CB1 cannabinoid receptor enrichment in the ependymal region of the adult human spinal cord

Cannabinoids are involved in the regulation of neural stem cell biology and their receptors are expressed in the neurogenic niches of adult rodents. In the spinal cord of rats and mice, neural stem cells can be found in the ependymal region, surrounding the central canal, but there is evidence that this region is largely different in adult humans: lacks a patent canal and presents perivascular pseudorosettes, typically found in low grade ependymomas. Using Laser Capture Microdissection, Taqman gene expression assays and immunohistochemistry, we have studied the expression of endocannabinoid system components (receptors and enzymes) at the human spinal cord ependymal region. We observe that ependymal region is enriched in CB1 cannabinoid receptor, due to high CB1 expression in GFAP+ astrocytic domains. However, in human spinal cord levels that retain central canal patency we found ependymal cells with high CB1 expression, equivalent to the CB1HIGH cell subpopulation described in rodents. Our results support the existence of ependymal CB1HIGH cells across species, and may encourage further studies on this subpopulation, although only in cases when central canal is patent. In the adult human ependyma, which usually shows central canal absence, CB1 may play a different role by modulating astrocyte functions.


Results and Discussion
We found that human ependymal region consistently expresses CB 1 cannabinoid receptor (CNR1 gene; Table 1). CB 1 receptor could be the target of locally produced 2-AG, since we also found expression (although non enrichment) of enzymes related with 2-AG synthesis and degradation: diacylglycerol lipase α (DAGLA), diacylglycerol lipase β (DAGLB), monoacylglycerol lipase (MGLL) and abhydrolase domain-containing proteins -6 (ABHD6) and -12 (ABHD12). On the contrary, we could not find consistent expression of enzymes related with direct anandamide synthesis or degradation (NAPE-phospholipase D and fatty acid amide hydrolase, respectively). However, it should be noted that alternative enzymatic routes have been described for AEA, involving glycerophosphodiester phosphodiesterase and N-acylethalnolamine-hydrolyzing acid amidase that have been not tested here 2 . We also did not find expression of CB 2 cannabinoid receptor or the related GPR55 receptor. In previous works, we observed expression (but not enrichment) of PPAR-α , another cannabinoid-related receptor 1 , in human ependymal region 9 .
When compared with ventral horn, only CNR1 (CB 1 receptor) was significantly enriched at the ependymal region (Table 1). Accordingly, we found a strong CB 1 immunoreactivity in central gray matter by immunohistochemistry (Fig. 1B-J). But CB 1 enrichment in adult humans ependyma is not equivalent to that found in rodents: In humans, CB 1 is expressed by astrocytes, forming part of the gliosis that accompanies central canal closure (Fig. 1C-E) and in the GFAP + hypocellular ribbon of perivascular pseudorosettes ( Fig. 1F-K) 9,10 . CB 1 receptor is also expressed in astrocytes from other spinal cord areas (Fig. 2), and its intensity is apparently related to high GFAP expression. Accordingly, a strong CB 1 expression has been reported in reactive astrocytes of human pathologies like spinocerebellar ataxia 11 or temporal lobe epilepsia 12 . The role of astrocytic CB 1 could be multiple: protection 13 , metabolism increase 14 , control of inflammation [15][16][17] , inhibition of glutamate transporters 18 or release of neurotransmitters such as glutamate 19 , ATP and D-serine 20 .
Interestingly, we obtained some sections from adult individuals in which parts of the central canal were patent. In those sections, we found ependymal cells with high expression of CB 1 receptors lining the canal (Fig. 1L-N), resembling those CB 1 HIGH cells described for rats and mice 6 . These cells were mostly GFAP-, except for a very dim expression at the apical pole (Fig. 1N), in contrast with strongly GFAP + cells embeded in the ependymal layer (Fig. 1M).
Our results support the existence of ependymal CB 1 HIGH cells across species, and may encourage further studies on this subpopulation, although only in cases when there is central canal patency, i.e. childhood and upper cervical levels 8,9 . But in the majority of adult ependyma, CB 1 is enriched in astrocyte domains, and cannabinoids may play a different role, that still might be relevant, in terms of homeostasis maintenance and response to injury.

Methods
Human tissue was obtained from the HUFA BioBank (Alcorcon, Spain) and the HUB-ICO-IDIBELL BioBank (Hospitalet de Llobregat, Spain). Samples were obtained from donor individuals deceased without clinical or histopathological involvement of the spinal cord (    Gene expression in human ependymal region. All procedures were performed according to our published protocol 9 . Briefly, fresh frozen spinal cord blocks were cut in 25 μ m thick sections and the ependymal region microdissected with a Laser Dissection Microscope. RNA extraction, amplification and reverse transcription were performed as previously described 9 . We also collected microdissected portions of ventral horn, which we used as a non-neurogenic, non-ependymal reference for gene expression. Gene expression was studied with Taqman PCR Assays (Life Technologies, Madrid, Spain) either incorporated in Taqman Low Density Arrays (DAGLA, #Hs00391374_m1; DAGLB, #Hs00373700_m1; MGLL, #Hs00200752_m1; NAPEPLD, #Hs00419593_m1) or in individual assays (ABHD6, #Hs00977889_m1; ABHD12, #Hs01018047_m1; CNR1, #Hs01038522_s1; CNR2, #Hs00361490_m1; Fast Real-Time PCR System. Data were analysed as described 9 using automatic detection of Ct, normalized with the endogenous gene (Δ Ct vs 18S). Only genes expressed in at least three out of four samples were considered as consistently expressed and included in statistics. Enrichment was defined as higher and statistically significant expression in ependymal region vs ventral horn (Student's t-test with Δ Cts, p < 0.05). To obtain folds of enrichment, we used Relative quantity formula, RQ = 2^− Δ Δ Ct.

Immunohistochemistry.
To improve signal to noise ratio and avoid autofluorescence, we amplified CB 1 immunoreactivity using Tyramide Signal Amplification System (TSA Plus Cyanine 3 System #NEL744001KT, Perkin Elmer, USA). Free floating vibratome sections (40 μ m) were rinsed on 0.1 M phosphate-buffered saline containing 0.5% bovine serum albumin + 0.3% Triton X-100. Endogenous peroxidase inhibition and antigen demasking were performed as described 9 . Sections were then blocked with TSA Blocking Solution (45′) and incubated for 2 days with primary antibodies diluted in rinse solution + 10% Normal Donkey Serum: guinea pig anti-CB 1 (1:2000, #CB1-GP-Af530-1, FSI, Japan), rabbit anti-GFAP (1:2000, #Z0334, DAKO, Spain) and mouse anti-Vimentin (1:300, #M0725, DAKO, Spain). Immunoreactivity was visualized by incubating sections with Alexa 488-, Alexa 555-and Alexa 633-secondary antibodies (1:1000, Invitrogen, Spain) or horseradish peroxidase donkey anti-guinea pig antibody (1:300, Jackson Immunoresearch, UK) followed by Tyramide-Cy3 diluted in TSA Amplification Buffer (1:50). Samples were analyzed with a LEICA SP5 confocal microscope. We ruled out the interference of nonspecific staining by omitting primary antibodies. We set the confocal parameters at a point where no signal was observed in these primary antibody controls and those settings were used for all the image acquisitions (Supplementary Figure 1A-F). Furthermore, as discussed in several reports, there is a variety of antibodies against CB 1 receptor, and some of them may show non-specific staining [21][22][23] . The specificity of CB 1 antibody used for this report has been extensively validated by other laboratories and ourselves in previous works 6,24,25 . We show here an additional validation in the Supplementary Figure 1 by using immunohistochemistry and TSA amplification on wild type (C57BL/6N) and CB 1 knockout mice tissue (kindly donated by Dr. Galve-Roperh 26 ). Using restrictive confocal parameters (as we did for humans), we got rid out of autofluorescence, background staining and most of the non-specific staining observed in the knockout mice that, in these conditions, is limited to a dim intracellular neuronal staining, largely different from that observed in the wild type mice (Supplementary Figure 1L-Q). All post-capture image modifications were identically performed for controls, including cropping, noise reduction and minor adjustments to optimize contrast and brightness.
To quantitatively support CB 1 enrichment in the astrocytic area, we calculated the fraction of CB 1 found in GFAP + vs GFAP − areas on confocal planes (image size 190 μ m × 190 μ m) using Fiji (http:// pacific.mpi-cbg.de). For this, we outlined GFAP borders using manual Threshold with Otsu Filter and used this ROI on the CB 1 image corresponding to the same confocal plane. We measured CB 1 + Area inside and outside the selection (GFAP + and GFAP − areas, respectively) and expressed them as % of total CB 1 staining (Fig. 1K). We used Student T-test for statistical comparisons.