Intra-arterially delivered MSCs was non-invasively and dynamically investigated by 3.0 T MRI after labeling with Fe3O4-DMSA-PLL in the present study. We showed that MRI could visualize the cells as soon as one hour after IA injection and was effective for tracking grafted MSCs within four weeks.
We choose beagle dog for IA transplantation because it is structurally similar to the human brain, has little variation in the size of the brain specimens and in the diameters of intracranial arteries despite the range of weight and technically easy for transfemoral catheterization of carotid artery 
. In our previous study 
, an embolic stroke model resembling lacunar infarction was got by proximal MCA occlusion in beagle dogs. In order to enlarge the cerebral infarction, proximal MCA occlusion in combination with temporal occlusion of the ipsilateral ICA which had been reported previously was used in our current study 
. Collateral supply was restricted by temporary ICA occlusion. Cerebral blood flow in the MCA territory was further reduced into the ischemic range which gave rise to the increase of infarction volume compared with single-thrombus approach. However, due to the variability of collateral circulation in dogs, heterogeneous infarction was observed in our study.
Results of labeling MSCs in vitro with home-synthesized SPIO demonstrated its usefulness to label MSCs without cell viability compromise which was compatible with widely used ferumoxides-PLL complex 
. Moreover, such labeled MSCs could be detected in vivo efficiently after IA transplantation by 3.0 T MRI. In agreement with previous studies 
, IA administration could produce a large number of MSCs in the target brain. The injected cells mainly distributed in the ipsilateral cerebral hemisphere of infarction while only few MSCs were found in the contralateral side.
However, no MSC was shown in dogs with occluded MCA at the time of transplantation. We considered two reasons for this. First, the main approach for MSCs entering brain may be the ipsilateral MCA. Thus, the flow status of that MCA before transplantation may play an important role in distribution and the amount of cells in the target tissue. Fewer MSCs detected in group B than group A could also confirm our speculation. Second, we considered that there should be a small amount of MSCs in the brain which was delivered to the host brain through other access, like anterior cerebral artery, as confirmed by PB staining in group C. However, our imaging methods may not be sensitive enough to detect small amount cells or tiny cell clusters.
In the present study, labeled MSCs were transplanted at one week after onset of ischemia. Strbian et al considered the blood-brain barrier (BBB) was continuously open for several weeks after focal cerebral ischemia 
. Komatsu et al also found that BBB may largely break for at least two weeks after cerebral infarction and remain insufficient even four weeks 
. Thus, we think that deficit BBB after cerebral ischemia may account for the mechanical trapping of MSCs from vessel to brain parenchyma as well as the possible cells immigration after transplantation.
The gradually fading of low signal intensity in the host brain was not only due to the cell division but also biodegradation and entry of iron into metabolic pathways 
. Even so, SPIO-labeled MSCs will retain the iron nanoparticles to a sufficient degree to produce hypointensity within four weeks on SWI and T2*
ratio could reflect the amount of grafted cells in the brain. But it is still difficult to directly extrapolate the exact number due to heterogeneity in the structure and composition of brain tissue and differences in MRI sequence and parameters 
The slower fading of hypointensity in PI area than in INP may support the concept that chemoattractive factors may be released in the site of brain lesion which gave rise to reservation and specific homing of transplanted cells to the PI area 
. PB staining confirmed that there were MSCs in the PI area.
Our results also showed that a smaller infarction on the day of cell transplantation seemed to be associated with fewer cells in the brain after IA delivery while dog with large infarction had more engrafted cells. But due to the small number of dogs, we didn’t perform statistical analyze. Li et al reported animal with smaller lesions (less than 10% of brain volume) at the time of transplantation have fewer grafted cells into the parenchyma 
. Our finding was consistent with theirs. The possible reason may be due to the following two reasons: First, small infarction was associated with mild BBB deficit which led to fewer mechanically trapped MSCs. Second, recruitment forces such as chemoattractive factors released by small infarction may not enough to attract lots of MSCs whereas these chemoattractant forces and possible other factors like brain edema caused excessive cell accumulation in the large infarction group 
Although IA administration showed early arrival and more transplanted MSCs in the target brain, there were still some risks. New infarction developed within 24 hours in two dogs. The likely cause was microembolism caused by transplanted MSCs which in turn led to local impeded cerebral blood flow 
.Besides, exogenous labeled-MSCs may cause epilepsy after transplantation which should be paid more attention in the future research.
There were some limitations of our study. First, the numbers of dogs with large and small cerebral infarction in group A were small, limiting the investigation of the relationship between initial ischemic volume and the amount of grafted cells in the brain. Second, the amount of grafted cells in the brain was analyzed by visual evaluation and T2* ratio. T2* value was measured by manually drawn ROIs which may lead to some bias. Third, MRI sensitivity of detecting labeled MSCs was affected by many factors, such as MRI protocol and software. Our imaging methods may not be sensitive enough for small amount of engrafted cells or tiny cell clusters. Finally, more animals were needed to confirm the reproducibility of our study.
In conclusion, it is feasible to transplant MSCs through IA route for cerebral infarction in a canine model. Successful IA administration showed diffuse distribution pattern, and large amounts of transplanted MSCs in the target brain. Both the flow status of ipsilateral MCA and infarction volume before transplantation may play an important role in the amount of grafted cells in the brain. In vivo MR imaging is useful to track SPIO-labeled MSCs for at least four weeks. However, more attention should be paid on the safety of IA approach considering the high ratio of adverse consequences.