MicroRNA in Situ Hybridization in the Human Entorhinal and Transentorhinal Cortex

MicroRNAs (miRNAs) play key roles in gene expression regulation in both healthy and disease brains. To better understand those roles, it is necessary to characterize the miRNAs that are expressed in particular cell types under a range of conditions. In situ hybridization (ISH) can demonstrate cell- and lamina-specific patterns of miRNA expression that would be lost in tissue-level expression profiling. In the present study, ISH was performed with special focus on the human entorhinal cortex (EC) and transentorhinal cortex (TEC). The TEC is the area of the cerebral cortex that first develops neurofibrillary tangles in Alzheimer's disease (AD). However, the reason for TEC's special vulnerability to AD-type pathology is unknown. MiRNA ISH was performed on three human brains with well-characterized clinical and pathological parameters. Locked nucleic acid ISH probes were used referent to miR-107, miR-124, miR-125b, and miR-320. In order to correlate the ISH data with AD pathology, the ISH staining was compared with near-adjacent slides processed using Thioflavine stains. Not all neurons or cortical lamina stain with equal intensity for individual miRNAs. As with other areas of brain, the TEC and EC have characteristic miRNA expression patterns. MiRNA ISH is among the first methods to show special staining characteristics of cells and laminae of the human TEC.


INTRODUCTION
Alzheimer's disease (AD) is a prevalent neurodegenerative disease that culminates in severe defi cits in cognition and autonomy. By defi nition, brains affl icted by AD contain two different neuropathological hallmarks -neurofi brillary tangles (NFTs) and neuritic amyloid plaques (NPs) The National Institute on Aging, and Reagan Institute Working Group on Diagnostic Criteria for the Neuropathological Assessment of Alzheimer's Disease (1997). NFTs are 'inclusion bodies' , composed of insoluble tau protein polymers that coalesce within neurons. NPs consists a roughly-spherical extracellular component that includes fi brillary polymers of the Aβ peptide, with nearby degenerating cell processes that contain tau polymers indistinguishable from those in NFTs.
Neuroanatomically, AD pathology manifests in a complex but well-characterized spatiotemporal sequence (Braak and Braak, 1991;Braak et al., 1993). Most clinico-pathological correlation studies indicate that cortical NFT density, assessed by Braak staging  or other means, is the parameter best correlated with the severity of AD cognitive impairment (Arriagada et al., 1992;Nelson et al., 2007bNelson et al., , 2008aNelson et al., ,b, 2009bSonnen et al., 2007). In the fi rst stages of the disease, NFTs are observed in medial temporal lobe structures (Braak and Braak, 1991).
The specifi c cerebral cortical subfi eld with earliest NFT formation in AD is the transentorhinal cortex (TEC) (Braak and Braak, 1992). The TEC usually occupies the medial bank of the perirhinal collateral sulcus, comprising ∼2-10 mm of the inexactly defi ned and phylogenetically variable Brodmann Area 35 (Schmidt et al., 1993;Taylor and Probst, 2008). As its name implies, the TEC constitutes a to serve key functions in neurodevelopment, synaptic plasticity, and neuroprotection (Kosik and Krichevsky, 2005;Cuellar et al., 2008;Smalheiser and Lugli, 2009). MiRNAs may have potentiated mammalian brain evolution by amplifying the complexity of nervous system gene expression regulation (Nelson and Keller, 2007;Heimberg et al., 2008). On the other hand, miRNAs also contribute to human illnesses, particularly in the pathogenesis of human neurodegenerative disease (Nelson et al., 2008b;Hebert and De Strooper, 2009). ISH shows important cerebral cortical lamina-specifi c patterns of miRNA expression that would be lost on most tissue-level expression studies (Mellios et al., 2008;Nelson and Wilfred, 2009), and these lamina-specifi c miRNA expression patterns could be relevant to AD (Wang et al., 2008).
We chose to study four miRNAs that are expressed in human brain: (1) miR-107, which we have shown may be relevant to AD pathogenesis and traumatic brain injury, and which may be involved in metabolic regulation (Wilfred et al., 2007;Wang et al., 2008;Redell et al., 2009;Tang et al., 2009). (2) miR-124, which is highly enriched in neurons and plays many important roles in neuronal gene expression regulation (Smirnova et al., 2005;Krichevsky et al., 2006;Makeyev et al., 2007;Tang et al., 2007). (3) miR-125b, which is expressed in many different cell types including the mammalian brain and which has been proposed to play a number of complex nervous system roles (Smirnova et al., 2005;Lukiw and Pogue, 2007;Ferretti et al., 2008;Le et al., 2009a,b). (4) miR-320, which is highly expressed in neurons and glial cells and which is dysregulated in prion disease (Nelson et al., 2007a;Saba et al., 2008).

MATERIALS AND METHODS
Brain tissue was obtained from University of Kentucky ADC Brain bank using appropriate IRB protocols. Details of subject recruitment, autopsies, and other analyses using the University of Kentucky ADC autopsy series are described elsewhere (Nelson et al., 2007b(Nelson et al., , 2008a. Criteria for inclusion in this study included post-mortem intervals (PMIs) under 5 h. Brain sections from two individuals without antemortem cognitive decline were included (Cases 1 and 2), and a third person with early AD (Case 3). Demographic and pathological parameters of each of the three cases used for this study are shown in Table 1.
Human brain ISH methods have been published (Nelson et al., 2006;Wang et al., 2008;Nelson and Wilfred, 2009). Very briefl y, post-mortem human brain sample was obtained via autopsy within 5 h of death. Tissue portions that included the EC were used. Brain tissues were fi xed in 4% paraformaldehyde overnight at 4 o C and then immersed in 20% sucrose (4 o C) for an additional 24-48 h. Tissue was cut to 25 microns on a freezing microtome and mounted onto premarked Superfrost® Plus slides. Cut tissue sections were allowed to air-dry for 30 min. The slides were then transferred to a −80 o C freezer, until subsequent processing as described. Digoxigenin-labeled locked nucleic acid probes (Exiqon, Woburn MA) were used and their presence visualized via anti-digoxigenin immunohistocyhemistry using protocols described in detail previously (Wang et al., 2008;Nelson and Wilfred, 2009).
Histological stains were performed on near-serial sections that were fi xed, cut, mounted, and frozen along with the sections used for ISH. Thiofl avine S (Polysciences, Inc., Warrington, Pennsylvania) was used as a 1% aqueous solution followed by differentiation in two changes of 80% ethanol. Nissl staining was accomplished using 0.1% Cresyl violet solution that was fi ltered immediately before use. After staining for 5 min, sections were differentiated in 95% ethanol and cleared in xylenes.

RESULTS
Sections of human TEC and EC were evaluated using miRNA ISH and several histological stains. Photomicrographs from each of the three cases are presented in Figures 1-3. These show ISH results for miR-107, miR-124, miR-125b, and miR-320 in the TEC and nearby structures in correlation to AD pathology (Thiofl avine S stained NFTs and NPs).
Since miR-320 and miR-124 showed distinctive cortical laminar staining in the TEC, a separate panel (Figure 4) shows the ISH results for near-serial section using miR-320 and miR-124 ISH and Thiofl avine S from Case 3. Note that the Thiofl avine S-stained NFT-bearing neurons are present in a band of cells that are relatively lacking in ISH stain for both miR-124 and miR-320. For both miR-124 and miR-320, there is an immediately more superfi cial band of cells that are labeled. As can be seen in Figures 1-3, the staining pattern in and near the TEC is relatively consistent with regard to this staining pattern of cells -the layer of miR-124 cells appears particularly characteristic. MiR-107 stains less strongly than the other probes. As expected, miR-125b, and to a lesser degree miR-107 and miR-320, appear to stain glial cells in addition to  neurons. In some of the sections there was artifactual staining in the white matter that was trimmed out of the photomicrographs. To see what the staining in the white matter looks like, please see SupplementaryMaterial.
Figure 5 provides a more comprehensive picture of the EC and the nearby subiculum. ISH shows results using probes for miR-320, miR-124, and miR-125b in Case 1. Note that the layer II ("pre-α") EC "islands" are relatively strongly labeled for miR-124, whereas the the superfi cial portions of the pre-, para-, and pro-subiculum are somewhat more strongly labeled for miR-125b. This pattern was also seen in other cases (data not shown).

DISCUSSION
ISH using a set of probes against brain-enriched miRNAs was used to assess cellular miRNA expression in the human TEC and surrounding structures. These data have both technical and theoretical implications. From a technical standpoint, the present study further underscores the importance of ISH as a technique to complement tissue-level miRNA expression profi ling. The pattern of miRNA expression also helps to refi ne the expectations for miRNA functions in the brain, in both normal and disease conditions. The pattern of ISH labeling in the human cerebral cortex affi rms that defi ning individual miRNAs as "neuronal" or "non-neuronal" is overly simplistic, because different populations of neurons -even within a tiny cell layer -can have distinct miRNA expression profi les.
There are some limitations to the current study. ISH is a relatively low-throughput technique and thus we only were able to thoroughly evaluate the results using a handful of miRNA probes, and brain sections from only three individuals' brains (all of these patients were elderly Caucasians). Although the results were consistent among the cases evaluated, it is possible that study of a larger population would result in greater variability or even completely disparate results. Also, it has been shown that post-mortem degradation of miRNAs in the human brain can be rapid and can affect different miRNAs at different rates (Sethi and Lukiw, 2009). In the current study we used brains that had relatively short PMIs (less than 5 h). However, using cases with even shorter PMIs may have revealed a different staining pattern. Finally, the technique that we used employs both ISH and Note that for both miR-124 and miR-320, it is layers with a relative paucity of ISH stain (arrows on the left) that are most intensely stained with the Thiofl avine S (arrows on the right). Pial surface is indicated in A with an asterisk. Scale bar = 300 microns.
immunohistochemistry (anti-digoxigenin antibody is used to detect the digoxigenin-labeled RNA probe). Like the great majority of histochemical staining methods, ISH is not a rigorously quantitative technique for evaluating gene expression. This is because of the many variables in tissue processing that are impossible to control between cases and even between different sections of the same case. As such it is more appropriate to evaluate critically the staining pattern, rather than the more detailed characteristics of staining intensity in any given section. We also are dissatisfi ed with employing near-serial sections to defi ne correlative staining patterns. Hence we are investigating the use of double-label methods to enable, in the future, more defi nite determination of human brain miRNA co-expression patterns.
Despite the limitations inherent to this type of study, the ISH technique reveal consistent staining patterns in the human cerebral cortex. The TEC itself is a small (<1 cm) but highly intriguing cortical subfi eld, being most developed in the human relative even to other primate species (Braak et al., 2000;Taylor and Probst, 2008). Perhaps most interesting is the cell population that seems to "dive" from the superfi cial layer II neurons of the EC, through the entire depth of layer III in the TEC, to approximate layer V neurons in the "temporal proneocortex" Braak, 1985, 1992). These are apparently the neurons most vulnerable to NFT development in the human cerebral cortex. These particular neurons show an apparent lack of high amounts of both miR-124 and of miR-320. The impact of these expression patterns are currently not understood. They may relate in some way with the remarkable dendritic plasticity shown by TEC and EC neurons (Arendt et al., 1998). However, neurons in the nearby pre-α (layer II) of the EC proper, which is also affected relatively early in AD, express high amounts of both miR-320 and miR-124. There are of course many other miRNAs with expression patterns overlaid on those of miR-107, miR-125b, miR-124, and miR-320. For all of these, the lamina-specifi c expression characteristics would not be possible to evaluate using most tissue-level expression profi ling platforms.
In conclusion, we have demonstrated novel miRNA expression patterns in the human TEC and EC. Neurodegenerative diseases tend to affect these cell populations for reasons that currently are poorly understood. MiRNAs are important components of a neuron's gene expression repertoire. Each neuronal subpopulation may express a distinct fraction of the uniquely-human miRNome. Thus,