Nash et al. , 2010. This bilateral representation of the orofacial region can be explained by transcallosal pathways or by a bilateral trigeminothalamic system. The involvement of transcallosal pathways in the bilateral activation pattern of the aforementioned brain regions in the registration of orofacial pain seems to be of limited importance according to the split-brain study of Stein et al. However, proper interpretation of the results of this study is hampered due to the small sample size (n = 1) (Stein et al. 1989). Additionally, Solstrand Dahlberg and colleagues report findings that a decreased connectivity of regions in the brainstem that are involved in orofacial pain with parts of the descending pain modulation system is present in patients suffering from migraine. However, they also conclude that the exact anatomy of the trigeminal pathways in the human brainstem remain elusive (Solstrand Dahlberg et al. 2018). In 2016, a review of tracing studies in animals and functional MRI studies in humans provided an overview of the trigeminal tracts in the brainstem (Henssen et al. 2016). Figure 1 depicts two trigeminothalamic tracts sprouting from the principal sensory nucleus (PSN), the ventral-and dorsal trigeminothalamic tract (vtt and dtt, respectively). The vtt is a mainly contralateral pathway and is known to be present in human specimens. The dtt, on the other hand, is a mainly ipsilateral pathway and is predominantly known from animal-based studies and has only been hypothesized to exist in humans (Henssen et al. 2018).
Based on the aforementioned results, we hypothesized that dtt would exist in the human brainstem too. To test this hypothesis, we investigated the neuroanatomy of the dtt in human brainstems using, multiple, advanced neuroimaging techniques, including post-mortem, 11.7 anatomical magnetic resonance imaging (MRI) and diffusion-weighted MRI (dMRI) and several post-processing techniques (e.g., tract density imaging (TDI) and tractography). To complement the MR findings, the same brainstems were sectioned to obtain histological transverse sections for polarized light imaging microscopy (PLI). PLI is a microscopy technique to quantify fiber orientation based on birefringence of the myelin sheath in histological brain sections and has been reported as a well-suited technique to validate MR findings (Axer et al. 2011b).

Acquisition of the post-mortem material
Three post-mortem brains were derived from the body donor program of the department of Anatomy at Radboud University Medical Center, Nijmegen, The Netherlands. The three donors had no known neurological diseases and none of the brains showed pathological deformities after death. All specimens were perfused via the femoral artery using 10% formaldehyde and a relatively short postmortem interval (< 24 h) was employed to limit the reduction of several diffusion values [apparent diffusion coefficient (ADC) and fractional anisotropy (FA)] (D'Arceuil and de Crespigny 2007; Schmierer et al. 2008). After embalmment, the specimens were fixated in 7.7% formaldehyde for at least 3 months to a maximum of 3 years as it was suggested that diffusivity measures remain stable up to 3 years after fixation (Dyrby et al. 2007). Then the brain was extracted from the skull. The brainstem and cerebellum were separated from the cerebrum via a transverse section, perpendicular to the neural axis, at the level of the superior colliculus. The cerebellum was dissected from the brainstem by a coronal cut through the middle cerebellar peduncle. All three parts were fixated in 7.7% formaldehyde and preserved for another 2 months. See Table 1 for additional information. This study was carried out in accordance with the recommendations of the CMO (Commissie Mensgebonden Onderzoek) region Arnhem-Nijmegen, Netherlands. The post-mortem specimens were acquired via the body donor program at the department of Anatomy of the Radboud university medical centre, Fig. 1 Overview of the trigeminothalamic tracts that arise from the PSN in a dorsal view. Cranial nerves indicated in Roman numerals; dPSN dorsal principal sensory nucleus; vPSN ventral principal sensory nucleus; *dorsal trigeminothalamic tract (dtt), **ventral trigeminothalamic tract (vtt) Nijmegen, Netherlands. All body donors in this program signed a written informed consent during their lifetime, permitting the use of their body and parts for science and teaching.

Magnetic resonance image acquisition
Prior to MR scanning, the three post-mortem brainstems were stored in a phosphate-buffered saline solution (PBS 0.1M, pH 7.4) for 1 week to reverse the decrease of the T2 relaxation rate of tissue introduced by formaldehyde (Shepherd et al. 2009). Next, the brainstems were placed for 24 h in Fomblin® (Solvay Solexis Inc), a susceptibilitymatched, hydrogen-free liquid. The brainstems were then placed in a vacuum chamber for 10 min to reduce free air in the lumina of the blood vessels. Ultimate maximum pressure within the vacuum chamber was set to 1350 mmHg to prevent cell membrane damages (Gonzalez-Rodriguez et al. 2016;Holzapfel et al. 2005), although the maximum pressure did not exceed 500 mmHg during the experiments as this was not necessary to remove air bubbles. Finally, each brainstem was placed in a 100 ml syringe, filled with Fomblin, for MR scanning. All imaging was performed on a 11.7T Bruker BioSpec Avance III preclinical MR system (Bruker BioSpin, Ettlingen, Germany) equipped with an actively shielded gradient set of 600 mT/m (slew rate 4570 T/m/s). A circular polarized resonator was used for signal transmission and an actively decoupled birdcage coil (Bruker Biospin) was used for receiving. Scanning was performed at room temperature (20 °C). T2* weighted images were acquired using a 3D multi-gradient echo sequence at 0.25 mm isotropic resolution with the following parameters: TR = 3314 ms; TE = 7 ms; and flip angle = 20°. The dMRI data were obtained using segmented spin-echo echo-planar imaging at 0.5 mm isotropic resolution. This sequence consisted out of four segments. A total of 256 gradient directions were employed at a b value of 4000 s/mm 2 , in addition to six images with no diffusion weighting (b = 0 s/mm 2 ) and the following gradient and pulse parameters: Δ = 12.5 ms, δ = 4.0 ms, TE = 30.70 ms, and TR = 13.75 s. The applied MR protocol was adapted from an empirically designed protocol reported in the literature (Kleinnijenhuis et al. 2013).

Probabilistic tractography
All processing of the MR images was performed with FSL (Jenkinson et al. 2012). To correct for eddy current artifacts and displacement between the different diffusion images, an eddy current correction was applied (Andersson and Sotiropoulos 2015). The dMRI data were pre-processed for tractography with the bedpostX algorithm that models multiple fiber orientations (n = 3) at each voxel (Behrens et al. 2007). To delineate the TT in our specimens, probabilistic tractography was performed using the Probtrackx2 algorithm. A manually defined seed mask (i.e., the starting point of tractography) was placed in the entry zone of the trigeminal nerve on the fractional anisotropy (FA) maps. To address the reproducibility of the drawing of the masks in each specimen, two neuroanatomists, independently, delineated the location of the trigeminal root entry zone on both sides of each brainstem. The overlap of both masks was used as the seeding masks for tractography on either the right or left side of each brainstem. Streamlines were drawn from each seed-voxel (n = 50,000 streamlines per voxel). Only streamlines that reached the dorsal part of the brainstem near the PSN-defined by a waypoint mask-were kept to avoid the inclusion of other large dominating white tracts that run through the brainstem (Büttner-Ennever and Horn 2013). Maximum intensity projections (MIPs) of the trigeminal tract streamlines along aspecified encoded color direction, which improve the visualization of general tractography results as they provide stereoscopic visualization, were overlaid on the fractional anisotropy, which highlights the tract locations.

Track density imaging
Track density imaging (TDI) is typically used to investigate white matter microstructure smaller than the imaging resolution by reinterpolation of the quantitative maps by constrained spherical deconvolution based on the response 23 Time between fixation and extraction of the brain (months) 3 2 2 Time between extraction of the brain and scanning (months) 2 2 2 function at a finer resolution (Calamante et al. 2010). TDI was applied to the dMRI data to study the TT trajectory and its possible bilateral projections. To generate the directionally encoded color TDI maps, whole brainstem probabilistic fiber-tracking was carried out with MRtrix as described by Calamante et al. (2010). Streamlines were generated (total number of streamlines: n = 500,000; 20 per seed voxel) from a large number of random seeds throughout the brainstem. Seeds are placed randomly throughout the brainstem as described by other before (Calamante et al. 2010(Calamante et al. , 2011. The total number of streamlines was calculated in each element of a grid that covers the brainstem. It should be noted that the sampling distance of the grid was smaller than the acquired voxel size, yielding TDI map with higher spatial resolution than the source dMRI data (0.1 mm isotropic). After tracking, the streamline density was used as intravoxel information to construct a super-resolution TDI image (Calamante et al. 2010).

Histological tissue processing
Histological sectioning was performed for the TT at the level of the trigeminal nerve entry zone. This part was separated from the rest of the brainstem with a transverse cut through the inferior olivary nucleus and the inferior colliculus. To prevent the formation of ice crystals during freezing, the brainstems were stored in a 30% sucrose PBS solution (PBS 0.1M, pH 7.4) at 4 °C for 1 week. The specimen was frozen using dry ice and serially sectioned with a HM 450 Sliding Microtome (Thermo Fisher Scientific Inc., Waltham, Massachusetts, USA) at a thickness of 100 microns. In a series of five axial sections, the first section was mounted on glass, creating an inter-slice gap of 500 microns.

Polarized light imaging
PLI utilizes the optical birefringence of nerve fibers, which can be attributed to the arrangement of lipid layers in the myelin sheaths wrapping the axons (Axer et al. 2011a, c). Birefringence alters the polarization state of the light that is passed through the specimen depending on the orientation of the myelinated nerve fibers. As such, taking a series of images while rotating the linear polarizer plane allows for quantification of fiber orientation at each pixel. The abovementioned principle of measurement is referred to as polarimetry. To realize PLI measurements of the transverse sections of the brainstems, two polarimeters were used: a large-area polarimeter (LAP) for single shot medium-resolution (64 microns per pixel) images covering the whole tissue section and a polarization microscope (PM) to scan a grid of high-resolution images (1.3 microns per pixel) in smaller regions of interest. Both types of polarimetric equipment work in a similar manner. Monochrome green light passes through a linear polarizer combined with a quarter wave retarder with its fast-axis oriented at 45° with respect to the polarizer plane. This creates circularly polarized light that passes through the specimen followed by another polarizer, which captures the change in polarization induced by tissue birefringence. By simultaneous rotation of all polarizing filters, the transmission intensity is systematically imaged at discrete filter rotation angles. A typical PLI raw image dataset comprised 18 snapshots at equidistant azimuth angles of 10°, ranging between 0° and 170°. On the LAP these measurements were supplemented by 8° tilt measurements to the cardinal points allowing for stereoscopic imaging to disambiguate the fiber elevation angle. Processing of all raw images created, due to the coarse resolution of the LAP-particular anatomical landmarks of the trigeminal pathways in the brainstem, were scanned additionally in the PM to obtain images of higher resolution. The PM used a reversed light pathway and rotates the polarizer only. A more extensive description of the polarimeters engaged during PLI is given by (Reckfort et al. 2015;Axer et al. 2011a, b).

T2* weighted MR images
The trigeminal nerve can be clearly observed entering the ventrolateral part of the pons, after which the hypo-intense signal of the tt can be seen traversing the brainstem. The tt penetrates the middle cerebellar peduncle (mcp) and slopes slightly through the pons towards a large hypo-intense bulge medial to the mcp. This enlarged portion of the tt represents the PSN (Fig. 2). The PSN is located lateral and dorsal to the medial lemniscus (ml) and ventral to the central tegmental tract. At Fig. 2D, two hypo-intense extensions of the PSN, dorsal towards the fourth ventricle (most probably dtt) and ventral towards the center of the brainstem (vtt), can be discriminated.

TDI-results
Figure 3 presents an overview of the results of the TDI analysis of the three specimens in transverse sections confirming the findings of T2* weighted MRI. It can be observed that the tt traverses through the mcp in a dorsomedial and caudal direction. A portion of it appears to join the ml, whereas the rest remains on its dorsomedial course through the tegmentum. After the bifurcation of the tt, two tracts can be observed: (1) a dorsomedial tract that courses in craniocaudal direction and (2) a more ventral tract, dorsal to the ml. Figure 4 shows a ventral and a dorsal region of the brainstem as maximum intensity projection of multiple coronal sections near the bifurcation. Their location is shown in (A). The ventral region (C-E) shows that the majority of fibers from the tt join the ml crossing to the contralateral hemisphere. The dorsal region (B-D) reveals a smaller tract coursing towards the raphe without crossing. Instead it turns into a cranio-caudal direction. In principle, these models reconfirm a bifurcation of the tt near the transverse pontine fibers and the ml into a dorsal and ventral part. The dorsal tract appears to remain on the ipsilateral side with regard to the trigeminal root from which it sprouts, while the ventral one crosses to the contralateral side of the brainstem. The dorsal tract can, therefore, be determined to be the dtt; the ventral one to be the vtt.

Tractography results
Tractography results are depicted stereoscopically by use of MIP images. Results from tractography show a bifurcation of the left (blue) and right tt (red) in all three specimens. From this bifurcation, two tracts can be recognized: a dorsomedial one, which courses in cranio-caudal direction remaining ipsilateral, and a ventral one crossing to the contralateral side at the dorsal surface of the ml. The dorsal tract can, therefore, be identified as dtt, the ventral one as vtt (Fig. 5).

Polarized light images
On the PLI images (Fig. 6), the trigeminal root is observed to enter the ventrolateral aspect of the brainstem. The tt then crosses the mcp after which it bifurcates into a ventral and dorsal division. The ventral division, recognized as the vtt, bends alongside the ml. The dorsal division, recognized as the dtt, runs towards the tegmentum and the fourth ventricle (Fig. 6). Figure 7 depicts the tt coursing towards the dorsomedial aspect of the brainstem. After exiting the mcp, the tt shows a bifurcation into two tracts. The ventral branch runs posterior to the ml and courses towards the midline of the brainstem. The dorsal tract runs towards the dorsal aspect of the brainstem. The dorsal tract furthermore shows dispersing fibers that run towards the cerebellum via the mcp and towards the dorsomedial area of the brainstem.

Discussion
For the first time, the existence of a bifurcation of the tt into a vtt and dtt in the human brainstem using (d)MRI, TDI and PLI has been reported. It is well known that the orofacial nociceptive afferents synapse in the ipsilateral trigeminal sensory nuclear complex (TSNC) cross the midline and ascend as the trigeminothalamic tract to the contralateral ventroposterior medial nucleus (VPM) of the thalamus. This study shows that the crossing tract can be identified as the vtt. The dorsal branch of the bifurcation could represent the hypothesized ipsilateral dtt.

Comparison with existing studies
In primates, evidence of this bilateral somatosensory pathway through the thalamus to the facial representation of the primary somatosensory cortex has been previously provided (Rausell and Jones 1991a, b). In humans, this bilateral representation pattern has been observed in functional MRI (fMRI) studies that investigated the activation patterns of orofacial pain in humans. These studies showed that unilateral noxious stimulation of the orofacial region evokes changes in both the contralateral and the ipsilateral thalami, insular and primary somatosensory cortices (Jantsch et al. 2005;Nash et al. 2010). Other studies suggested that a bifurcation of the trigeminal nerve can be found at the level of the brainstem, based on the blood oxygen level dependent (BOLD) activation patterns in experimental orofacial pain settings (Nash et al. 2009(Nash et al. , 2010, which is consistent with the results of this paper. Furthermore, it is known from developmental brain research that the raphe of the brainstem is a generator of attractants and repellents that channel axonal growth to the proper side of the central nervous system, which causes the crossing of the major somatosensory and motor pathways. However, embryological development studies in mice have shown that a minority of the axons that sprout from the PSN do not cross the midline (Kivrak and Erzurumlu 2013). The vtt has been described to consist of fibers that sprout from the ventral part of the PSN and the cranial two-thirds of the spinal nucleus (SN). The axons of the second-order neurons in the aforementioned nuclei decussate along the dorsal border of the ml and are, therefore, called the trigeminal lemniscus (Torvik 1957;Smith 1975;Matsushita et al. 1982). The function of the vtt is thought to be the conduction of vital sensory information, including nociceptive information. This study shows that the vtt indeed runs posterior to the ml and some imaging techniques even show that there is no clear distinction between the ml and the vtt. However, the high-resolution PLI images show that fibers of the ml and fibers of the vtt do not entwine along their courses.
The dtt has been described to consist of fibers originating from the dorsal part of the PSN and the caudal two-thirds of the SN (Burton and Craig 1979;Matsushita et al. 1982). It has been assumed that the dtt is responsible for conducting pain, temperature, and mechanoreceptive stimuli from the anterior part of the facial region and the intraoral cavity [for a review see (Henssen et al. 2016)].
This anatomical study provides further evidence of the existence of both the dtt and vtt; however, it cannot verify that both tracts are involved in the conduction of noxious stimuli. It is well known that the primary afferents of the somatosensory system can be subdivided according to myelinization and size. The different fiber types cluster the afferents into group A-C, of which the unmyelinated C-fibers and thin, myelinated Aδ-fibers subserve nociceptive transmission, whereas B-fibers represent thin preganglionic fibers of the autonomic nervous system (Erlanger and Gasser 1937). Central white matter, on the other hand, cannot be allocated to nociceptive systems on the basis of its architecture alone. For this reason, although the dtt seems to be present in humans as well, its function cannot be determined by this study.

Technical considerations
The use of diffusion magnetic resonance imaging (dMRI) and tractography has become a popular method to study white matter pathways (Catani and Thiebaut de Schotten 2008). However, due to the limited imaging resolution of in vivo dMRI (typically 1.0-2.0 mm), its outcomes should be interpreted with care, especially in the presence of small fiber bundles. However, a recent study of Setsompop showed that on a clinical 3T scanner, imaging resolution of wholebrain data could be acquired of 0.66 mm and 0.76 mm isotropic. This indicates that new efficient methods that are capable of acquiring high-resolution diffusion data in vivo might be possible in the near future (Setsompop et al. 2018). Nevertheless, the applied MRI protocol in this study was found to further increase resolution to a spatial resolution of 0.2 mm isotropic for the T2*-images and 0.5 mm for the diffusion data. Although this resolution is higher than the resolution obtained in the study of Setsompop et al., it must be underlined that the present data was acquired on a preclinical scanning system, using 11.7T in an ex-vivo setting, whereas the setting of Setsompop et al. was a clinical 3T in vivo setting. One of the limitations of this study is formed by the fact that MR imaging parameters were not based on distribution of quantitative longitudinal relaxation time (T1) and transverse relaxation time (T2) and mean diffusivity (D) (Beaujoin et al. 2017(Beaujoin et al. , 2018. Therefore, corresponding histograms of T1, T2 and D values which could have been The vtt turns alongside the posterior wall of the ml and courses towards the raphe of the brainstem. The dtt runs towards the dorsal aspect of the brainstem splitting up again towards the cerebellum via the mcp (green-aquamarine) and towards the mesencephalic nucleus (purple). ml Medial lemniscus, tt trigeminal tract, vtt ventral trigeminothalamic tract, dtt dorsal trigeminothalamic tract used to optimize the imaging protocol, are lacking from this study. On the other hand, an adapted protocol that came forth from empirical investigation (Kleinnijenhuis et al. 2013) was applied and yielded the aforementioned results. Furthermore, it is known that tractography can produce plausiblelooking bundles in locations where the white matter tract does not actually exist, indicating that tractography produces substantial amounts of false-positive fiber bundles (Maier-Hein et al. 2017). To optimize the investigation of white matter microstructure, TDI was used as this technique uses reinterpolation of the quantitative maps at a finer resolution than the imaging resolution, which allows researchers to study white matter structure smaller than the imaging resolution (Calamante et al. 2010). As such, dMRI of ex vivo tissue allows for further evaluation of the tractography results by direct comparison against histological reference measures within the same specimen (Miller et al. 2011). Furthermore, complementation of the tractography data by use of PLI has been suggested to overcome the coarse sensitivity of diffusion tractography with respect to the identification of sharp bends, twists, and undulations (Foxley et al. 2016).
One of the limitations of this study is formed by the fact that the trigeminal tracts were not followed up to the thalamus due to the size of the coil of the 11.7T MR-scanner. Therefore, we cannot confirm that the dtt is connected to the thalamus as has been proposed in animal-based studies (Torvik 1957). Furthermore, the restricted number of specimens and the lack of other complementary neuroimaging techniques like optical coherence tomography (OCT) form another limitation of this study. OCT provides high-resolution fiber architecture data at approximately 1 micron spatial resolution. OCT is analogous to ultrasound techniques as it measures the backscattered light of a sample. Specifically, it is sensitive to variations of the refraction index in different types of tissue. For example, in 2014, Magnain and colleagues suggested that OCT can be used to generate accurate 3D reconstructions of histological layers in the cerebral cortex (Magnain et al. 2014). Furthermore, polarizationsensitive OCT is a special form of OCT that uses the intrinsic optical properties of back-scattering and birefringence of neural tissue. In this way, polarization-sensitive OCT has been used to image the fiber orientation and cyto-and myeloarchitecture of human nervous tissue and to visualize volumetric reconstructions of human brain tissue ). 3D-PLI, however, also provides the possibility to investigate three-dimensional fiber orientations in sectioned tissue. Each section is measured separately and basically provides a 3D vector field description of fiber orientations, forming the basis for tractography (Axer et al. 2011c). Finally, this study did not use tracers, whereas most of the knowledge on which we base our neuroanatomical insights regarding the trigeminal connections are derived from animal-based tracer studies [for a review see (Henssen et al. 2016)]. In 2013, Seehaus et al. used fluorescent, retrograde tracers to stain projection fibers in human tissue, showing that tracing studies can be performed in a post-mortem setting on adult human neural tissue. However, the tracer was injected into the subcortical white matter (approximately 500-1000 µm from the cortex) of the upper part of the temporal lobe. After 48 months, the tracer had traced a tract with a maximal length of 13 mm, which makes it unsuitable for human tracing studies. The same study, however, reported that DTI is capable of reflecting the shape and orientation of nerve pathways (Seehaus et al. 2013).
Future research should investigate the functionality of the vtt and the dtt to determine the clinical significance of functional entities like the TSNC with respect to different trigeminothalamic tracts.

Conclusion
This study shows that the tt bifurcates at the level of the human brainstem into two pathways which have previously been infrequently described in birds, reptiles, rodents, and non-human primates. These fiber tracts course in a ventral and dorsal direction and are, therefore, called the vtt and dtt, respectively. Connectivity, functionality, and clinical significance of these pathways must be determined in future studies.
Funding No funding was received for conducting this research.

Compliance with ethical standards
Conflict of interest The authors declare that they have had no conflict of interest in the conduction of this research.
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creat iveco mmons .org/licen ses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.