Understanding the Role of Dicer in Astrocyte Development

The Dicer1 allele is used to show that microRNAs (miRNAs) play important roles in astrocyte development and functions. While it is known that astrocytes that lack miRNAs are dysregulated, the in vivo phenotypes of these astrocytes are not well understood. In this study, we use Aldh1l1-EGFP transgene, a marker of astrocytes, to characterize mouse models with conditional Dicer1 ablation (via either human or mouse GFAP-Cre). This transgene revealed novel features of the defective astrocytes from the absence of miRNA. Although astrocyte miRNAs were depleted in both lines, we found histological and molecular differences in the Aldh1l1-EGFP cells between the two Cre lines. Aldh1l1-EGFP cells from hGFAP-Cre mutant lines displayed up-regulation of Aldh1l1-EGFP with increased proliferation and a genomic profile that acquired many features of wildtype primary astrocyte cultures. In the young mGFAP-Cre mutant lines we found that Aldh1l1-EGFP cells were disorganized and hyperproliferative in the developing cerebellum. Using the Aldh1l1-EGFP transgene, our work provides new insights into the roles of miRNAs in astrocyte development and the features of astrocytes in these two mouse models.


Introduction
Conditional Dicer1 alleles allow researchers to show the importance of miRNAs in developmental processes, including astrocyte development and function [1][2][3][4]. While studies have shown that astrocytes lacking miRNAs are dysregulated, the molecular changes that occur to these astrocytes are unclear. In this study, we use the Aldh1l1-EGFP transgene, a recently characterized marker for astrocytes, to characterize the changes to astrocytes in two different mouse models where mature miRNAs are ablated in astrocytes, via hGFAP-Cre or mGFAP-Cre.
MiRNAs are endogenous short hairpin non-coding RNAs that regulate the function and development of cellular processes by inhibiting the synthesis of gene products [5,6]. Dicer1 encodes a ribonuclease that cleaves miRNAs into their mature functioning form. Studies have used a conditional Dicer1 allele to show that the loss of miRNAs in neural precursor cells result in dysregulated brain development and functions [3,5,[7][8][9][10]. Although Dicer1 is also absent in astrocytes in these models, these studies focused on the effects of losing miRNAs on neuronal differentiation and survival and did not characterize the impact of miRNA depletion on astrocytes [3,7,[10][11][12]. When Dicer1 is ablated in astrocyte precursor cells, some studies have shown that staining of GFAP is altered [3,4,9]. The roles of miRNAs in astrocyte functions were further examined in another study using Cre transgenes that were expressed more specifically in astrocytes. In that study, the ablation of Dicer1 in astrocytes resulted in non-cell autonomous neurodegeneration in the cerebellum [1]. While that study indicated that astrocytes appeared immature at postnatal day 30 (P30), earlier developmental defects of the astrocytes were not assessed. Additionally, in both mouse models, many features of the astrocytes lacking mature miRNAs remain unknown.
Here, we utilized the Aldh1l1-EGFP transgene, a pan-astrocyte marker, to characterize the morphological and molecular phenotypes of astrocytes in the absence of Dicer1 [13,14]. We assessed Aldh1l1-EGFP cells in two different mouse models where Dicer1 was ablated by astrocyte Cre lines. One Cre line expressed before and the other line expressed after astrogliogenesis. We found that Aldh1l1-EGFP cells exhibited distinct dysregulated features. Forebrain Aldh1-l1-EGFP cells in the mouse model where Dicer1 was ablated early (hGFAP-Cre) had features of immature astrocytes and primary astrocytes, whereas forebrain Aldh1l1-EGFP cells in the mouse model where Dicer1 was ablated later (mGFAP-Cre) did not have obvious defects during development. As previously reported, astrocytes had dysregulation in the developing cerebellum in the mice generated from mGFAP-Cre. In using the Aldh1l1-EGFP transgene, we found additional defects of the astrocytes in the mGFAP-Cre model at an earlier time frame than previously described [1]. The use of Aldh1l1-EGFP transgene allowed us to identify several novel features of astrocytes in mouse models where miRNAs are ablated from astrocytes.

Materials and Methods Mice
BAC Aldh1l1-EGFP transgene were generated by GENSAT. hGFAP-Cre and mGFAP-Cre (line 77.6) lines were obtained from the Jackson laboratory. Mice with conditional Dicer1 allele were obtained from the McManus lab at UCSF [15]. These conditional Dicer1 alleles contained lox sites flanking exon 23 which is excised in the presence of Cre. This exon encodes most of the second RNaseIII domain, necessary to convert precursor miRNAs into mature forms when inactivated [15]. mGFAP-Cre Dicer1 and hGFAP-Cre Dicer1 experiments were conducted in C57/B6 background and mixed FVB and C57/B6 backgrounds, respectively. Animals of both sexes were used. The care and treatment of the animals were conducted under the strict guidelines of the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. Animals were anesthetized with 2.5% avertin solution according to guidelines. During euthanasia, mice were first exposed to carbon dioxide followed by cervical dislocation. The protocol was approved by the Institutional Animal Care and Use Committee of the University of California San Francisco (Approval Number: AN089663-01E). Primers for animal genotypes include the following: hGFAP-CreF-ACTCCTTCATAAAGCCCT hGFAP-CreR-ATCACTCGTTGCATCGACCG EGFPForward-CCTACGGCGTGCAGTGCTTCAGC EGFPReverse-CGGCGAGCTGCACGCTGCCGTCCTC Dicer1F-CCTGACAGTGACGGTCCAAAG Dicer1R-CATGACTCTTCAACTCAAACT wt: 351 bp; floxed allele: 420 bp mGFAP-CreF-CCGGGCTGCCACGACCAA mGFAP-CreR-GGCGCGGCAACACCATTTTT

Immunohistochemistry
To obtain tissues for immunohistochemistry, animals were deeply anesthetized and perfused with 4% PFA in PBS. After dissection, tissues were incubated in fixative for two days and then transferred to 30% sucrose (PBS) solution for another two days. After two days of incubation in 30% sucrose incubation, fixed tissues were frozen in OCT media and sliced to 20 μm sections. For staining, samples were permeabilized in ice cold acetone for 10 minutes. The samples were then washed with PBS before being blocked for 1 hour in room temperature with blocking buffer (5% normal goat serum, NGS, and 0.01% triton-X in PBS). Samples were stained with primary antibodies overnight at 4°C and stained with secondary antibodies for two hours at room temperature the following day. Primary antibodies used were GFAP (Dako, 1/500) and ki67 (Thermo, 1/400). Anti-rabbit secondary fluorochromes conjugated with Cy3 (Jackson Immu-noResearch) were used as secondaries. Slides were mounted with Vectashield (Vector Laboratories). Visualization of the samples was performed with either Leica DMI 6000B or Zeiss Axio imager M1 microscopes. Images were processed with ImageJ (NIH) and Zen (Zeiss) softwares.

Primary astrocyte cultures
Primary astrocytes were obtained from P2 forebrain. Preparation was performed as described in Cahoy et al. 2008 [16]. Primary cells were used after 2 weeks in vitro with a passage at 1 week. For FACS isolation, cells were detached using. 25% Trypsin-EDTA (Life Technologies), then resuspended in Leibovitz L-15 media (Life Technologies) before FACS. After sorting, Trizol (Life Technologies) was added to the cells for subsequent RNA isolation.

Counting of cells in micrograph images
Counting of cells was performed using ImageJ (NIH). For the brains, 20 μm sections were used and images of cerebral cortex were taken with Leica DMI 6000B. Using ImageJ, areas of 500 μm 2 unit was marked and numbers of positive cells were counted. The significance levels were calculated using Student's t test, unpaired. Graphs were made using Prism software (GraphPad Software). For mGFAP:CKO cerebellum, ki67 positive cells were counted if they localized between Bergmann glia and pial surface of the cerebellum.

Cellular preparation of in vivo Aldh1l1-EGFP positive cells for FACS
Isolation procedure of the EGFP cells was adapted from Cahoy et al. While we followed the published protocol, we modified the incubation time depending on the ages of the forebrains. P7 forebrains were digested for 25 minutes while P16 and older samples were digested for 80 minutes in papain based digestion media.

Flow cytometry
EGFP positive cells were sorted as described in Molofsky et al [17]. Arias III (BD Bioscience) machine was used to perform the sorting. For each experiment, samples lacking Aldhl1-EGFP were used as negative control. EGFP positive cells were gated on forward/side scatter, DAPI (Life Technologies) exclusion for live cells, and EGFP signals. Samples were sorted into PBS containing 3% FBS. All samples were re-sorted. After collection of data, samples were analyzed using FlowJo (Tree Star). Once the samples were isolated, the cells were centrifuged for 5 min at 1000 g. The cell pellets were then resuspended in Trizol (Life Technologies) for RNA isolation.
To compare the EGFP intensity of the Aldh1l1-EGFP cells, comparison of the wildtype and mutant cells were always performed on the same day. To determine the differences in number of Aldh1l1-EGFP positive cells, the percentage of EGFP positive cells were determined only from live cell population that were gated for forward/side scatter and excluded DAPI. Representative examples of gating strategies are in S2 Fig, S7 Fig, S9 Fig, S10 Fig, and S11 Fig. We used similar gating strategies for animals of the same age in both hGFAP:CKO and mGFAP: CKO models.

RNA isolation
Total RNA was isolated using Trizol (Life Technologies) following manufacturer's protocol. For FACS samples, roughly 10,000 cells were used to isolate total RNA. In the small quantity samples, 250 μg/ml of glycogen was added as a carrier. Samples were precipitated in isopropanol at -20°C for at least two hours. RNA was reconstituted in 50 μl of RNase/DNase free water. Subsequently, the RNA was cleaned with RNA Clean and Concentrator-5 columns (Zymo) and treated with DNase (Life Technologies) in columns following Zymo protocol. 11 μl of RNase/DNase free water was used to elute the RNA. Total RNA was quantified using Agilent Pico Chip (Agilent).

Microarray
Microarray experiments, which used 2 ng of total RNA, were performed according to manufacturer's protocol. Samples were amplified using NuGEN Pico V2 (NuGEN) kit and hybridized onto Affymetrix Mouse Gene 1ST chip.
For the generation of the fold change, normalization of the microarray data, gene ontology list, and statistical significance of fold changes, we used the ArrayStar (DNASTAR) analysis program. Normalization was performed using Robust Multi-array Average (RMA) algorithms. Data from 3 samples each of wildtype P7 Aldh1l1-EGFP cells, hGFAP:CKO P7 Aldh1l1-EGFP cells and wildtype primary Aldh1l1-EGFP cells were normalized together. Significance levels were determined by moderated t-test [18]. Given the small sample size that was used, the moderated ttest is better suited at reducing the background noise of the data [18]. False-discovery-rates (FDR) was determined using Benjamini-Hochberg test [19]. Microarray data are available in the ArrayExpress database (www.ebi.ac.uk/arrayexpress) under accession number E-MTAB-2995.
To identify upstream miRNA regulators, we used Ingenuity Pathway Analysis (IPA) to analyze all genes that were differentially up-regulated in mutant samples. As a negative control in the analysis, we used differentially down-regulated genes in the mutant samples. IPA uses both experimentally verified and highly confident prediction data for the analysis. The significance levels were calculated using Fisher's exact tests.
Gene ontology was performed using ArrayStar (DNASTAR). Grouping was performed on genes that had 2 folds difference between wildtype and hGFAP:CKO samples and have p-values less than 0.05.

Real time qPCR
Total RNA was converted to cDNA using the SuperScript Kit (Life Technologies). For FACSsorted cells with low concentrations, we used 5 μl of total RNA to convert into cDNA. Gapdh was used for normalizing the samples. Real time qPCR reactions were performed on ABI 3770 using SYBR Green Master Mix (Life Technologies). The significance levels were calculated using unpaired Student's t test.
Dicer1Forward-ACCAGCGCTTAGAATTCCTGGGAG Dicer1Reverse-GCTCAGAGTCCATTCCTTGC Dicer1 qPCR primers are located in exon 23 of Dicer1 thus in the event of recombination, the amplicons will not be made. GapdhF-TTGATGGCAACAATCTCCAC

Aldh1l1-EGFP positive cells are dysregulated in hGFAP-Cre Dicer1/ Dicer1 mutant forebrain
Previous studies reported changes in astrocytes when ablation of Dicer1 occurs before astrogliogenesis. Studies have found increased GFAP staining in these models while other found a decrease [3,4]. Here, we used the Aldh1l1-EGFP transgene to further explore these astrocytes in models where miRNA ablation occurs before astrogliogenesis. In these reporter mice, EGFP proteins are activated by the Aldh1l1 promoter, which resulted in the staining of astrocyte cell bodies. While some EGFP may localize to the processes of the astrocytes, the majority of the signals stay in the cell bodies. We combined Cre expressed under the human GFAP promoter with homozygous floxed Dicer1 mice to generate astrocyte conditional knockout mice (hGFAP:CKO). The hGFAP-Cre transgene is activated in radial glial cells at E14.5 before the onset of astrogliogenesis in the forebrain [20]. Similar to other models where activation of Cre occurs in radial cells during neurogenesis, hGFAP:CKO mutant mice had smaller forebrains (Fig 1a and 1b) [3,10,11,21,22]. The mutant mice were ataxic and died between postnatal days 18 to 26. In these brains, a large number of Aldh1l1-EGFP positive cells were present in the forebrains and spinal cords of P16 mice (Fig 1c-1h, additional images of forebrain Aldh1-l1-EGFP positive cells are shown in S1 Fig). This result indicates that astrogliogenesis can occur in the absence of mature miRNAs. Thus our data agreed with reports that found an increase in astrocytes when mature miRNAs were ablated during neurogenesis. However, while there were abundant Aldh1l1-EGFP cells, both histological data and FACS analysis indicated that hGFAP:CKO Aldh1l1-EGFP cells had increased EGFP signals in all regions of the central nervous system (Fig 1c-1i and S2 Fig). We further assessed Aldh1l1-EGFP cells in the hGFAP:CKO mutant forebrains and found the presence of more Aldh1l1-EGFP positive cells in hGFAP:CKO mutant than in wildtype forebrains. Aldh1l1-EGFP cell numbers in hGFAP:CKO mutant forebrains (P16 wildtype (WT) = 122.2 ±19.2 n = 3, P16 hGFAP:CKO = 255.4±44.1 n = 3, p<0.05 per 500 μm 2 unit, t test, unpaired, S3 Fig) were greater than in wildtype forebrains. In the cortex, astrocytes divide locally until around P7 [23]. We focused our investigation in P7 forebrains where astrocytes are still dividing to determine if proliferation was different. At this stage, hGFAP:CKO forebrain Aldh1l1-EGFP cells had increased proliferation compared to wildtype mice (13.6% ± 5% co-localization of ki67/EGFP for hGFAP:CKO vs 4% ± 1% ki67/EGFP for WT, n = 3 per genotype, p < 0.05, by t-test, unpaired, Evaluation of Aldh1l1-EGFP cell phenotype in mGFAP-Cre Dicer1/ Dicer1 mutant forebrain While the hGFAP:CKO mutants died prematurely and had smaller brains, animals appeared normal during development when we ablated the floxed Dicer1 alleles with Cre expressed under the mGfap promoter (mGFAP:CKO, line 77.6). These animals then developed ataxia and seizure at 6 weeks of age and also died prematurely at approximately 2 months [1]. In these lines, Cre is expressed early in postnatal life but after the occurrence of astrogliogenesis. This differs from human GFAP promoter, which is activated in radial glial cells during embryonic development and results in a deletion of Dicer1 more exclusive to astrocytes.
Next we assessed these mutants to determine if Aldh1l1-EGFP positive cells also had up-regulated signals. Histological sections from mGFAP:CKO mutant did not exhibit noticeable upregulation of EGFP signals in P16 forebrains (Fig 2a-2c). FACS analysis also did not show increased levels of fluorescence in Aldh1l1-EGFP positive cells (Fig 2d and S7 Fig). mGFAP:CKO mutants began to develop ataxia at 6 weeks and displayed astrogliosis which can be identified using GFAP antibodies (S8 Fig). In two-month old wildtype forebrains, EGFP signals were low and difficult to distinguish from background, whereas mutant cells had higher and more distinct EGFP signals (Fig 2e-2h and S9 Fig). Thus, while developing mGFAP:CKO animals did not have a distinct difference in forebrain Aldh1l1-EGFP signals, symptomatic mGFAP:CKO animals displayed up-regulated EGFP phenotype similar to hGFAP:CKO mutants.

Aldh1l1-EGFP cells are disorganized in the cerebellums of developing mGFAP:CKO mice
Although mGFAP:CKO mutant cerebellums appear to develop normally, neurodegeneration in the cerebellum occurs as these animals age. Histological examination of two-month-old symptomatic animals indicated cellular loss and disorganization of Aldh1l1-EGFP positive cells in the cerebellum (Fig 3a-3d). Additionally, real-time qPCR analysis of Aldh1l1 levels indicated up-regulation in the mutant cerebellum (Fig 3e). During development, mGFAP:CKO mice did not exhibit ataxic behavior and gross features of their cerebellum appeared normal. However, we found abnormal localization of Aldh1l1-EGFP positive cells in P16 mGFAP:CKO cerebellums (Fig 3f-3h). While the molecular layers of wildtype cerebellums exhibited no clear Aldh1l1-EGFP cell bodies, mGFAP:CKO cerebellum molecular layers contained Aldh1l1-EGFP positive cells (Fig 3f-3h). EGFP signals also appeared elevated in the mutant Bergman glial cells (Fig 3f-3h), which is reflected in up-regulation of Aldh1l1 transcript levels in the cerebellums of P16 mGFAP:CKO mice (Fig 3o). Additionally, we found small but significant up-regulation and down-regulation of Gfap and Glast, respectively (Fig 3o). Taken together, these data indicate that mGFAP:CKO cerebellums are dysregulated by P16.
Previous studies by Tao et. al. suggested that mGFAP:CKO cerebellum astrocytes are less differentiated [1]. We therefore assessed if Aldh1l1-EGFP positive cells have increased proliferation during development. Using ki67 antibodies as a marker for proliferation, we found that mGFAP:CKO cerebellums had increased number of ki67 positive cells throughout the molecular layer, while wildtype P16 cerebellum did not have any measurable dividing cells in the molecular layer (Fig 3i-3k). We found that 87.3% ± 1.5% of these ki67 positive cells in the molecular layer co-localized with Aldh1l1-EGFP positive cells (out of 27.5 ± 2.12 ki67 cells in molecular layer, n = 2 mutants) (Fig 3l-3n). Thus in the absence of mature miRNAs in mGFAP:CKO cerebellum, Aldh1l1-EGFP in the molecular layer are actively proliferating at P16.
The expression profile of hGFAP:CKO Aldh1l1-EGFP cells displays immature and primary culture astrocyte phenotypes The introduction of the Aldh1l1-EGFP transgene into these mice allowed us to isolate and analyze the molecular features of Aldh1l1-EGFP positive cells from young forebrains using FACS. Given the change in EGFP signals of the P7 hGFAP:CKO forebrains, we isolated Aldh1-l1-EGFP positive cells from these mutants to perform further molecular characterization (S10 Fig). We obtained live cells that were 95-98% Aldh1l1-EGFP-positive after double sorting. Using qRT-PCR, we found down-regulation of exon 23 in Dicer1 mRNA, which indicated that recombination occurred in these cells (Fig 4). Evaluation with qRT-PCR of known astrocyte genes in mutant cells showed that Aqp4 and Glast were significantly down-regulated in mutant cells, suggesting that these cells were deficient in expressing genes involved in proper astrocyte functions (Fig 4, p-values for Aqp4 and Glast <0.05, t test, unpaired). Gfap expression was not significantly down-regulated, which suggested that these cells already acquired astrocyte identity and were in the process of maturing (Fig 4, p-value for Gfap was 0.09, t test, unpaired). We also found Fgfr3 transcript levels to be similar between hGFAP:CKO and wildtype cells, which suggests that acquisition of astrocyte identity occurred (Fig 4). Nes, a developmental marker, had a small significant increase, indicating immature phenotype in these mutant cells (Fig 4) [ 16,24]. Overall, the down-regulation of Aqp4 and up-regulation of Nes suggest that mutant Aldh1l1-EGFP cells exhibit an immature molecular signature [16]. To identify other features of the dysregulated Aldh1l1-EGFP cells, we performed microarray expression profiling of FACS isolated Aldh1l1-EGFP positive cells. Expression profiling was performed using RNA of Aldh1l1-EGFP positive cells isolated from P7 wildtype and hGFAP: CKO forebrains (n = 3 per genotype) and the Gene 1.0ST microarray chip. S1 Table contains the normalized expression data with corresponding p-values. In mutant cells, many genes associated with mature astrocytes were down-regulated and genes associated with immature astrocytes were up-regulated. For example, Aqp4 and Slc1a2 (-2.22, FDR = .08 and -2.12, FDR = .05, respectively, S1 Table) were down-regulated. Additionally genes associated with immature phenotypes such as Nes and Vim were up-regulated (1.78, FDR = .07 and 2.34, FDR = .05, respectively, S1 Table) [24]. We also analyzed the microarray data on genes involved in cell division to determine if they aligned with the histology data. Although we found that several genes involved in cell division such as Ccna1, Ccnb1, Ccnb2, and Cdkn1a had increased levels, their expression changes were not significant and did not indicate active proliferation. One possible explanation for this difference between the expression data and histological data may be that only a small subset of these mutant cells are proliferating. Our localization data with ki67 showed that only 13% of Aldh1l1-EGFP cells co-localized with this marker, thus the expression profile may not be sensitive enough to capture the change.
We used gene ontology to uncover additional features that were disrupted in hGFAP:CKO mutant Aldh1l1-EGFP positive cells. We found that differentially regulated genes (greater than or equal to 2-fold differences with p-value <0.05) in mutant Aldh1l1-EGFP positive cells fall into metabolic pathways such as cholesterol and lipid synthesis pathways (Table 1). While direct assessment of cholesterol level of the brain will need to be assessed, the change in expression of this of group of genes may indicate that the cholesterol levels of brain are altered [25].
We noticed that many of the down-regulated genes in hGFAP:CKO astrocytes were also down-regulated when we compared the FACS P7 wildtype cells to primary culture expression [16]. To investigate this further, we generated expression data of forebrain primary astrocytes from wildtype mice that were grown in culture for two weeks. Additionally, we sought to limit the heterogeneity by isolating Aldh1l1-EGFP positive cells in these cultures. At two weeks, primary astrocyte cultures had 50-75% Aldh1l1-EGFP positive cells (S11 Fig). When hGFAP: CKO cells and primary Aldh1l1-EGFP cells were compared to wildtype Aldh1l1-EGFP FACS cells, we found that hGFAP:CKO and primary culture Aldh1l1-EGFP cells shared many of the same differences from wildtype FACS isolated Aldh1l1-EGFP cells (Tables 2 and 3). 76% of the genes that were 2-fold down-regulated between hGFAP:CKO and wildtype FACS Aldh1-l1-EGFP cells were also down-regulated in primary astrocytes ( Table 2, column 2 and 4). On the other hand, 61% of the genes up-regulated in hGFAP:CKO Aldh1l1-EGFP cells were also up-regulated in the primary astrocyte cultures when compared to wildtype Aldh1l1-EGFP cells ( Table 3, column 2 and 4). These data suggest that P7 hGFAP:CKO astrocytes acquired many of the molecular signatures of primary astrocyte cultures.

Identification of miRNAs expressed in P7 Aldh1l1-EGFP cells and their potential targets
To better understand how miRNAs may regulate astrocyte development and function, we identified the miRNAs that were expressed in P7 forebrain Aldh1l1-EGFP cells using Taqman Low Density Arrays. Additionally, we used Ingenuity Pathway Analysis (IPA) software to evaluate The Role of Dicer in Astrocyte Development potential upstream regulator of the microarray data. Using the Taqman Low Density Arrays, we found that Aldh1l1-EGFP cells expressed many miRNAs (S2 Table). Several of the miRNAs on the list from P7 wildtype forebrain astrocytes such as miR-21, miR-223, miR-146a and miR-181 (S2 Table) have been previously shown to regulate astrocyte functions [26][27][28][29][30]. However, the roles of most of these miRNAs in astrocytes are not known. We used IPA to determine whether The Role of Dicer in Astrocyte Development genes with increased expression levels in hGFAP:CKO Aldh1l1-EGFP cells are targets of these miRNAs (S2 Table). Given that the effect of miRNAs on a gene may be subtle, we included all up-regulated genes for this analysis. As a negative control for this analysis, we also performed this IPA analysis on down-regulated genes. IPA identified various upstream regulators for many of these genes including miRNAs. While IPA identified several miRNAs expressed in astrocytes with the up-regulated set of genes, the data set only predicted two miRNAs with the down-regulated genes (S3 Table). The potential targets of miRNAs expressed by Aldh1l1-EGFP positive cells are listed in Table 4 (miRNAs predicted to be upstream regulator but did not have expression in Aldh1l1-EGFP cells are not shown on the list). While the relationship of these miRNAs and their potential targets will need to be further evaluated, the IPA analysis was able to identify two miRNAs, miR-181a and miR-125b that have been previously reported to regulate astrocytes [28,31,32]. Interestingly, we found that many of the genes are shared potential targets of different miRNAs in this analysis (Table 4), which suggests that different Aldh1l1-EGFP miRNAs may have redundant targets to ensure that a particular gene is down-regulated. Further experimental analysis of these miRNAs individually will be necessary to determine if they share similar targets and to determine the mechanism in which they regulate astrocyte maturation and functions.

Discussion
While other studies mainly used GFAP antibodies to understand the role of miRNAs in astrocyte development, here we used a pan-astrocyte marker, Aldh1l1-EGFP, to characterize novel changes of astrocytes in two mouse Dicer1 models. Our study found that the ablation of Dicer1 with hGFAP-Cre or mGFAP-Cre produced different phenotypes in astrocytes as assessed from the features of Aldh1l1-EGFP cells.
In this study, we found that Aldh1l1-EGFP positive cells in young hGFAP:CKO forebrains displayed up-regulated EGFP signals and molecular phenotypes of immature and primary astrocytes, while Aldh1l1-EGFP positive cells appeared normal in young mGFAP:CKO forebrains. One possible reason for this difference may be that miRNAs affect astrocytes differently during different developmental time points similar to how neurons are affected differently depending on the timing of the ablation of Dicer1 [8]. Cre is activated prior to astrogliogenesis in the hGFAP:CKO line, whereas it is expressed after astrogliogenesis in the mGFAP:CKO line. Another reason may be that the hGFAP:CKO astrocytes are in a reactive state since brain malformations occur in the hGFAP:CKO line. Aldh1l1 is known to be up-regulated in reactive astrocytes [14]. mGFAP:CKO Aldh1l1-EGFP cells also had a more dramatic increase in EGFP signals during symptomatic disease state, further suggesting that the reactive state of astrocytes may be contributing to the phenotype that we see in the hGFAP:CKO mutants. The molecular profile that we obtained from hGFAP:CKO Aldh1l1-EGFP cells contained some genes that have been implicated in the profile of reactive astrocytes [33]. For example, one of the most highly expressed genes in the reactive astrocyte data was Cp, which was also highly up-regulated in the hGFAP:CKO mutant samples. However, other highly expressed genes in reactive astrocytes such as Lcn did not appear in our database. Gfap was also not up-regulated in hGFAP:CKO mutant animals. Reactive astrocytes are now known to be heterogeneous depending on the context of the brain injury [33,34]. Therefore the possibility that the hGFAP:CKO Aldh1l1-EGFP cells are a type of previously unknown reactive astrocyte cannot be ruled out. Thus the phenotype seen in hGFAP:CKO Aldh1l1-EGFP cells may be the result of both the cells being in a reactive state and the loss of mature miRNAs.
Using the Aldh1l1-EGFP transgene, we did not notice gross changes of astrocytes in the mGFAP:CKO forebrain at P16. The Aldh1l1-EGFP transgene did allow us to observe changes to Aldh1l1-EGFP cells in mGFAP:CKO cerebellum at P16. Previous study did not notice differences in P15 mGFAP:CKO cerebellum. We found that Aldh1l1-EGFP cells were localized in the molecular layer of mutant P16 cerebellum while wildtype Aldh1l1-EGFP cells were not. Additionally, these Aldh1l1-EGFP cells in the molecular layers were proliferating. At this stage, we also found small but significant up-regulation of Gfap. Thus our data indicate that dysregulated features are present in P16 mGFAP:CKO mice.
Our genomic approach allowed us to further characterize astrocytes in hGFAP:CKO mice as well as identifying several interesting features. One interesting aspect was that the cholesterol gene pathway was down-regulated in mutant cells. While additional research will be required to assess if cholesterol levels are affected in mutant brains, this gene expression signature is potentially interesting for future studies since cholesterols from astrocytes have been shown to be involved in the formation of synapse in neurons [25].
Currently, the in vivo relevance of primary astrocytes is not well understood, although it has been hypothesized that primary cells may be more similar to reactive astrocytes [16,33]. Here we were able to discover another situation where in vivo cells resembled many of the characteristics of primary astrocytes.
In conclusion, our data provide new insights into the phenotypes of astrocytes in mouse models where Dicer1 is deleted with different astrocyte Cre lines. We used the Aldh1l1-EGFP transgene to describe previously unknown aspects of astrocytes in these mutants. Additionally, this study showed that Aldh1l1-EGFP transgene can be a useful marker in understanding asymptomatic developing phases of mGFAP:CKO neurodegenerative disease model and potentially in understanding the roles of developmental astrocytes in other neurodegenerative diseases.