Subcellular RNA-seq for the Analysis of the Dendritic and Somatic Transcriptomes of Single Neurons

[Abstract] In neurons, local translation in dendritic and axonal compartments allows for the fast and ondemand modification of the local proteome. As the last few years have witnessed dramatic advancements in our appreciation of the brain’s neuronal diversity, it is increasingly relevant to understand how local translation is regulated according to cell type. To this end, both sequencing-based and imaging-based techniques have recently been reported. Here, we present a subcellular single cell RNA sequencing protocol that allows molecular quantification from the soma and dendrites of single neurons, and which can be scaled up for the characterization of several hundreds to thousands of neurons. Somata and dendrites of cultured neurons are dissected using laser capture microdissection, followed by cell lysis to release mRNA content. Reverse transcription is then conducted using an indexed primer that allows the downstream pooling of samples. The pooled cDNA library is prepared for and sequenced in an Illumina platform. Finally, the data generated are processed and converted into a gene vs. cells digital expression table. This protocol provides detailed instructions for both wet lab and bioinformatic steps, as well as insights into controls, data analysis, interpretations, and ways to achieve robust and reproducible results.

specialization of these compartments is, in part, accomplished via the selective transport and local translation of mRNAs (Holt et al., 2019). To characterize the neuronal local transcriptome, many studies have performed bulk RNA profiles in brain regions enriched in dendrites and axons (Zhong et al., 2006;Cajigas et al., 2012;Glock et al., 2020), on synaptic particles isolated from tissue (Hafner et al., 2019), or from neuronal cultures in chambers that separate cell bodies and neurites (Gumy et al., 2011;Poon et al., 2006). These studies have revealed that protein functions such as synaptic transmission, cytoskeletal regulation, and translation itself (among others) are encoded in the local transcriptome (Holt et al., 2019). However, as recent advancements in single cell transcriptomics have revealed, the brain contains a complex array of neuronal types, raising the question of how variable the local transcriptome is across diverse cell types (Wang et al., 2020;Perez et al., 2021).
To address this question, single cell resolution of the local transcriptome is needed. There are three main challenges to implementing such an approach: (1) the isolation of mRNAs from distinct subcellular compartments of a single neuron, (2)   A. Images showing the dissection of the soma and dendrites of a neuron using LCM. B. Library preparation workflow, showing the sequences, primers, and key enzymes used at every step.
Altogether, this method can serve as a powerful tool to achieve an unbiased investigation of cell type effects in the local transcriptome, and can be implemented across cells derived from different brain regions, developmental stages, or species. Additionally, it may be used to study single cell responses to pharmacological treatments, or other manipulations that induce changes in cell states (e.g., paradigms of synaptic plasticity). It may also be useful to study the local transcriptome of other polarized cell types, such as astrocytes (Sakers et al., 2017;Mazaré et al., 2021), or epithelial cells (Moor et al., 2017). Finally, it should be possible to adjust this protocol to profile non-coding RNAs, such as small RNAs    3. Open the PALMRobo software. Using the 20× objective, identify and register locations for microdissection. We recommend the collection of no more than 48 samples per plate (see Note 1): the somata and dendritic arbors of 16 neurons amenable for dissection, 12 somata whose dendritic arbors are not amenable for collection, and 4 empty cuts (see Note 2). Thus, at this point the location of 28 neurons and 4 empty regions should be saved. Neurons amenable for collection have isolated somata free of processes, AND isolated dendritic arbors in which most processes can be unambiguously assigned to the same neuron (See example in Figure 1A and Supplementary File 2). Avoid neurons at the edges of the coverslip, as these are inefficiently catapulted.
4. Switch to the 40× objective, go to the first registered location, and take a picture. 5. For somata or empty regions, select AutoLPC from the Cut Tools menu, and the circle drawing tool to delineate the soma or region of interest. For dendrites, select LineAutoLPC from the Cut Tools menu, and the free-hand drawing tool to delineate processes. Because of the continuous nature of neuronal compartments, there is no obvious point where the soma ends and dendrites begin, and thus what we consider the border between the two compartments is ultimately arbitrary. In our experiments, we delineate the soma as the cellular area containing and surrounding the nucleus, which dilates out until drastic decrements in width suddenly occur. The processes that continue are considered dendrites. However, processes that are qualitatively thinner than the rest are excluded as these might be axons (see example in Figure  18. Repeat steps 5 to 17 for each location registered in step 3. We recommend organizing samples shown in Figure 4A in each 96-well plate. 19. Proceed immediately to the next step. 3. Place plate in thermal cycler and run the following program (lid set to 105°C): Step 1 (Protein Digestion): 50°C for 10 min Step 2 (Protease Inactivation): 75°C for 10 min 6. Place plate in thermal cycler and run the following program (lid set to 105°C): Step 1 (Reverse Transcription): 55°C for 10 min Step 2 (Enzyme Inactivation): 80°C for 10 min 6. Place plate in thermal cycler and run the following program (lid set to 105°C): Step 1 (Taq Activation): 98°C for 3 min Step 2 (Denaturation): 98°C for 20 s Step 3 (Annealing): 67°C for 15 s Step 4 (Extension): 72°C for 6 min (Repeat steps 2-4 for a total of 21 cycles*) Step 5 (Final Extension): 72°C for 5 min Step 6: 12°C hold *Cycle number may need to be optimized, as samples with high starting amounts of RNA will require less cycles, and those with low starting amounts will require more. 7. It's safe to stop and store PCR reactions at -20°C for at least 3 months. for best practices when performing DNA purification using magnetic beads). 11. Let plate stand in magnetic stand for 5 min, or until the bead pellet looks dry. If cracks begin to appear in the pellet, proceed immediately to next step.     3. Before analyzing the digital gene expression table, additional quality evaluations and data cleaning are necessary. First, ERCCs should be used to evaluate the quality of library preparation and sequencing: outliers with little to no ERCCs sequenced should be discarded.
Second, somatic or dendritic samples with a number of RNA molecules comparable to that of empty cuts, should also be discarded. Finally, as described below, unsupervised dimensionality  (Table 1