.5-bromo-2’deoxyuridine (BrdU) is a nucleoside analog that is also commonly used for marking cell proliferation, but detection of the BrdU requires strong acids for DNA denaturation to reveal the antigen for antibody targeting, which can affect the integrity of dsDNA and other cellular structures that may be desired for DAPI-labelling or IHF. Antibody-based detection may also have background fluorescence and be variable between experiments or within the same tissue (Chehrehasa et al., 2009). One problem with both methods is that as cells continue to proliferate, the modified nucleosides are divided between progeny. This may result in variable dilution of the fluorescent output of the labelled cells, depending on the number of divisions before terminal differentiation. Therefore, glomerular neurons derived from neuroblasts dividing at P1 may label more brightly in the olfactory bulb than those from dividing SVZ astrocyte progenitor cells.
Make this 1st paragraph of the section. Also insert a lead-in sentence: “To determine, …with 517, … we performed..in order to” P1 was chosen as the EdU injection date to label VZ-SVZ tissue at the same developmental stage as prior experiments from de Chevigny et al (2012). They observed an increase in dopaminergic PGCs in the olfactory bulb 15 days after Pax6 overexpression and miR-7 knockdown experiments on P1 VZ-SVZ. In contrast to de Chevigny however, we allowed 60 days post-injection before harvesting the olfactory bulb. At adult stages, when BrdU is injected into mice at P60, BrdU+NeuN+ cells in the glomerular layer peak between 1 to 2 months later (Winner et al., 2002). This time period was also suitable because the majority of TH+ PGCs and CalR+ PGCs are thought to be generated between birth and adulthood (McLean and Shipley, 1988; De Marchis et al., 2007). Following BrdU injection into adult mice SVZ, the quantity of TH+BrdU+ and CalR+BrdU+ cells in the glomerular layer peaked between 45-60 days post-injection before plateauing and/or beginning to turn over (Winner et al., 2002; Kohwi et al., 2007). For these reasons, we chose to harvest 60 days after injection to allow PGCs derived from labelled radial glia and SVZ progenitor cells the time to migrate to the olfactory bulb. The source about how radial glia give OB interneurons? And how we will be technically labelling all cells dividing, including neuroblasts and TACs that came from the SVZ. So our labelling isn’t specific to the SVZ but in the olfactory bulb the EdU+NeuN+ cells should be derived from radial glia patterned at the SVZ and their progeny.
Olfactory bulb tissue preparation and immunofluorescence
To improve olfactory bulb histology, the tissue was fixed inside the skull prior to dissection, as the tissue is soft and subject to damage when dissecting it out unfixed. Once dissected out, a small amount of forebrain was left attached to the bulbs to hold them together and provide orientation when bulbs from multiple pups were placed inside the same mould. When sectioning, at least three intervening sections of 14um each were discarded between retained sections, so as to ensure that a single cell was not represented across multiple sections and to provide a broader representation of cells across the olfactory bulb.
Two conditions were used for each mouse: immunofluorescence (IHF) for tyrosine hydroxylase (TH), a marker of dopaminergic cells (Figure 1), or calretinin (CalR), a marker of calretinin-expressing cells (Figure 2). These markers delineate non-overlapping populations of glomerular interneurons (Kosaka et al., 1995; Hack et al., 2005; De Marchis et al., 2007). Both conditions also included IHF of NeuN, DAPI, and a Click-It reaction. NeuN is a nuclear-localized marker of mature neurons. NeuN labelling was used to a) more easily discern what was a true TH+EdU+ or CalR+EdU+ cell and b) be able to normalize the number of TH+ or CalR+ postnatal-born cells to the number of EdU+NeuN+ cells in each image. DAPI nuclear staining was used to for confirmation in cell counting that EdU and NeuN labelling were indeed within real nuclei. The Click-It reaction adds the fluorescent Alexa Fluor 647 azide to EdU nucleosides incorporated into the DNA of all cells that were in S-phase upon EdU injection on postnatal day 1.
To get enough cells for a representative sample, approximately 20 images were taken per mouse, per condition, from at least two different sections. A couple animals did not have enough tissue sectioned and available for imaging to reach 20 images per condition, however no animal had less than 15 images collected. Post-natal TH expression is not known to be regionally different across the olfactory bulb across dorsoventral, rostrocaudal, or mediolateral axes (McLean and Shipley, 1988), however the lateral glomerular layer appears to be wider and contains more cells compared to medial glomerular layer (Dellovade, Pfaff and Schwanzel-Fukuda, 1998). To get fair representation for all conditions and animals, images were collected in a non-overlapping fashion around the whole olfactory bulb, with from an equal proportion of medial and lateral glomerular layer. ** need to mention here and in methods that the experiment was done blind. Explain how it was done
ImageJ cell counting
Originally, I was counting double-labelled (EdU+NeuN+) cells by putting the blue EdU channel and red NeuN channels together and looking by eye for colocalization by toggling between the two channels and across multiple z-stacks. This was unreliable because the deep blue-coloured 647 azide staining of EdU was hard to distinguish from the black background for confident counting. To overcome this problem, I created a threshold to convert the image to 16-bit black-and-white format, followed by an region-of-interest (ROI) selection around EdU labelling (Figure 3B,C,D). This allowed me to quickly and consistently define what was EdU-labelled and what was not; it also allowed for easy comparison to DAPI and NeuN nuclear labelling, and TH+ or CalR+ staining. I chose the pre-set IJ_IsoData threshold in FIJI because it was the option that most consistently selected what I would have considered EdU labelling across multiple images without missing obvious labelling. Threshold selection also accommodated small or punctate EdU
Figure 1. A representative image of olfactory bulb glomerular layer collected at 20X magnification for counting. TH is a marker for dopaminergic neurons, NeuN is a marker for mature neurons, and EdU is a modified nucleoside incorporated into the DNA of cells proliferating at P1.
Figure 2. A representative image of olfactory bulb glomerular layer collected at 20X magnification for counting. CalR is a marker for calretinin-expressing neurons, NeuN is a marker for mature neurons, and EdU is a modified nucleoside incorporated into the DNA of cells proliferating at P1.
fluorescence, as the EdU incorporated during S phase may be diluted in subsequent divisions after P1 (Chehrehasa et al., 2009). Upon imaging, some EdU fluorescence (Alexa 647 azide) was oversaturated, but I allowed this to accommodate fainter or punctate fluorescence that still marked cells dividing at P1.Something about how punctate fluorescence was brought up just below saturation when imaging. EdU was later compared against DAPI staining to confirm that it was not aberrant fluorescence (Figure 4). Both EdU and DAPI staining can appear punctate in nuclei as they both target AT-rich regions of DNA.I used an over-under threshold feature on FIJI to view NeuN and TH or CalR staining (Figure 3E), as it allowed me to easily hide background noise and view staining of true cells. I did not restrict this to a set threshold feature on FIJI however, as the quality of IHF varied between individual trials and sometimes within tissue in the same olfactory bulbs. The EdU selection was then overlaid onto NeuN staining for cell counting (Figure 3F).
One drawback to using a mouse anti-NeuN primary antibody was that the donkey-anti-mouse secondary also fluorescently labelled any vasculature that intersects the olfactory bulb, due to circulating mouse antibodies in the blood. However labelled vasculature was often overexposed and easily identified by morphology as different from NeuN+ neurons. Any EdU+ selection on top of suspected vasculature NeuN+ labelling was not counted.
Figure 3. Representative images depicting methodology for counting cells. A) ROI selection defining the glomerular layer B) IJ_IsoData threshold applied to EdU labelling C) Selection created from EdU image D) Selection overlaid onto original EdU channel E) NeuN labelling viewed with over-under feature F) EdU threshold selection overlaid onto NeuN for EdU+NeuN+ counting
Figure 4. EdU is consistently localized with glomerular nuclei. Blue shows fluorescence from 647 azide bound to EdU, while magenta shows DAPI staining. Crosshairs intersect an EdU+DAPI+ cell in the orthogonal view in FIJI, generating a composite of Z-stacks in the X and Y directions. EdU and DAPI fluorescence is intermixed, indicating co-localization.
My criteria for selecting cells as double-positive were that 1) the EdU+ selection needed to be localized within DAPI nuclear staining 2) the selection needed to be overlaid clearly on top of NeuN nuclear labelling from the same z-stack image and 3) NeuN staining was clear across multiple z-stacks, not just one. If these criteria were met, I created a selection around the EdU+NeuN+ cell using the hand-drawn ROI tool. These ROIs were colour-coded separate from the EdU selection and could then be overlaid onto the channel containing Alexa 488 fluorescence for TH or CalR. The same over-under procedure and criteria #2 and #3 were used to confirm triple-labelling. I also drew colour-coded ROI selections around these cells and counted the total number of TH+EdU+NeuN+ selections and CalR+EdU+NeuN+ selections each.
If fulfillment of criteria #2 and #3 were uncertain, I compared co-localization using the orthogonal view tool in FIJI, whereby a composite of the z-plane was created by combining the pixels from the five z-stacks, for the x and y directions each. EdU, NeuN, and TH or CalR needed to be visualized together directly. Figure 5 depicts an orthogonal view of a TH+EdU+NeuN+ cell, while Figure 6 depicts an orthogonal view of a CalR+EdU+NeuN+ cell. If different staining was stacked on top of each other in the z-plane, it signified that they were not co-localized and may represent different cells entirely; these cells were not counted.
A small number of cells were observed that were TH+EdU+ or CalR+EdU+ but not NeuN+. These cells were not included since they could not be normalized to the population of EdU+NeuN+ cells. The majority of TH+EdU+ or CalR+EdU+ cells were also NeuN+.
Figure 5. A representative confocal image of a glomerular TH+EdU+NeuN+ cell in
orthogonal view in FIJI. TH is visualized in green, EdU punctae in blue, and NeuN in red. The cross-hairs intersect the triple-positive cell, generating a composite of Z-stacks in the X and Y directions. In the orthogonal sections, nuclear EdU and NeuN are intermixed, with somatic TH staining on either side, indicating co-localization.
Figure 6. A representative confocal image of a glomerular CalR+EdU+NeuN+ cell in orthogonal view in FIJI. TH is visualized in green, EdU punctae in blue, and NeuN in red. The cross-hairs intersect the triple-positive cell, generating a composite of Z-stacks in the X and Y directions. In the orthogonal sections, nuclear EdU and NeuN are intermixed, with somatic CalR staining on either side, indicating co-localization.
Figure 7. Representative image of IHF for CalR condition, at 60X magnification. The standard arrow indicates a CalR+EdU+NeuN+ cell, while the chevron arrow indicates a EdU+NeuN+ cell.
Figure 8. Representative image of IHF for TH condition, at 60X magnification. The standard arrow indicates a TH+EdU+NeuN+ cell, while the chevron arrow indicates a EdU+NeuN+ cell.
Double-positive and triple-positive cells were counted for 517MUT TH, WT TH, 517MUT CalR, and WT CalR conditions, with n=6 per genotype. Tables 1a,b show the total number of cells counted. The percentage of EdU+NeuN+ cells also TH+ or CalR+ were calculated for each animal (Tables 2a,b).
2-way ANOVA with Sidak’s multiple comparisons was used for statistical analysis. Sidak’s multiple comparisons were appropriate because we wanted to know if mutation of the 517 miR-7 target site impact TH+ PGC subtype and/or CalR PGC subtype generation. 2-way ANOVA was used because both TH+ and CalR+ PGCs both come from VZ-SVZ progenitor cells, and therefore we wanted to know if changes in the generation of one subtype may impact the generation of the other. The mean of %EdU+NeuN+ cells for TH 517MUT was 13.6% while TH WT was 13.82%; the mean difference between the two conditions was -0.212 (p=0.9902), 95% CI: -4.282 to 3.858. The mean of %EdU+NeuN+ cells for CalR 517MUT was 23.13% while CalR WT was 20.04%; the mean difference between the two conditions was 3.084 (p=0.1572), 95% CI: -0.9862 to 7.154. Therefore, there is no significant difference between the two genotypes for the TH condition, nor the CalR condition. Figure 9 summarizes these results; each bar shows the mean +/- 95% CI of each condition. In two-way ANOVA, the p-value for interaction between genotype and PGC subtype was p=0.0936, while genotype was p=0.3389, indicating that the 517MUT mice were not significantly different than WT mice. PGC subtype was p<0.0001, indicating that the number of CalR+ cells counted were significantly different than the number of TH+ cells counted.
Figure 9. The proportion of EdU+NeuN+ cells that are also TH+ or CalR+, in WT mice and 517MUT mice. Bars show the mean with 95% confidence intervals. The TH 517MUT-WT difference is -0.212% (p=0.9902) and the CalR 517MUT-WT difference is +3.084% (p=0.1572). n=6 for both genotypes. Analysis was 2-way ANOVA with Sidak’s multiple comparisons.
Discussion
Effect of the 517 mutation
There is no statistically significant difference in TH+ subtype generation between 517MUT mice and WT mice. This may suggest that disruption of the 517 site by two-base pair mutation was insufficient to relieve miR-7 repression of Pax6 and shift progenitor cell fate towards TH+ PGCs. de Chevigny et al. (2012) on the other hand, observed a statistically significant increase in TH+ PGC subtype when they suppressed miR-7 action on Pax6 via electroporation of a miR-7 sponge into the lateral wall of the SVZ. In our experiment, the effect on CalR+ PGCs generation was also not statistically significant between genotypes, however CalR+ PGCs did trend upward in 517MUT mice compared to WT. This trend is unexpected, since CalR+ subtype generation is not dependent on Pax6 and generation of TH+ PGCs following Pax6 overexpression was thought to happen at the expense of the CalR+ subtype (Waclaw et al., 2006; Brill et al., 2008). It is possible that despite the many images analyzed and cells counted, the natural variation between images and animals may be too great to reasonably and feasibly detect statistically significant changes between genotypes.
TH and CalR proportions
With respect to the proportions of periglomerular subtypes, the data generated from this experiment are somewhat difficult to situate within other literature on the topic. Some studies focus on neurogenesis in strictly adult mice, rather than the early postnatal mice, and no other study has used the same triple-labelling counting method as we have. BrdU labelling is directly comparable to EdU labelling (Chehrehasa et al., 2009), however the location of the injection is important. Some studies injected BrdU locally at the SVZ, while others injected intraperitoneally or subcutaneously as we did with EdU. Local injections of BrdU are aimed to label only SVZ progenitor cells, while systemic injections will label all cells in S phase throughout the body. Following systemic injection of BrdU, approximately half of BrdU+ cells co-labelled with NeuN (Winner et al., 2002). Therefore, following a systemic injection, the percentage of BrdU+ cells colabelled for TH+/CalR+ is not as comparable without also labelling for NeuN, due to local gliogenesis in the olfactory bulb (Fukushima et al., 2002). NeuN+ cells generated after birth are primarily from cells migrating from the SVZ-RMS pathway, with very few from local neurogenesis (Fukushima et al., 2002). There are some oligodendrocytes generated in early postnatal stages however, primarily from NG2-expressing progenitors (Aguirre and Gallo, 2004), so studies not labelling for NeuN+ may underestimate the percentage of postnatal-born neuron subtypes. Several studies have looked at either TH+BrdU+ or NeuN+BrdU+ normalized to BrdU+ cells in the olfactory bulb, or the total numbers of these cells, but not the proportions of triple-labelled cells (e.g. TH+EdU+NeuN+) as we have.
For example, we found that of the EdU+NeuN+ cells counted, on average 13% of them were also TH+ for both the 517MUT and WT conditions. One study injected BrdU once locally at the SVZ of two-day old rats; they harvested the olfactory bulbs three weeks later and found that about 10% of the BrdU+ cells in the periglomerular layer were also TH+ (Betarbet et al., 1996). The percentage of TH+BrdU+ from the total number of BrdU+ cells in this study is likely underestimated compared to our study, due to their use of an incubation period shorter than what is estimated for TH+ differentiation (Winner et al., 2002; Kohwi et al., 2005), as well the potential inclusion of any BrdU+ SVZ-derived glial cells. Our experiment has a longer incubation period and is focused on neurons, which may account for the slightly elevated proportion of TH+ cells (13%) to the results of the experiment by Betarbet et al. (10%). Additionally, our EdU injection would have labelled the specified neuroblasts in the RMS that undergo a couple divisions en route to the olfactory bulb—these cells would have evaded local BrdU labelling at the SVZ.
Kohwi et al. (2007) injected 60-day old mice with BrdU intraperitoneally and collected mice at different survival times for analyses. After BrdU injection, CalR+ PGCs are generated a little more quickly than TH+ PGCs, but both peak at around 45-60 days post-injection. They estimate that the ratio between TH:CalR neurons at P90 is approximately 1:2. This is somewhat comparable to results from De Marchis et al. (2007), who found the TH:CalR ratio to be about 1:1.8, although they labelled the SVZ directly with fluorogold and harvested 30 days after injection. Both of these studies use adult mice and do not label for neurons specifically when counting the proportions of labelled cells. Regardless, the results from our approach do compare somewhat to these studies—we found TH subtype generation to be little more than half that of CalR generation. In combination with the observation that the proportion of CalR:TH neurons generated from the SVZ does not change substantially between early postnatal and adult mice (Kohwi et al., 2007), the proportions of subtypes that we observed fit in with the results from other studies that use different approaches.
Cell-autonomous and transcriptional regulation of periglomerular identity
Periglomerular identity appears to be regulated in a cell-autonomous fashion. Merkle, Mirzadeh, and Alvarez-Buylla (2007) stereotactically labelled distinct populations of SVZ radial glia at P0 with adenovirus-cre recombinase to force GFP expression in a localized set of cells. Radial glia from distinct locations in the VZ-SVZ gave rise to distinct PGC subtype populations in the olfactory bulb 40 days later, e.g. TH+ PGCs from dorsal progenitors, CalR+ from anterior and medial progenitors. Transplanting GFP-labelled early postnatal radial glia to corresponding or different regions of a host VZ-SVZ produced no change in the periglomerular subtype generated by these cells, suggesting progenitor cell fate was pre-specified (Merkle, Mirzadeh and Alvarez-Buylla, 2007). Other experiments confirm that Pax6 acts cell-autonomously to inform PGC subfate. GFP-expressing Pax6Sey/Sey precursor cells (null for Pax6), grafted into WT adult SVZ, were able to produce neuroblasts capable of migrating and fully differentiating in the olfactory bulb into mature neurons (NeuN+), however they were explicitly unable to generate TH+ PGCs, unlike Pax6+/+ grafts (Kohwi et al., 2005). Additionally, heterozygous Pax6Sey/+grafts were also unable to produce TH+PGCs as well, confirming the Pax6 dosage-dependence of TH+ PGC generation (Kohwi et al., 2005).
Some work has been done to identify the mechanism by which Pax6 promotes TH+ PGC fate. Pax6 works with other transcription factors, including Dslx2 and Meis2, which both interact with and are required alongside Pax6 for adult SVZ-generated TH+ PGC fate (Brill et al., 2008; Agoston et al., 2014). Pax6 and Dslx2 form dimers in the SVZ and the olfactory bulb, and Meis2 is colocalized with them in the RMS and olfactory bulb. In particular, inhibition of either Dslx2 or Meis2 expression or action inhibited TH+ subtype formation, while overexpression in the RMS can generate an increased proportion of TH+ PGCs in the glomerular layer (Hack et al., 2005; Brill et al., 2008; Agoston et al., 2014). Expression of the tyrosine hydroxylase gene was found to be controlled by Meis2, as it directly interacts with its upstream regulatory elements. That Meis2 is not expressed until later in the migratory pathway, particularly the RMS and olfactory bulb, is consistent with the observation that specified PGC precursors do not become TH+ until they reach the olfactory bulb (Baker et al., 1983). The mechanisms by which Meis2 expression is regulated as well as the exact pathways, targets, and interaction of these transcription factors are yet unknown.
Pax6 and SVZ-RMS neurogenesis
Given the well-characterized role of Pax6 elsewhere in other tissues on neurogenesis and cell proliferation, it is possible that changes in Pax6 dosage may result in different rates of neurogenesis overall. Hack et al. (2005) observed dynamic regulation of Pax6 in the journey throughout the adult SVZ, RMS, and OB. Specifically, they saw relatively low Pax6 expression in SVZ, upregulated expression in most neuroblasts of the RMS, followed by persistence of Pax6 in TH+ PGCs and an absence of Pax6 expression in CalR+ PGCs. Overexpression of Pax6 in the RMS directed neuroblasts towards a PGC cell fate at the expense of granule cells, while injection of a dominant-negative version into the RMS reduced PGC formation. Additionally, a transcription factor called Olig2 opposes Pax6 activity, is expressed in precursor and transit amplifying cells in the SVZ, promotes astrocyte identity, and is downregulated in neuroblasts (Hack et al., 2005). They suggest that Pax6 not only has a role in specifying PGCs towards TH+ fate, but that Pax6 is required for overall neurogenesis in the RMS, against Olig2. In the VZ-SVZ however, de Chevigny et al. (2012) did not observe any changes in neurogenesis following overexpressed Pax6 or miR-7. They place the VZ-SVZ gradients of miR-7 and Pax6 at the level of PGC subtype specification, not at neurogenesis. It is possible that Pax6 has a dual role in the SVZ-OB pathway: subtype specification in the VZ-SVZ, and neurogenesis in the RMS. Given the upregulation of Pax6 in the RMS for neurogenesis, it would be interesting to observe if there is a concurrent downregulation of Pax6-acting miRNAs in migrating neuroblasts. Downregulation of miR-7 may be required in neuroblasts derived from the ventrolateral wall, which were fated for CalR+ identity by high miR-7 and low Pax6 expression in precursor cells.
EdU as a marker of cell division
When labelling for cell proliferation, one caveat to using nucleoside analogs such as EdU or BrdU is that they specifically label DNA synthesis, not cell division itself. Because of this, there is the possibility that modified nucleosides may be incorporated into the DNA during repair of nicked or damaged sequences. However, BrdU+ cells analyzed shortly after labelling do not express the mature neuronal marker NeuN, suggesting that within mature neurons undergoing DNA repair, nucleoside analogs do not sufficiently incorporate for detection (Winner et al., 2002). We also have little reason to suspect that the 517MUT condition would differ substantially to the WT mice on the basis of DNA damage and repair.
Once concern is that incorporation of EdU is harmful to dividing cells. EdU has been used extensively to label proliferating cells in vitro and was observed to be very robust in vivo as well, whereby all BrdU-labelled cells are also EdU-labelled (Chehrehasa et al., 2009). EdU has the benefit of having little to no background or aberrant fluorescence, compared to antibody-detection of BrdU, and did not require the same potentially damaging conditions for DNA-denaturing required for BrdU detection. For these reasons, EdU was preferable to BrdU for this experiment. However, there is some evidence that EdU may have other detrimental effects. Ponti et al. (2013) examined cell cycle of SVZ astrocytes (NSCs) and their progeny in adult neurogenesis. In their supplementary materials, they provide evidence that EdU may be cytotoxic to V-SVZ cells attempting a second S phase, compared to the use of CldU nucleoside analogs. Up to 36 hours after injection, the quantity of EdU+ and CldU+ NSCs was comparable, but timepoints past this saw a reduction of EdU+ cells and an increase in pyknotic nuclei, compared to CldU+ cells (Ponti et al., 2013). This may be a confounding factor, if EdU toxicity affects PGC subtypes differently throughout our relatively long-term experiment. To investigate this possibility, we will be performing the same PGC subtype characterization experiments on BrdU-injected mice as a control. It may be that use of EdU knocks down total PGC generation number, without injuring the generation of TH+ or CalR+ PGCs preferentially. However, if this issue proves to be a concern, repeating the experiment with BrdU nucleosides may be necessary.
In vivo miRNA target mutagenesis
Some studies have attempted to assess the function of miRNAs on their targets by knocking out the gene for the miRNA completely. There is a number of complications associated with this approach. First, many miRNAs exist within families of a number of similar miRNAs that may have function on the same target genes. miR-7 represents a family containing three conserved genes in different genomic loci: in mice, they are called miR-7a-1, miR-7a-2, and miR-7b. Knocking out all miR-7 genes in a mouse model may prove difficult, however, disrupting a specific target may prevent the action of all of them at that site. Second, a single miRNA may regulate multiple targets—it is then difficult to distinguish the effect on a single target from other potential consequences (Bartel and Chen, 2004). Additionally, phenotypic effects may be secondary to the miRNA deletion, e.g. de-repression of a transcription factor results in up- or down-regulation of genes, which are not directly regulated by miRNAs. In sum, deletion of a miRNA does not illuminate a direct relationship between the miRNA and a single target, and masks the molecular mechanisms that lead to observed phenotypic changes (Staton and Giraldez, 2011). On the other hand, overexpression experiments with miRNAs may identify target genes but provide limited information about target-miRNA interaction under normal circumstances. Other methods to probe the interaction between a target site and its cognate miRNA may provide more useful information.
The most common methodology to assess the function of recognition sites experimentally is fusion of the sequences surrounding and including the MRE site, or the full 3’UTR, to a reporter, and then express this construct in a cell line made to also express the miRNA of interest (e.g. luciferase assays). This method, while useful, does not establish the significance of the miRNA-site interaction in vivo, as it cannot reproduce the natural context of the interaction (Bassett et al., 2014). In particular, excising or moving individual MREs from their place in the 3’UTR can artificially render them non-functional, giving a false negative readout (Didiano and Hobert, 2008). Target protector (TP) antisense oligonucleotides are used as a means to examine MRE function in vivo, by blocking the predicted recognition site from miRNA interaction (Choi, Giraldez and Schier, 2007). This method overcomes some of the problems with reporter assays, but it also has several limitations: TPs morpholinos may have off-target effects if sequences similar to the target are present elsewhere in the genome; potential off-target effects are currently poorly characterized. Additionally, the TP may not fully bind to the target, providing incomplete protection of the site from complementary miRNA that is difficult to assess and may hide the exact nature of the miRNA-mRNA interaction (Staton and Giraldez, 2011). Finally, TP morpholinos are best used for transient purposes as they are injected or transfected into cells and consequently don’t last permanently throughout development (Staton and Giraldez, 2011).
The common problem for all of these methods is that they don’t directly examine the functional relevance of a specific miR site in a specific 3’UTR, within a biological system. In vivo MRE mutagenesis overcomes some of these problems. To date, this approach has been performed very rarely. In 2008, Dorsett et al. used LoxP-mediated gene targeting in mice to replace nucleotides forming the only miR-155 recognition site in the Aicda 3’UTR with a GC-rich sequence not predicted to match the seed site of any miRNAs. Aicda gene produces activation-induced cytidine deaminase (AID), an important gene for somatic hypermutation in B lymphocytes, as well as chromosomal translocations. Aicda155 mice had an increase in Aicda mRNA lifespan and protein levels, as well as an increase in translocational activity known to be oncogenic in Burkitt’s lymphoma, which is associated with miR-155-deficiency (Dorsett et al., 2008). This experiment established miR-155 as a tumour suppressor and was the first experiment to mutate an MRE directly. In 2014, Bassett et al. produced targeted MRE gene mutations in zebrafish with transcription activator-like effector nucleases (TALENS) and in drosophila with CRISPR-Cas9 gene editing and homology-directed repair. They were able to confirm bona fide targets—for miR-430 in the lefty2 3’UTR, and for the miRNA bantam in the enabled 3’UTR, in zebrafish and drosophila respectively—and analyze their functional relevance. Again, MRE mutation enabled direct analysis of the phenotypic consequences from miRNA-MRE interaction. Our experiment, whereby the periglomerular phenotypic consequences of a miR-7 MRE mutation are characterized, adds to the emerging conversation on miRNA-target interaction, the functional significance of miRNA target sites, as well as the networks of gene expression and regulation that work to specify adult-born PGC subtypes in particular.
517 and 655 target sites
A few studies have identified Pax6 as a true target of miR-7 and used mutations of different MREs in the 3’UTR to confirm this interaction in vitro. Zhao et al. (2012) conducted a luciferase assay in a COS-7 cell line with the mouse miR-7 locus cloned into it, as well as the reporter-fused Pax6 3’UTR carrying two putative miR-7 binding sites: the 517 site (GTTTTCCA) and the 655 site (GTCTTCC). Transfection of the miR-7 locus with the WT Pax6 3’UTR-fused reporter into cells produced a significant decrease in luciferase activity compared to cells transfected with the WT 3’UTR-reporter only. In miR-7-expressing cells, mutation of the 517 (GTTTTCCAGTTTTGGA) and the 655 (GTCTTCCGTCCGAA) seed sites relieved repression of the 3’UTR reporter nearly entirely (Zhao et al., 2012). This may suggest that miR-7 cooperativity at one or both of these sites mediates miR-7 downregulation of Pax6. They did not conduct this experiment with mutation of one site at a time however, so it is unclear how much each site contributes to Pax6 repression or if they work cooperatively to mediate this repression. De Chevigny et al. (2012) and Kredo-Russo et al. (2012) also used a similar luciferase approach in HEK293T cell lines, although they identified and mutated the 655 site, but not the 517 site.
The identification of 655 as another potentially functional site for miR-7 target recognition suggests that miR-7 may act on multiple sites to mediate Pax6 expression in the VZ-SVZ and the arising PGC subtypes. Global analysis found that repression of a mRNA by a certain miRNA not only depends on seed site complementarity, but also on the number of target sites and the distance between these binding sites (Hon and Zhang, 2007). Bridget Ryan’s luciferase assays found that mutation of the 517 or 655 sites individually did not restore Pax6 3’UTR-fused reporter activity from miR-7 repression, but that mutation of both 517 and 655 together did. For this reason, we are generating additional mouse lines: one with the 655 seed site disrupted (GTCTTCCGTCCGAA), and one with both 517 and 655 sites disrupted. The 517 site may indeed be functional in vivo, but the effect of mutation on PGC subtype may be hidden by Pax6 repression mediated through the 655 site.
Other miRNAs acting on Pax6
It has been observed that transcription factors and genes crucial to development are often regulated by multiple miRNAs (John et al., 2004; Hon and Zhang, 2007), therefore miR-7 may not be the only miRNA acting on the Pax6 transcript. This possibility was examined by de Chevigny et al. (2012). Of the predicted conserved miR binding sites in the Pax6 3’UTR, only those for miR-375, miR-7, and miR-365 significantly reduced luciferase levels relative to cells transfected with an empty vector. They quantified the absolute expression levels of these miRNAs from microdissections of the dorsal and lateral walls of the VZ-SVZ. miR-7 expression was enriched in the lateral wall versus the dorsal wall at both the molecular and cellular levels; a later experiment separated the dorsolateral aspect from the ventrolateral, and miR-7 was found to be very low in the dorsal, intermediate in the dorsolateral, and high in the ventrolateral, suggesting a gradient of expression. Negligible levels of miR-375 were detected in either dorsal or lateral wall, while miR-375 was weakly expressed in the dorsal wall, with slightly increased expression in the lateral wall. It may be possible that miR-375 represses Pax6 expression in the ventrolateral SVZ as well, but miR-7 was considered of primary interest by de Chevigny et al. due to its steep, opposing gradient to Pax6. Additionally, electroporation of a miR-7 sponge construct (repetitive sequences complementary to miR-7) into the lateral wall produced an increased number of TH+ PGCs in the olfactory bulb 15 days later, while electroporation into the dorsal wall had no effect on TH+ PGC generation. It appears that miR-7 is the foremost miRNA regulating the gradient of Pax6 expression in radial glia of the postnatal VZ-SVZ.
Regulatory roles for miR-7
The effect of miRNAs on their targets is frequently modest, and they may function in conferring robustness to complex networks by operating in feed-forward and feedback regulation loops (Tsang, Zhu and van Oudenaarden, 2007; Baek et al., 2008). In drosophila, miR-7 has been found to participate with transcription factors in feedback and feedforward loops important to development; specifically, miR-7 buffered the response or output of these systems against environmental perturbation (Li et al., 2009). In developing larvae, miR-7 was found to work in a network to activate expression of a protein called Ato (atonal) and repress expression of Yan. Ato expression enables transcription of genes required to specify proneural cluster cells to adopt sensory organ precursor fate, eventually giving rise to olfactory and proprioceptor organs. miR-7 mutant larvae showed no defect in Ato or Yan expression under uniform laboratory conditions (Li et al., 2009). Perturbation of the environment by fluctuating the temperature between 31ºC and 18ºC resulted in a strong decrease in Ato expression and a strong increase in Yan in miR-7 mutants, consistent with the miR-7 mutant being unable to regulate their expression. WT larvae showed no changes in expression even with temperature fluctuation.
For our experiment, if this role for miR-7 is conserved to mice, it suggests that interference with miR-7 action (i.e. 517MUT) may not produce measurable changes in the expression of Pax6 and consequent NSC fates under uniform laboratory conditions. Exposing the 517MUT mice to some form of olfactory variation or stress may be required to significantly alter the proportion of postnatal-born PGC subtypes. It may be possible that miR-7 has a role in conferring robustness to PGC subtype generation in the olfactory system. Survival of TH+ PGCs is dependent upon olfactory stimulation: odor deprivation or nostril-blocking decreased their survival, while odor enrichment increased survival (Sawada et al., 2011). Odor enrichment has also been found to increase the number of newly-born neurons in mouse adult olfactory bulb and improve olfactory memory (Rochefort et al., 2002). Specifically, TH+ PGCs may mediate olfactory system plasticity in these cases: olfactory enrichment with the same fragrances from Rochefort et al. (2002) increased TH+ PGC neurogenesis specifically, while having no effects on CalR+ or CalB+ subtype generation (Bonzano et al., 2014). Given the responsiveness of TH+ PGC generation to the sensory environment, miR-7 and Pax6 may function in a feedback network of transcription factors to accommodate, facilitate, or mediate this responsiveness. A fluctuating olfactory environment may be a suitable model to test this hypothesis.
Additional experiments
While this experiment examined the systemic effect of mutating a putative miR-7 target site in the Pax6 3’UTR, it has not actually confirmed directly if the 2bp 517 mutation was effective in disrupting miR-7 interaction in vivo. To examine this, Bridget Ryan will use a complementary oligonucleotide to capture and purify Pax6 transcript from cells dissected from the postnatal SVZ. This method will provide a sample of Pax6 mRNA with all bound miRNAs still attached (Vencken et al., 2014). We then plan to use quantitative PCR to identify all transcript-associated miRNAs, as well as to confirm if miR-7 directly interacts with Pax6 3’UTR in the SVZ and if the 517 mutation (and/or 655 mutation) disrupt miR-7 association.
Additionally, this experiment operates on the assumption that following an inability for miR-7 to regulate the Pax6 3’UTR, an increase in the domain of Pax6 protein expression into the ventral aspect will increase the proportion of TH+ PGC identity. This experiment has not evaluated the actual domain of expression of Pax6 in the postnatal SVZ of 517MUT and WT mice. It is unknown at this point whether the 517 mutation enabled an expansion of Pax6 expression. It may be that the 517 site alone was insufficient to change Pax6 expression, and this is the reason why there was no measurable change in TH+ PGC subtypes. However, it may be that it did function to change Pax6 expression in the SVZ but did not have a measurable effect on subtype proportion in the olfactory bulb. To investigate these differences, we plan to harvest brains from 517MUT and WT mice (and 655MUT and 517+655MUT mice), section them, and use quantitative immunohistofluorescence to measure Pax6 protein expression in the postnatal SVZ. Within images, we will divide the SVZ into equal domains along the dorsoventral axis with an ImageJ script, and quantify Pax6 fluorescence in each domain as a percentage of dorsal-most Pax6 expression. This way, we might detect Pax6 domain expansion between mice following miR-7 recognition site mutagenesis.
Conclusion and significance
This thesis examined the consequences of in vivo mutagenesis of a single putative miR-7 target site in the Pax6 3’UTR on post-natal-born PGC subtype. Mutation of the 517 site was found to be insufficient to alter the proportion of TH+PGCs in the olfactory bulb glomerular layer generated 60 days after mitotic labelling at P1. This work is the third to examine the effect of in vivo miRNA target site mutagenesis (Dorsett et al., 2008; Bassett et al., 2014), and is the first, to our knowledge, to use CRISPR-Cas9 gene editing towards this goal in mammals. Future work includes characterizing the phenotypes of mice with a mutation at the 655 site and 517+655 sites together, as well as determining if miR-7 site mutation relieves Pax6 repression in the VZ-SVZ. This research adds to the conversation on the phenotypic function of miRNA regulation as well as the mechanisms by which Pax6 is regulated and periglomerular subtype is specified in the mammalian olfactory bulb.
2018-4-12-1523565337